|
|
|
/*
|
|
|
|
* Copyright (C) The c-ares project
|
|
|
|
*
|
|
|
|
* Permission to use, copy, modify, and distribute this
|
|
|
|
* software and its documentation for any purpose and without
|
|
|
|
* fee is hereby granted, provided that the above copyright
|
|
|
|
* notice appear in all copies and that both that copyright
|
|
|
|
* notice and this permission notice appear in supporting
|
|
|
|
* documentation, and that the name of M.I.T. not be used in
|
|
|
|
* advertising or publicity pertaining to distribution of the
|
|
|
|
* software without specific, written prior permission.
|
|
|
|
* M.I.T. makes no representations about the suitability of
|
|
|
|
* this software for any purpose. It is provided "as is"
|
|
|
|
* without express or implied warranty.
|
|
|
|
*
|
|
|
|
* SPDX-License-Identifier: MIT
|
|
|
|
*/
|
|
|
|
#include "ares-test.h"
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
extern "C" {
|
|
|
|
#include "ares_private.h"
|
|
|
|
}
|
|
|
|
|
|
|
|
// library initialization is only needed for windows builds
|
|
|
|
#ifdef WIN32
|
|
|
|
#define EXPECTED_NONINIT ARES_ENOTINITIALIZED
|
|
|
|
#else
|
|
|
|
#define EXPECTED_NONINIT ARES_SUCCESS
|
|
|
|
#endif
|
|
|
|
|
|
|
|
namespace ares {
|
|
|
|
namespace test {
|
|
|
|
|
|
|
|
TEST(LibraryInit, Basic) {
|
|
|
|
EXPECT_EQ(EXPECTED_NONINIT, ares_library_initialized());
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_library_init(ARES_LIB_INIT_ALL));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_library_initialized());
|
|
|
|
ares_library_cleanup();
|
|
|
|
EXPECT_EQ(EXPECTED_NONINIT, ares_library_initialized());
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(LibraryInit, UnexpectedCleanup) {
|
|
|
|
EXPECT_EQ(EXPECTED_NONINIT, ares_library_initialized());
|
|
|
|
ares_library_cleanup();
|
|
|
|
EXPECT_EQ(EXPECTED_NONINIT, ares_library_initialized());
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(LibraryInit, DISABLED_InvalidParam) {
|
|
|
|
// TODO: police flags argument to ares_library_init()
|
|
|
|
EXPECT_EQ(ARES_EBADQUERY, ares_library_init(ARES_LIB_INIT_ALL << 2));
|
|
|
|
EXPECT_EQ(EXPECTED_NONINIT, ares_library_initialized());
|
|
|
|
ares_library_cleanup();
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(LibraryInit, Nested) {
|
|
|
|
EXPECT_EQ(EXPECTED_NONINIT, ares_library_initialized());
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_library_init(ARES_LIB_INIT_ALL));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_library_initialized());
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_library_init(ARES_LIB_INIT_ALL));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_library_initialized());
|
|
|
|
ares_library_cleanup();
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_library_initialized());
|
|
|
|
ares_library_cleanup();
|
|
|
|
EXPECT_EQ(EXPECTED_NONINIT, ares_library_initialized());
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(LibraryInit, BasicChannelInit) {
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_library_init(ARES_LIB_INIT_ALL));
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
EXPECT_NE(nullptr, channel);
|
|
|
|
ares_destroy(channel);
|
|
|
|
ares_library_cleanup();
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(LibraryTest, OptionsChannelInit) {
|
|
|
|
struct ares_options opts;
|
|
|
|
int optmask = 0;
|
|
|
|
memset(&opts, 0, sizeof(opts));
|
|
|
|
opts.flags = ARES_FLAG_USEVC | ARES_FLAG_PRIMARY;
|
|
|
|
optmask |= ARES_OPT_FLAGS;
|
|
|
|
opts.timeout = 2000;
|
|
|
|
optmask |= ARES_OPT_TIMEOUTMS;
|
|
|
|
opts.tries = 2;
|
|
|
|
optmask |= ARES_OPT_TRIES;
|
|
|
|
opts.ndots = 4;
|
|
|
|
optmask |= ARES_OPT_NDOTS;
|
|
|
|
opts.udp_port = 54;
|
|
|
|
optmask |= ARES_OPT_MAXTIMEOUTMS;
|
|
|
|
opts.maxtimeout = 10000;
|
|
|
|
optmask |= ARES_OPT_UDP_PORT;
|
|
|
|
opts.tcp_port = 54;
|
|
|
|
optmask |= ARES_OPT_TCP_PORT;
|
|
|
|
opts.socket_send_buffer_size = 514;
|
|
|
|
optmask |= ARES_OPT_SOCK_SNDBUF;
|
|
|
|
opts.socket_receive_buffer_size = 514;
|
|
|
|
optmask |= ARES_OPT_SOCK_RCVBUF;
|
|
|
|
opts.ednspsz = 1280;
|
|
|
|
optmask |= ARES_OPT_EDNSPSZ;
|
|
|
|
opts.nservers = 2;
|
|
|
|
opts.servers = (struct in_addr *)malloc((size_t)opts.nservers * sizeof(struct in_addr));
|
|
|
|
opts.servers[0].s_addr = htonl(0x01020304);
|
|
|
|
opts.servers[1].s_addr = htonl(0x02030405);
|
|
|
|
optmask |= ARES_OPT_SERVERS;
|
|
|
|
opts.ndomains = 2;
|
|
|
|
opts.domains = (char **)malloc((size_t)opts.ndomains * sizeof(char *));
|
|
|
|
opts.domains[0] = strdup("example.com");
|
|
|
|
opts.domains[1] = strdup("example2.com");
|
|
|
|
optmask |= ARES_OPT_DOMAINS;
|
|
|
|
opts.lookups = strdup("b");
|
|
|
|
optmask |= ARES_OPT_LOOKUPS;
|
|
|
|
optmask |= ARES_OPT_ROTATE;
|
|
|
|
opts.resolvconf_path = strdup("/etc/resolv.conf");
|
|
|
|
optmask |= ARES_OPT_RESOLVCONF;
|
|
|
|
opts.hosts_path = strdup("/etc/hosts");
|
|
|
|
optmask |= ARES_OPT_HOSTS_FILE;
|
|
|
|
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init_options(&channel, &opts, optmask));
|
|
|
|
EXPECT_NE(nullptr, channel);
|
|
|
|
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel2 = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_dup(&channel2, channel));
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_NE(nullptr, channel2);
|
|
|
|
|
|
|
|
struct ares_options opts2;
|
|
|
|
int optmask2 = 0;
|
|
|
|
memset(&opts2, 0, sizeof(opts2));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_save_options(channel2, &opts2, &optmask2));
|
|
|
|
|
|
|
|
// Note that not all opts-settable fields are saved (e.g.
|
|
|
|
// ednspsz, socket_{send,receive}_buffer_size).
|
|
|
|
EXPECT_EQ(opts.flags, opts2.flags);
|
|
|
|
EXPECT_EQ(opts.timeout, opts2.timeout);
|
|
|
|
EXPECT_EQ(opts.tries, opts2.tries);
|
|
|
|
EXPECT_EQ(opts.ndots, opts2.ndots);
|
|
|
|
EXPECT_EQ(opts.maxtimeout, opts2.maxtimeout);
|
|
|
|
EXPECT_EQ(opts.udp_port, opts2.udp_port);
|
|
|
|
EXPECT_EQ(opts.tcp_port, opts2.tcp_port);
|
|
|
|
EXPECT_EQ(1, opts2.nservers); // Truncated by ARES_FLAG_PRIMARY
|
|
|
|
EXPECT_EQ(opts.servers[0].s_addr, opts2.servers[0].s_addr);
|
|
|
|
EXPECT_EQ(opts.ndomains, opts2.ndomains);
|
|
|
|
EXPECT_EQ(std::string(opts.domains[0]), std::string(opts2.domains[0]));
|
|
|
|
EXPECT_EQ(std::string(opts.domains[1]), std::string(opts2.domains[1]));
|
|
|
|
EXPECT_EQ(std::string(opts.lookups), std::string(opts2.lookups));
|
|
|
|
EXPECT_EQ(std::string(opts.resolvconf_path), std::string(opts2.resolvconf_path));
|
|
|
|
EXPECT_EQ(std::string(opts.hosts_path), std::string(opts2.hosts_path));
|
|
|
|
|
|
|
|
ares_destroy_options(&opts);
|
|
|
|
ares_destroy_options(&opts2);
|
|
|
|
ares_destroy(channel);
|
|
|
|
ares_destroy(channel2);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(LibraryTest, ChannelAllocFail) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel;
|
|
|
|
for (int ii = 1; ii <= 25; ii++) {
|
|
|
|
ClearFails();
|
|
|
|
SetAllocFail(ii);
|
|
|
|
channel = nullptr;
|
|
|
|
int rc = ares_init(&channel);
|
|
|
|
// The number of allocations depends on local environment, so don't expect ENOMEM.
|
|
|
|
if (rc == ARES_ENOMEM) {
|
|
|
|
EXPECT_EQ(nullptr, channel);
|
|
|
|
} else {
|
|
|
|
ares_destroy(channel);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(LibraryTest, OptionsChannelAllocFail) {
|
|
|
|
struct ares_options opts;
|
|
|
|
int optmask = 0;
|
|
|
|
memset(&opts, 0, sizeof(opts));
|
|
|
|
opts.flags = ARES_FLAG_USEVC;
|
|
|
|
optmask |= ARES_OPT_FLAGS;
|
|
|
|
opts.timeout = 2;
|
|
|
|
optmask |= ARES_OPT_TIMEOUT;
|
|
|
|
opts.tries = 2;
|
|
|
|
optmask |= ARES_OPT_TRIES;
|
|
|
|
opts.ndots = 4;
|
|
|
|
optmask |= ARES_OPT_NDOTS;
|
|
|
|
opts.udp_port = 54;
|
|
|
|
optmask |= ARES_OPT_UDP_PORT;
|
|
|
|
opts.tcp_port = 54;
|
|
|
|
optmask |= ARES_OPT_TCP_PORT;
|
|
|
|
opts.socket_send_buffer_size = 514;
|
|
|
|
optmask |= ARES_OPT_SOCK_SNDBUF;
|
|
|
|
opts.socket_receive_buffer_size = 514;
|
|
|
|
optmask |= ARES_OPT_SOCK_RCVBUF;
|
|
|
|
opts.ednspsz = 1280;
|
|
|
|
optmask |= ARES_OPT_EDNSPSZ;
|
|
|
|
opts.nservers = 2;
|
|
|
|
opts.servers = (struct in_addr *)malloc((size_t)opts.nservers * sizeof(struct in_addr));
|
|
|
|
opts.servers[0].s_addr = htonl(0x01020304);
|
|
|
|
opts.servers[1].s_addr = htonl(0x02030405);
|
|
|
|
optmask |= ARES_OPT_SERVERS;
|
|
|
|
opts.ndomains = 2;
|
|
|
|
opts.domains = (char **)malloc((size_t)opts.ndomains * sizeof(char *));
|
|
|
|
opts.domains[0] = strdup("example.com");
|
|
|
|
opts.domains[1] = strdup("example2.com");
|
|
|
|
optmask |= ARES_OPT_DOMAINS;
|
|
|
|
opts.lookups = strdup("b");
|
|
|
|
optmask |= ARES_OPT_LOOKUPS;
|
|
|
|
optmask |= ARES_OPT_ROTATE;
|
|
|
|
opts.resolvconf_path = strdup("/etc/resolv.conf");
|
|
|
|
optmask |= ARES_OPT_RESOLVCONF;
|
|
|
|
opts.hosts_path = strdup("/etc/hosts");
|
|
|
|
optmask |= ARES_OPT_HOSTS_FILE;
|
|
|
|
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
for (int ii = 1; ii <= 8; ii++) {
|
|
|
|
ClearFails();
|
|
|
|
SetAllocFail(ii);
|
|
|
|
int rc = ares_init_options(&channel, &opts, optmask);
|
|
|
|
if (rc == ARES_ENOMEM) {
|
|
|
|
EXPECT_EQ(nullptr, channel);
|
|
|
|
} else {
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, rc);
|
|
|
|
ares_destroy(channel);
|
|
|
|
channel = nullptr;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
ClearFails();
|
|
|
|
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init_options(&channel, &opts, optmask));
|
|
|
|
EXPECT_NE(nullptr, channel);
|
|
|
|
|
|
|
|
// Add some servers and a sortlist for flavour.
|
|
|
|
EXPECT_EQ(ARES_SUCCESS,
|
|
|
|
ares_set_servers_csv(channel, "1.2.3.4,0102:0304:0506:0708:0910:1112:1314:1516,2.3.4.5"));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel, "1.2.3.4 2.3.4.5"));
|
|
|
|
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel2 = nullptr;
|
|
|
|
for (int ii = 1; ii <= 18; ii++) {
|
|
|
|
ClearFails();
|
|
|
|
SetAllocFail(ii);
|
|
|
|
EXPECT_EQ(ARES_ENOMEM, ares_dup(&channel2, channel)) << ii;
|
|
|
|
EXPECT_EQ(nullptr, channel2) << ii;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct ares_options opts2;
|
|
|
|
int optmask2 = 0;
|
|
|
|
for (int ii = 1; ii <= 6; ii++) {
|
|
|
|
memset(&opts2, 0, sizeof(opts2));
|
|
|
|
ClearFails();
|
|
|
|
SetAllocFail(ii);
|
|
|
|
EXPECT_EQ(ARES_ENOMEM, ares_save_options(channel, &opts2, &optmask2)) << ii;
|
|
|
|
// May still have allocations even after ARES_ENOMEM return code.
|
|
|
|
ares_destroy_options(&opts2);
|
|
|
|
}
|
|
|
|
ares_destroy_options(&opts);
|
|
|
|
ares_destroy(channel);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(LibraryTest, FailChannelInit) {
|
|
|
|
EXPECT_EQ(ARES_SUCCESS,
|
|
|
|
ares_library_init_mem(ARES_LIB_INIT_ALL,
|
|
|
|
&LibraryTest::amalloc,
|
|
|
|
&LibraryTest::afree,
|
|
|
|
&LibraryTest::arealloc));
|
|
|
|
SetAllocFail(1);
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_ENOMEM, ares_init(&channel));
|
|
|
|
EXPECT_EQ(nullptr, channel);
|
|
|
|
ares_library_cleanup();
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifndef WIN32
|
|
|
|
TEST_F(LibraryTest, EnvInit) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EnvValue v1("LOCALDOMAIN", "this.is.local");
|
|
|
|
EnvValue v2("RES_OPTIONS", "options debug ndots:3 retry:3 rotate retrans:2");
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
ares_destroy(channel);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(LibraryTest, EnvInitModernOptions) {
|
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EnvValue v1("LOCALDOMAIN", "this.is.local");
|
|
|
|
EnvValue v2("RES_OPTIONS", "options debug retrans:2 ndots:3 attempts:4 timeout:5 rotate");
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
|
|
|
|
channel->optmask |= ARES_OPT_TRIES;
|
|
|
|
channel->optmask |= ARES_OPT_TIMEOUTMS;
|
|
|
|
|
|
|
|
struct ares_options opts;
|
|
|
|
memset(&opts, 0, sizeof(opts));
|
|
|
|
int optmask = 0;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_save_options(channel, &opts, &optmask));
|
|
|
|
EXPECT_EQ(5000, opts.timeout);
|
|
|
|
EXPECT_EQ(4, opts.tries);
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(LibraryTest, EnvInitAllocFail) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel;
|
|
|
|
EnvValue v1("LOCALDOMAIN", "this.is.local");
|
|
|
|
EnvValue v2("RES_OPTIONS", "options debug ndots:3 retry:3 rotate retrans:2");
|
|
|
|
for (int ii = 1; ii <= 10; ii++) {
|
|
|
|
ClearFails();
|
|
|
|
SetAllocFail(ii);
|
|
|
|
channel = nullptr;
|
|
|
|
int rc = ares_init(&channel);
|
|
|
|
if (rc == ARES_SUCCESS) {
|
|
|
|
ares_destroy(channel);
|
|
|
|
} else {
|
|
|
|
EXPECT_EQ(ARES_ENOMEM, rc);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
TEST_F(DefaultChannelTest, SetAddresses) {
|
|
|
|
ares_set_local_ip4(channel_, 0x01020304);
|
|
|
|
byte addr6[16] = {0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08,
|
|
|
|
0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x10};
|
|
|
|
ares_set_local_ip6(channel_, addr6);
|
|
|
|
ares_set_local_dev(channel_, "dummy");
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(DefaultChannelTest, SetSortlistFailures) {
|
|
|
|
EXPECT_EQ(ARES_ENODATA, ares_set_sortlist(nullptr, "1.2.3.4"));
|
|
|
|
EXPECT_EQ(ARES_EBADSTR, ares_set_sortlist(channel_, "111.111.111.111*/16"));
|
|
|
|
EXPECT_EQ(ARES_EBADSTR, ares_set_sortlist(channel_, "111.111.111.111/255.255.255.240*"));
|
|
|
|
EXPECT_EQ(ARES_EBADSTR, ares_set_sortlist(channel_, "1 0123456789012345"));
|
|
|
|
EXPECT_EQ(ARES_EBADSTR, ares_set_sortlist(channel_, "1 /01234567890123456789012345678901"));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel_, "xyzzy ; lwk"));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel_, "xyzzy ; 0x123"));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(DefaultChannelTest, SetSortlistVariants) {
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel_, "1.2.3.4"));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel_, "1.2.3.4 ; 2.3.4.5"));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel_, "129.1.1.1"));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel_, "192.1.1.1"));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel_, "224.1.1.1"));
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_set_sortlist(channel_, "225.1.1.1"));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(DefaultChannelTest, SetSortlistAllocFail) {
|
|
|
|
for (int ii = 1; ii <= 3; ii++) {
|
|
|
|
ClearFails();
|
|
|
|
SetAllocFail(ii);
|
|
|
|
EXPECT_EQ(ARES_ENOMEM, ares_set_sortlist(channel_, "12.13.0.0/16 1234::5678/40 1.2.3.4")) << ii;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef USE_WINSOCK
|
|
|
|
TEST(Init, NoLibraryInit) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_ENOTINITIALIZED, ares_init(&channel));
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef HAVE_CONTAINER
|
|
|
|
// These tests rely on the ability of non-root users to create a chroot
|
|
|
|
// using Linux namespaces.
|
|
|
|
|
|
|
|
|
|
|
|
// The library uses a variety of information sources to initialize a channel,
|
|
|
|
// in particular to determine:
|
|
|
|
// - search: the search domains to use
|
|
|
|
// - servers: the name servers to use
|
|
|
|
// - lookup: whether to check files or DNS or both (e.g. "fb")
|
|
|
|
// - options: various resolver options
|
|
|
|
// - sortlist: the order of preference for IP addresses
|
|
|
|
//
|
|
|
|
// The first source from the following list is used:
|
|
|
|
// - init_by_options(): explicitly specified values in struct ares_options
|
|
|
|
// - init_by_environment(): values from the environment:
|
|
|
|
// - LOCALDOMAIN -> search (single value)
|
|
|
|
// - RES_OPTIONS -> options
|
|
|
|
// - init_by_resolv_conf(): values from various config files:
|
|
|
|
// - /etc/resolv.conf -> search, lookup, servers, sortlist, options
|
|
|
|
// - /etc/nsswitch.conf -> lookup
|
|
|
|
// - /etc/host.conf -> lookup
|
|
|
|
// - /etc/svc.conf -> lookup
|
|
|
|
// - init_by_defaults(): fallback values:
|
|
|
|
// - gethostname(3) -> domain
|
|
|
|
// - "fb" -> lookup
|
|
|
|
|
|
|
|
NameContentList filelist = {
|
|
|
|
{"/etc/resolv.conf", "nameserver 1.2.3.4\n"
|
|
|
|
"sortlist 1.2.3.4/16 2.3.4.5\n"
|
|
|
|
"search first.com second.com\n"},
|
|
|
|
{"/etc/hosts", "3.4.5.6 ahostname.com\n"},
|
|
|
|
{"/etc/nsswitch.conf", "hosts: files\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerChannelInit,
|
|
|
|
"myhostname", "mydomainname.org", filelist) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
std::vector<std::string> actual = GetNameServers(channel);
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
std::vector<std::string> expected = {"1.2.3.4:53"};
|
|
|
|
EXPECT_EQ(expected, actual);
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(2, channel->ndomains);
|
|
|
|
EXPECT_EQ(std::string("first.com"), std::string(channel->domains[0]));
|
|
|
|
EXPECT_EQ(std::string("second.com"), std::string(channel->domains[1]));
|
|
|
|
|
|
|
|
HostResult result;
|
|
|
|
ares_gethostbyname(channel, "ahostname.com", AF_INET, HostCallback, &result);
|
|
|
|
ProcessWork(channel, NoExtraFDs, nullptr);
|
|
|
|
EXPECT_TRUE(result.done_);
|
|
|
|
std::stringstream ss;
|
|
|
|
ss << result.host_;
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
|
|
|
|
EXPECT_EQ("{'ahostname.com' aliases=[] addrs=[3.4.5.6]}", ss.str());
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerSortlistOptionInit,
|
|
|
|
"myhostname", "mydomainname.org", filelist) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
struct ares_options opts = {0};
|
|
|
|
int optmask = 0;
|
|
|
|
optmask |= ARES_OPT_SORTLIST;
|
|
|
|
opts.nsort = 0;
|
|
|
|
// Explicitly specifying an empty sortlist in the options should override the
|
|
|
|
// environment.
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init_options(&channel, &opts, optmask));
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(0, channel->nsort);
|
|
|
|
EXPECT_EQ(nullptr, channel->sortlist);
|
|
|
|
EXPECT_EQ(ARES_OPT_SORTLIST, (channel->optmask & ARES_OPT_SORTLIST));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
NameContentList fullresolv = {
|
|
|
|
{"/etc/resolv.conf", " nameserver 1.2.3.4 \n"
|
|
|
|
"search first.com second.com\n"
|
|
|
|
"lookup bind\n"
|
|
|
|
"options debug ndots:5\n"
|
|
|
|
"sortlist 1.2.3.4/16 2.3.4.5\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerFullResolvInit,
|
|
|
|
"myhostname", "mydomainname.org", fullresolv) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(std::string("b"), std::string(channel->lookups));
|
|
|
|
EXPECT_EQ(5, channel->ndots);
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Allow path for resolv.conf to be configurable
|
|
|
|
NameContentList myresolvconf = {
|
|
|
|
{"/tmp/myresolv.cnf", " nameserver 1.2.3.4 \n"
|
|
|
|
"search first.com second.com\n"
|
|
|
|
"lookup bind\n"
|
|
|
|
"options debug ndots:5\n"
|
|
|
|
"sortlist 1.2.3.4/16 2.3.4.5\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerMyResolvConfInit,
|
|
|
|
"myhostname", "mydomain.org", myresolvconf) {
|
|
|
|
char filename[] = "/tmp/myresolv.cnf";
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
struct ares_options options = {0};
|
|
|
|
options.resolvconf_path = strdup(filename);
|
|
|
|
int optmask = ARES_OPT_RESOLVCONF;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init_options(&channel, &options, optmask));
|
|
|
|
|
|
|
|
optmask = 0;
|
|
|
|
free(options.resolvconf_path);
|
|
|
|
options.resolvconf_path = NULL;
|
|
|
|
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_save_options(channel, &options, &optmask));
|
|
|
|
EXPECT_EQ(ARES_OPT_RESOLVCONF, (optmask & ARES_OPT_RESOLVCONF));
|
|
|
|
EXPECT_EQ(std::string(filename), std::string(options.resolvconf_path));
|
|
|
|
|
|
|
|
ares_destroy_options(&options);
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Allow hosts path to be configurable
|
|
|
|
NameContentList myhosts = {
|
|
|
|
{"/tmp/hosts", "10.0.12.26 foobar\n"
|
|
|
|
"2001:A0:C::1A foobar\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerMyHostsInit,
|
|
|
|
"myhostname", "mydomain.org", myhosts) {
|
|
|
|
char filename[] = "/tmp/hosts";
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
struct ares_options options = {0};
|
|
|
|
options.hosts_path = strdup(filename);
|
|
|
|
int optmask = ARES_OPT_HOSTS_FILE;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init_options(&channel, &options, optmask));
|
|
|
|
|
|
|
|
optmask = 0;
|
|
|
|
free(options.hosts_path);
|
|
|
|
options.hosts_path = NULL;
|
|
|
|
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_save_options(channel, &options, &optmask));
|
|
|
|
EXPECT_EQ(ARES_OPT_HOSTS_FILE, (optmask & ARES_OPT_HOSTS_FILE));
|
|
|
|
EXPECT_EQ(std::string(filename), std::string(options.hosts_path));
|
|
|
|
|
|
|
|
ares_destroy_options(&options);
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
NameContentList hostconf = {
|
|
|
|
{"/etc/resolv.conf", "nameserver 1.2.3.4\n"
|
|
|
|
"sortlist1.2.3.4\n" // malformed line
|
|
|
|
"search first.com second.com\n"},
|
|
|
|
{"/etc/host.conf", "order bind hosts\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerHostConfInit,
|
|
|
|
"myhostname", "mydomainname.org", hostconf) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(std::string("bf"), std::string(channel->lookups));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
NameContentList svcconf = {
|
|
|
|
{"/etc/resolv.conf", "nameserver 1.2.3.4\n"
|
|
|
|
"search first.com second.com\n"},
|
|
|
|
{"/etc/svc.conf", "hosts= bind\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerSvcConfInit,
|
|
|
|
"myhostname", "mydomainname.org", svcconf) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(std::string("b"), std::string(channel->lookups));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
NameContentList malformedresolvconflookup = {
|
|
|
|
{"/etc/resolv.conf", "nameserver 1.2.3.4\n"
|
|
|
|
"lookup garbage\n"}}; // malformed line
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerMalformedResolvConfLookup,
|
|
|
|
"myhostname", "mydomainname.org", malformedresolvconflookup) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(std::string("fb"), std::string(channel->lookups));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Failures when expected config filenames are inaccessible.
|
|
|
|
class MakeUnreadable {
|
|
|
|
public:
|
|
|
|
explicit MakeUnreadable(const std::string& filename)
|
|
|
|
: filename_(filename) {
|
|
|
|
chmod(filename_.c_str(), 0000);
|
|
|
|
}
|
|
|
|
~MakeUnreadable() { chmod(filename_.c_str(), 0644); }
|
|
|
|
private:
|
|
|
|
std::string filename_;
|
|
|
|
};
|
|
|
|
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerResolvConfNotReadable,
|
|
|
|
"myhostname", "mydomainname.org", filelist) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
MakeUnreadable hide("/etc/resolv.conf");
|
|
|
|
// Unavailable /etc/resolv.conf falls back to defaults
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerNsswitchConfNotReadable,
|
|
|
|
"myhostname", "mydomainname.org", filelist) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
// Unavailable /etc/nsswitch.conf falls back to defaults.
|
|
|
|
MakeUnreadable hide("/etc/nsswitch.conf");
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(std::string("fb"), std::string(channel->lookups));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerHostConfNotReadable,
|
|
|
|
"myhostname", "mydomainname.org", hostconf) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
// Unavailable /etc/host.conf falls back to defaults.
|
|
|
|
MakeUnreadable hide("/etc/host.conf");
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerSvcConfNotReadable,
|
|
|
|
"myhostname", "mydomainname.org", svcconf) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
// Unavailable /etc/svc.conf falls back to defaults.
|
|
|
|
MakeUnreadable hide("/etc/svc.conf");
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
NameContentList rotateenv = {
|
|
|
|
{"/etc/resolv.conf", "nameserver 1.2.3.4\n"
|
|
|
|
"search first.com second.com\n"
|
|
|
|
"options rotate\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerRotateInit,
|
|
|
|
"myhostname", "mydomainname.org", rotateenv) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(ARES_TRUE, channel->rotate);
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerRotateOverride,
|
|
|
|
"myhostname", "mydomainname.org", rotateenv) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
struct ares_options opts = {0};
|
|
|
|
int optmask = ARES_OPT_NOROTATE;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init_options(&channel, &opts, optmask));
|
|
|
|
optmask = 0;
|
|
|
|
ares_save_options(channel, &opts, &optmask);
|
|
|
|
EXPECT_EQ(ARES_OPT_NOROTATE, (optmask & ARES_OPT_NOROTATE));
|
|
|
|
ares_destroy_options(&opts);
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Test that blacklisted IPv6 resolves are ignored. They're filtered from any
|
|
|
|
// source, so resolv.conf is as good as any.
|
|
|
|
NameContentList blacklistedIpv6 = {
|
|
|
|
{"/etc/resolv.conf", " nameserver 254.192.1.1\n" // 0xfe.0xc0.0x01.0x01
|
|
|
|
" nameserver fec0::dead\n" // Blacklisted
|
|
|
|
" nameserver ffc0::c001\n" // Not blacklisted
|
|
|
|
" domain first.com\n"},
|
|
|
|
{"/etc/nsswitch.conf", "hosts: files\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerBlacklistedIpv6,
|
|
|
|
"myhostname", "mydomainname.org", blacklistedIpv6) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
std::vector<std::string> actual = GetNameServers(channel);
|
|
|
|
std::vector<std::string> expected = {
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
"254.192.1.1:53",
|
|
|
|
"[ffc0:0000:0000:0000:0000:0000:0000:c001]:53"
|
|
|
|
};
|
|
|
|
EXPECT_EQ(expected, actual);
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(1, channel->ndomains);
|
|
|
|
EXPECT_EQ(std::string("first.com"), std::string(channel->domains[0]));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
NameContentList multiresolv = {
|
|
|
|
{"/etc/resolv.conf", " nameserver 1::2 ; ;;\n"
|
|
|
|
" domain first.com\n"},
|
|
|
|
{"/etc/nsswitch.conf", "hosts: files\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerMultiResolvInit,
|
|
|
|
"myhostname", "mydomainname.org", multiresolv) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
std::vector<std::string> actual = GetNameServers(channel);
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
std::vector<std::string> expected = {"[0001:0000:0000:0000:0000:0000:0000:0002]:53"};
|
|
|
|
EXPECT_EQ(expected, actual);
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(1, channel->ndomains);
|
|
|
|
EXPECT_EQ(std::string("first.com"), std::string(channel->domains[0]));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
NameContentList systemdresolv = {
|
|
|
|
{"/etc/resolv.conf", "nameserver 1.2.3.4\n"
|
|
|
|
"domain first.com\n"},
|
|
|
|
{"/etc/nsswitch.conf", "hosts: junk resolve files\n"}};
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerSystemdResolvInit,
|
|
|
|
"myhostname", "mydomainname.org", systemdresolv) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(std::string("bf"), std::string(channel->lookups));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
NameContentList empty = {}; // no files
|
|
|
|
CONTAINED_TEST_F(LibraryTest, ContainerEmptyInit,
|
|
|
|
"host.domain.org", "domain.org", empty) {
|
`ares_channel` -> `ares_channel_t *`: don't bury the pointer (#595)
`ares_channel` is defined as `typedef struct ares_channeldata *ares_channel;`. The problem with this, is it embeds the pointer into the typedef, which means an `ares_channel` can never be declared as `const` as if you write `const ares_channel channel`, that expands to `struct ares_channeldata * const ares_channel` and not `const struct ares_channeldata *channel`.
We will now typedef `ares_channel_t` as `typedef struct ares_channeldata ares_channel_t;`, so if you write `const ares_channel_t *channel`, it properly expands to `const struct ares_channeldata *channel`.
We are maintaining the old typedef for API compatibility with existing integrations, and due to typedef expansion this should not even cause any compiler warnings for existing code. There are no ABI implications with this change. I could be convinced to keep existing public functions as `ares_channel` if a sufficient argument exists, but internally we really need make this change for modern best practices.
This change will allow us to internally use `const ares_channel_t *` where appropriate. Whether or not we decide to change any public interfaces to use `const` may require further discussion on if there might be ABI implications (I don't think so, but I'm also not 100% sure what a compiler internally does with `const` when emitting machine code ... I think more likely ABI implications would occur going the opposite direction).
FYI, This PR was done via a combination of sed and clang-format, the only manual code change was the addition of the new typedef, and a couple doc fixes :)
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
ares_channel_t *channel = nullptr;
|
|
|
|
EXPECT_EQ(ARES_SUCCESS, ares_init(&channel));
|
|
|
|
std::vector<std::string> actual = GetNameServers(channel);
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
std::vector<std::string> expected = {"127.0.0.1:53"};
|
|
|
|
EXPECT_EQ(expected, actual);
|
|
|
|
|
Dynamic Server List (#594)
This PR makes the server list a dynamic sorted list of servers. The sort order is [ consecutive failures, system config index ]. The server list can be updated via ares_set_servers_*(). Any queries currently directed to servers that are no longer in the list will be automatically re-queued to a different server.
Also, any time a failure occurs on the server, the sort order of the servers will be updated so that the one with the fewest consecutive failures is chosen for the next query that goes on the wire, this way bad or non-responsive servers are automatically isolated.
Since the server list is now dynamic, the tracking of query failures per server has been removed and instead is relying on the server sort order as previously described. This simplifies the logic while also reducing the amount of memory required per query. However, because of this dynamic nature, it may not be easy to determine the server attempt order for enqueued queries if there have been any failures.
If using the ARES_OPT_ROTATE, this is now implemented to be a random selection of the configured servers. Since the server list is dynamic, its not possible to go to the next server as configuration could have changed between queries or attempts for the same query.
Finally, this PR moved some existing functions into new files to logically separate them.
This should address issues #550 and #440, while also setting the framework to implement #301. #301 needs a little more effort since it configures things other than the servers themselves (domains, search, sortlist, lookups), which need to make sure they can be safely updated.
Fix By: Brad House (@bradh352)
1 year ago
|
|
|
EXPECT_EQ(1, channel->ndomains);
|
|
|
|
EXPECT_EQ(std::string("domain.org"), std::string(channel->domains[0]));
|
|
|
|
EXPECT_EQ(std::string("fb"), std::string(channel->lookups));
|
|
|
|
|
|
|
|
ares_destroy(channel);
|
|
|
|
return HasFailure();
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
} // namespace test
|
|
|
|
} // namespace ares
|