We've had reports of user-after-free type crashes in Windows cleanup
code for the Event Thread. In evaluating the code, it appeared there
were some memory leaks on per-connection handles that may have remained
open during shutdown, while trying to resolve that it became apparent
the methodology chosen may not have been the right one for interfacing
with the Windows AFD system as stability issues were seen during this
debugging process.
Since this system is completely undocumented, there was no clear
resolution path other than to switch to the *other* methodology which
involves directly opening `\Device\Afd`, rather than spawning a "peer
socket" to use to queue AFD operations.
The original methodology chosen more closely resembled what is employed
by [libuv](https://github.com/libuv/libuv) and given its widespread use
was the reason it was used. The new methodology more closely resembles
[wepoll](https://github.com/piscisaureus/wepoll).
Its not clear if there are any scalability or performance advantages or
disadvantages for either method. They both seem like different ways to
do the same thing, but this current way does seem more stable.
Fixes#798
Fix By: Brad House (@bradh352)
AppVeyor has gotten progressively slower over the last year or so. Move
some Windows builds to GitHub actions which is considerably faster.
Fix By: Brad House (@bradh352)
Completely rework the autotools build system, issues have cropped up due to the complexity and could cause issues on even semi-modern Linux systems (Ubuntu 20.04 for example).
Changes include:
Remove all curl/xc/cares m4 helper files, they go overboard on detections of functions and datatypes. Go back to more plain autoconf macros as they've come a long way over the years.
Use known systems and heuristics to determine datatypes for functions like send() and recv(), rather than the error prone detection which required thousands of permutations and might still get it wrong.
Remove unneeded configure arguments like --enable-debug or --enable-optimize, its more common for people to simply pass their own CFLAGS on the command line.
Only require CARES_STATICLIB definition on Windows static builds, its not necessary ever for other systems, even when hiding non-public symbols.
Remove some function and definition detections that were never used in c-ares
The test framework is now embedded into the toplevel configure system, there was no need to chain build the test system as it is never built externally to c-ares.
As a side-effect of the changes, a configure run completes in about 25% of the original time.
This has been tested on various Linux distributions (of varying age), FreeBSD, MacOS, Windows (via MSYS2 with Mingw), and Solaris10/11 (by @dfandrich), AIX 7.3 (by @dfandrich). It is not unlikely that this may have broken more esoteric or legacy systems, and we'll likely need to be ready to accept bug reports and patches, but it has removed over 10k lines of build system code. It is very likely any issues that crop up will add far fewer lines of code to fix such systems.
Fixes Bug: #670
Fix By: Brad House (@bradh352)
GoogleTest should be unbundled. Google changed their guidance a few years back and modern versions of google test cannot build the bundling code file.
This PR also updates to use C++14 as is required by modern GoogleTest versions.
Fixes Bug: #506
Fix By: Brad House (@bradh352)
Some environments may send router advertisements on a link setting their link-local (fe80::/10) address as a valid DNS server to the remote system. This will cause a DNS entry to be created like `fe80::1%iface`, since all link-local network interfaces are technically part of the same /10 subnet, it must be told what interface to send packets through explicitly if there are multiple physical interfaces.
This PR adds support for the %iface modifier when setting DNS servers via `/etc/resolv.conf` as well as via `ares_set_servers_csv()`.
For MacOS and iOS it is assumed that libresolve will set the `sin6_scope_id` and should be supported, but my test systems don't seem to read the Router Advertisement for RDNSS link-local. Specifying the link-local dns server on MacOS via adig has been tested and confirmed working.
For Windows, this is similar to MacOS in that the system doesn't seem to honor the RDNSS RA, but specifying manually has been tested to work.
At this point, Android support does not exist.
Fixes Bug #462
Supersedes PR #463
Fix By: Brad House (@bradh352) and Serhii Purik (@sergvpurik)
When building for UWP (WindowsStore), additional headers are needed and some functions are not available. This also adds AppVeyor CI/CD support to catch these issues in the future.
Fix By: Deal (@halx99) and Brad House (@bradh352)
AppVeyor was using Visual Studio 2015 along with old versions of MinGW. Update to the latest AppVeyor provides and also add an MSYS2 build test using MinGW which will use the bleeding edge version.
When researching #590 this also uncovered a bug in cmake not properly detecting if_indextoname() on windows. This has been corrected as well as the underlying issue reported in #590.
Fix By: Brad House (@bradh352) and Jonas Kvinge (@jonaski)
The acountry utility required a third party DNSBL service from nerd.dk in order to operate. That service has been offline for about a year and there is no other comparable service offering. We are keeping the code in the repository as an example, but no longer building it.
Fixes: #537
Fix By: Brad House (@bradh352)
All files have their licence and copyright information clearly
identifiable. If not in the file header, they are set separately in
.reuse/dep5.
All used license texts are provided in LICENSES/
Originally started by Daniel Stenberg (@bagder) with #123, this patch reorganizes the c-ares source tree to have a more modern layout. It also fixes out of tree builds for autotools, and automatically builds the tests if tests are enabled. All tests are passing which tests each of the supported build systems (autotools, cmake, nmake, mingw gmake). There may be some edge cases that will have to be caught later on for things I'm not aware of.
Fix By: Brad House (@bradh352)
The rc4 function iterates over a buffer of size buffer_len who's maximum
value is INT_MAX with a counter of type short that is not guaranteed to
have maximum size INT_MAX.
In circumstances where short is narrower than int and where buffer_len
is larger than the maximum value of a short, it may be possible to loop
infinitely as counter will overflow and never be greater than or equal
to buffer_len.
The solution is to make the comparison be between types of equal width.
This commit defines counter as an int.
Fix By: Fionn Fitzmaurice (@fionn)
Brad House
Daniel Stenberg
Fionn Fitzmaurice