As bazel folks are looking at getting auto generation of module maps going and
the importing of sources files causes issues there. We were only do it to
hack around some of the apple linker behaviors around objc classes and
categories, but even that isn't complete and CocoaPods was already doing -ObjC,
and developers not using pods could have still needed it to ensure everything
was linked anyways; so drop the hack of importing sources.
Add the `nullable` qualifier to return types of Objective-C methods that
can return a nil due to errors. This change makes these methods
compatible with the Swift 2 try-catch syntax.
- Correct some cases sources were compiled into the static lib and the tests.
- Enable Xcodes code coverage support on the unittests. We aren't complete on
coverage, but having the data always there should make it easier to chip away
at this going forward.
- Drop method in tests that isn't used, wire up a validator in another test.
There was a twist code path (that some times showed up due to what happened to
be in memory in failure cases), that would cast a bogus wire type into the
enum, and then fall through switch statements.
Resolve this by validating all wire types when parsing tags and throwing the
error at that point so it can't enter the system.
As added safety, stick in a few asserts for apis that get passed tags to ensure
they also are only seeing valid data.
Bonus: Tweak the parsing loop to skip some work when we get the end marker
(zero tag) instead of still looping through all the fields.
Note: Breaking API change on the Dictionary classes.
The numeric value classes were using "Value" in the naming, but this silently
collided with the KVC category on NSObject; meaning KVC code could break up a
keypath and call these selectors with the wrong types leading to crashes (even
though the code all would compile cleanly).
- Rename the methods to use the "type" instead of literal "Value".
- Update all the impls and tests.
- Enable the warning that will catch issues like this in the future.
Fixes https://github.com/google/protobuf/issues/1616
- Better docs in the generator for the different options that can be passed
during an invoke of protoc.
- Add named_framework_to_proto_path_mappings_path to pass the path to a file
containing mappings of frameworks for different proto files.
- Update the generation to use the mapping to change the #import directives
it creates.
Note: the changes in helpers is mostly moving code within the fine, and then
a small change to expose the parsing so a passed on class can consume the line.
Fixes https://github.com/google/protobuf/issues/1457
- Add a protoc objc option (generate_for_named_framework) to set the name of
the framework all generated sources will be in.
- Tweak some comments/naming to make it clear what is the Protobuf framework
vs. the framework for generated code.
- Update the objc README to document the new generation option to protoc.
This is working towards https://github.com/google/protobuf/issues/1457.
Add more context to GPBCodedInputStream failures.
Have GPBMessage parsing apis extract out the GPBCodedInputStream information and expose it.
Update HeaderDocs with pointers to all error domains/codes.
Expand the unittests to cover the full set of errors reported.
Fixes https://github.com/google/protobuf/issues/1618
Working on https://github.com/google/protobuf/issues/1599, specifically:
- Turn on more warnings that the Xcode UI calls out with individual controls.
- Manually add:
-Wundef
-Wswitch-enum
- Manually add and then diable in the unittests because of XCTest's headers:
-Wreserved-id-macro
-Wdocumentation-unknown-command
- Manually add -Wdirect-ivar-access, but disable it for the unittests and in
the library code (via #pragmas to suppress it). This is done so proto users
can enable the warning.
- Move the ObjC tests into the list and exclude them on linux, this will change
where in the order they start, since they are longer, it will have other
things run in parallel instead of them ending up last and taking the longest.
- Switch to the Xcode 7.3 image.
- Drop the use of xctool and stream line things through the full_mac_build.sh
script. This means we end up with only one build script instead of two.
- Tweaks to the mac build script:
- Make iOS Xcode version support explicit
- Support Debug/Release only building
- Change the OS X min parallel count to 2 to better deal with VMs.
- Split the travis ios tests into the two Xcode Configurations as the logs are
choking travis.
- Add generator constant for the default framework name.
- Add generator api for making the CPP symbol from the name.
- Add generator api to see if it is a bundled proto file.
- Output a CPP conditional and two imports for the core library headers.
- Add helper for generating the #import for file headers to deal with the
framework imports.
- Add a reference from the unittests to a WKT to use that to inspect how
imports generate.
- Update the podspec to define the CPP symbol and require pods 1.0 (or later).
Fixes https://github.com/google/protobuf/issues/1457
- Env solution doesn't seem to always work, use template pod files and copy
them in place instead.
- Flush the pods cache before and after runs.
- Make pod install verbose to have the info incase something goes wrong.
- Add an Xcode 6.3 created default iOS Project.
- Add an Xcode 6.3 created default OS X Project.
- Add Podfiles to for both that use Protobufs from within the tree.
- Add a script to run the tests (and cleanup) to help confirm the state of the
Protobuf.podspec and sources.
cp -r foo/ bar/ in linux will create a bar/foo directoy. In the
objectivec generate descritpor case, well known types will be created in
objectivec/google/google/protobuf/.. if the command is run under linux.
Adding the trailing period fixes the behavior inconsistency.
Both methods weren't checking the has_bits (where the bools are stored), so
it resulted in invalid results.
Add a test that should shake out something like this in the future also.
proto2 syntax allows the first enum to have a non zero value. This means any
field using that default has a non zero default without having an explicit
default being set. So when deciding what runtime info is needed, don't rely
on an explicit default, always check that the values aren't zero.
Fixes https://github.com/google/protobuf/issues/1453
- Always generated into a temp directory so we can see if things changed.
- Add a flag to control exiting with error when stale vs updating.
This should let the continuous builds error out when ObjC needs to have the
checked in sources updated.
This seems to be some code evolution side effects. Back when there was a custom
string class, we couldn't really error when we finally saw the string was bad
so we had to return the empty string, but now that full validation is done
up front, it can error out.
This will lower the amount of dispatch_semaphores created per Message when the
full object tree isn't walked in a way that would require them to be created.
Uses a dispatch_once_t for one time init of the dispatch_semaphore.
NOTE: This is a binary breaking change as structure sizes have changed size
and/or order.
- Drop capturing field options, no other options were captured and other mobile
targeted languages don't try to capture this sort information (saved 8
bytes for every field defined (in static data and again in field descriptor
instance size data).
- No longer generate/compile in the messages/enums in descriptor.proto. If
developers need it, they should generate it and compile it in. Reduced the
overhead of the core library.
- Compute the number of has_bits actually needs to avoid over reserving.
- Let the boolean single fields store via a has_bit to avoid storage, makes
the common cases of the instance size smaller.
- Reorder some flags and down size the enums to contain the bits needed.
- Reorder the items in the structures to manually ensure they are are packed
better (especially when generating 64bit code - 8 bytes for every field,
16 bytes for every extension, instance sizes 8 bytes also).
- Split off the structure initialization so when the default is zero, the
generated static storage doesn't need to reserve the space. This is batched
at the message level, so all the fields for the message have to have zero
defaults to get the saves. By definition all proto3 syntax files fall into
this case but it also saves space for the proto2 that use the standard
defaults. (saves 8 bytes of static data for every field that had a zero
default)
- Don't track the enums defined by a message. Nothing in the runtime needs it
and it was just generation and runtime overhead. (saves 8 bytes per enum)
- Ensure EnumDescriptors are started up threadsafe in all cases.
- Split some of the Descriptor initialization into multiple methods so the
generated code isn't padded with lots of zero/nil args.
- Change how oneof info is feed to the runtime enabling us to generate less
static data (8 bytes saved per oneof for 64bit).
- Change how enum value informat is capture to pack the data and only decode
it if it ends up being needed. Avoids padding issues causing bloat of 64bit,
and removes the needs for extra pointers in addition to the data (just the
data and one pointer now).
- Convert most of the core library headers over to HeaderDoc format.
- Switch the generated comments over to HeaderDoc.
- Create GPBCodedOutputStream_PackagePrivate and move some things into there
that should be more internal.
- Extend GPB*ObjectDictionary to support generic syntax.
- Update the generator to output generics so the enclosed type is exposed for compiler checks.
- Use generics in a the public interfaces.
- Update the generated sources that are checked in.
Thomas Van Lenten
Dia Kharrat
Feng Xiao
Nathan Wong
Dimitris Koutsogiorgas
Sergio Campamá
Jon Wall
Jisi Liu
Geoffrey Wiseman
Dongjoon Hyun