The Undefined Behavior sanitizer flags one part of the unittests for this.
For default values for `bytes` we write a length on the front of a c-string
in the static data, apparently the compiler/linker doesn't always make this
4 byte aligned, so it get flagged for undefined/degraded performance. Avoid
this by using memcpy instead.
The generated code for enums needs atomics support, so generate the
import instead of relying on it via transitive imports. This will
make future changes to this likely likely to break generated code
and runtime support are mixed.
Followup to https://github.com/google/protobuf/pull/4184.
Followup to https://github.com/google/protobuf/pull/4184, keep the
import to not break any existing generated code that isn't regenerated
when they update to the newer protobuf code.
Stop having travis build the test matrix and make it explicit so extra
settings are only on the test that need the, this helps make the travis
UI a little cleaner (linux hosted stuff doesn't list an Xcode version),
as well as avoid unneeded requirements limiting the travis pool something
can run on.
Small improvements to objectivec/DevTools/full_mac_build.sh to avoid some
repeated scripting and hopefully make things more reliable work with Xcode
9.x.
Don't mark the iOS tests as able to fail, hopefully with the new settings
they are stable and won't give false signals.
Simplify the case for xcode versions and fixup xcode 9.[0-1].
Add runtime asserts (that can be disabled in release) that verify
that the types being get/set for messages using the C Api match
the type in the descriptor for the field being get/set.
Remove unnecessary retain/releases for GPBFieldDescriptors when
creating implementations with blocks. The descriptors do not need
to be memory managed because they are essentially global constructs
that are guaranteed to be around the lifetime of the app.
Remove unreferenced 'GPBMessageSignatureProtocol' class that is just taking up
space in the Objective C runtime information by hanging the protocol it needs to
declare off of GPBRootObject instead.
Small binary size reduction, but more importantly a removal of a class that
appears to be unused when statically analyzed.
- Xcode 8.3.3 bumped the simulators to 10.3.1, so 10.3 fails. Just use
the "latest" naming instead to avoid issues.
- Also switch to just testing an old 32bit build and new 64bit build as
there isn't any code in protobuf that should care about form facts.
(Looks like the iPad Pro simulator names aren't always constantly, on
some developer machines they are "iPad Pro (9.7 inch)" while on others
they are "iPad Pro (9.7-inch)"; so just avoid using them.)
This should never happen, but if someone is swizzling or do other
hooking of methods, anything is possible, so this seems slighty
safer than they returning NO.
- Fix up -copyWithZone: to not leave the two registries sharing
some of the storage by using -addExtensions:.
- Improve -addExtensions: to clone the sub dict when there is
nothing to merge into.
- A ExtensionRegistry unittests.
- Update project schemes to not have extra things in perf scheme.
- Ensure extensions resolution/wiring is happening directly on the
messageClass (incase someone is doing odd things our out classes).
- Make the extension message check match the other class checks in
for mergeFrom/isEqual/etc.
Apple recently updated the docs on dispatch_once to point out
that the storage for the dispatch_once_t must be static or global,
but not something that was ever used before as the implementation
doesn't use a memory barrier. So we drop the use and create the
semaphore when needed and use an atomic swap deal with any
threading races.
- Don't prune the extension registry as that can lead to failures when two
threads are racing.
- If adding the method fails, check and see if it already is bound to decide
the return result. Deals with threading races binding the methods.
Thomas Van Lenten
Jonathan Dierksen
kvukic
Feng Xiao
Adam Cozzette
dmaclach
Sergio Campama
xuwei-k
Sergio Campama
Jisi Liu
cclauss