> That's a pretty old version. Does this implementation correspond to the > current version N3928? (In particular, the change (made in v5; the > version you reference is v4, and the current version is v6) that it's > guaranteed that leaving execution of the scope of a defer by a noreturn > function call, such as longjmp or exit, or a signal, does not execute the > deferred block, when previously it was implementation-defined whether this > happens. Note that the description of Example 3 hasn't been properly > updated for that; to be corrected in the ballot draft.)
the behaviour already matched that, so no change is needed afaict > > Since we allow mixed-language code to throw exceptions (e.g. C++) across C > code built with -fexceptions, it would be a good idea also to have test > coverage for how this feature interacts with exceptions. And I think the > GCC manual should document how the feature interacts with relevant GNU > extensions (the cleanup attribute - you already have a test using that - > and throwing exceptions across C code, at least). added a test and documentation, we do trigger deferred blocks on -fexception stack unwind, same as [[gnu::cleanup]] and clang's _Defer implementation
68990292A7A98C5E.asc
Description: application/pgp-keys
signature.asc
Description: PGP signature
