[llvm-dev] [LLVM 10][Windows/MSVC] llvm.eh.sjlj.setjmp does not lower to _setjmp() call any more

[Reid Kleckner via llvm-dev]

The timeline appears to coincide with Windows EH support for ARM64:
https://releases.llvm.org/8.0.0/tools/clang/docs/ReleaseNotes.html#windows-support


I looked at the logs, but it's not immediately obvious when the behavior
changed. If you want to dig deeper, it would be straightforward to bisect
this behavior change to a specific revision, and the commit message should
be explanatory.

On Fri, May 8, 2020 at 2:35 PM Alexander Mitin via llvm-dev <
llvm-dev at lists.llvm.org> wrote:

> Hello,
>
> In our project we use llvm.eh.sjlj.setjmp intrinsic to establish a
> recovering chain which implies that it's 'longjmp' counterpart would
> be called from a C code, so it uses a runtime implementation of
> longjmp().
> For LLVM 7 llvm.eh.sjlj.setjmp lowers to a call to Windows CRT
> function _setjmp(). After migrating to LLVM 10 we've found that it's
> no longer true.
> Is it done intentionally? If yes, why? The doc for llvm.eh.sjlj.setjmp
> says that this intrinsic is compatible with the GCC __builtin_setjmp
> implementation, so why isn't it compatible with MSVC CRT (in case of
> using MSVC CRT)?
>
> --
> Alexander Mitin,
> Instantiations, Inc.
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200511/03f1decc/attachment.html>