This patch is meant to track bug fixes and other patches that we want to merge into 3.9.1.
The following patch should be merged into 3.9.1 r279955 - Fix pair::operator=(TupleLike&&) https://github.com/llvm-mirror/libcxx/commit/235d71f053d0a4160c67ec343712ae1223570c0d
This one too: http://llvm.org/viewvc/llvm-project?view=revision&revision=279871
Tom, I'm assigning to you, as a stable release manager, for 3.9.1.
@Renato Are you aware this is currently binned under libc++?
No, why?
Sometimes libunwind stuff gets put here. Rebinning to new bugs to make this generic to LLVM.
r280190 should be applied as well. Which fixes: - http://llvm.org/PR12298 - http://llvm.org/PR29123
"Don't reduce the width of vector mul if the target doesn't support SSE2." https://llvm.org/bugs/show_bug.cgi?id=30298 https://reviews.llvm.org/D24288 https://reviews.llvm.org/rL280837
In FreeBSD's copy of 3.9 we have already applied the following post-3.9.0 patches. These would all be nice to have in 3.9.1. r280837 (bug 30298): Don't reduce the width of vector mul if the target doesn't support SSE2. r280040 + r280041 (bug 19098): [PowerPC] Add support for -mlongcall r280188 (bug 26856): [PowerPC] Don't spill the frame pointer twice r280350 (bug 26761): Add ISD::EH_DWARF_CFA, simplify @llvm.eh.dwarf.cfa on Mips, fix on PowerPC r280705 (bug 26519): [PPC] Claim stack frame before storing into it, if no red zone is present r281632 (bug 29140): Silence false positive diagnostics regarding passing an object of enumeration type to va_start().
I'd like to nominate the fix for PR30199 (in r281957) to the 3.9.1 release.
(In reply to comment #9) > In FreeBSD's copy of 3.9 we have already applied the following post-3.9.0 > patches. These would all be nice to have in 3.9.1. > > r280837 (bug 30298): Don't reduce the width of vector mul if the target > doesn't support SSE2. Nadav has given his OK - shall I merge this into the 3.9 release branch?
(In reply to comment #11) > (In reply to comment #9) > > In FreeBSD's copy of 3.9 we have already applied the following post-3.9.0 > > patches. These would all be nice to have in 3.9.1. > > > > r280837 (bug 30298): Don't reduce the width of vector mul if the target > > doesn't support SSE2. > > Nadav has given his OK - shall I merge this into the 3.9 release branch? Yes, please merge it.
I'd like to propose https://llvm.org/bugs/show_bug.cgi?id=30596 (https://reviews.llvm.org/D25039) for merging to 3.9.1: [X86][AVX] Ensure broadcast loads respect dependencies. This was fixed at r283070: http://llvm.org/viewvc/llvm-project?view=revision&revision=283070 With a preliminary test update at r282613: http://llvm.org/viewvc/llvm-project?view=revision&revision=282613
I would like to nominate bug 30342. This is impacting rust, see https://github.com/rust-lang/rust/issues/36474
Please merge http://llvm.org/viewvc/llvm-project?rev=286129&view=rev Log: [OPENMP] Fixed codegen for __real/__imag expressions in atomic constructs.
This is a regression introduced in clang 3.9: https://llvm.org/bugs/show_bug.cgi?id=29087 which has a patch under review that, once accepted, should probably be part of some clang 3.9.x release: https://reviews.llvm.org/D23765
Hi Tom, This is an important one to get into 3.9.1: https://llvm.org/bugs/show_bug.cgi?id=31035 Or Clang up to 3.8 won't be able to compile compiler-rt 3.9.1 (as it can't compile 3.9.0 already). We'll get a commit to trunk as soon as possible, then I'll create a "merge bug" for the rev, but I'm assing that bug as a dep here to make sure we don't forget about it. cheers, --renato
I would like to nominate bug 29163, which hurts Rust pretty badly: https://github.com/rust-lang/rust/issues/36023
(In reply to comment #18) > I would like to nominate bug 29163, which hurts Rust pretty badly: > https://github.com/rust-lang/rust/issues/36023 Please, create a new bug called "Merge [commit] into 3.9.1" and make it a blocker to this one, depending on Bug #29163, and write on the body why is this important". This short description with a link to the github issue is enough. It seems we have documentation for most steps on our side, but this one above is not there. I'll add it.
(In reply to comment #13) > I'd like to propose https://llvm.org/bugs/show_bug.cgi?id=30596 > (https://reviews.llvm.org/D25039) for merging to 3.9.1: > > [X86][AVX] Ensure broadcast loads respect dependencies. > > This was fixed at r283070: > http://llvm.org/viewvc/llvm-project?view=revision&revision=283070 > > With a preliminary test update at r282613: > http://llvm.org/viewvc/llvm-project?view=revision&revision=282613 These have been merged.
(In reply to comment #14) > I would like to nominate bug 30342. This is impacting rust, see > https://github.com/rust-lang/rust/issues/36474 I've created a ticket to track this: https://llvm.org/bugs/show_bug.cgi?id=31139
(In reply to comment #15) > Please merge http://llvm.org/viewvc/llvm-project?rev=286129&view=rev > > Log: > [OPENMP] Fixed codegen for __real/__imag expressions in atomic constructs. I have opened https://llvm.org/bugs/show_bug.cgi?id=31140 to track this request.
(In reply to comment #17) > Hi Tom, > > This is an important one to get into 3.9.1: > > https://llvm.org/bugs/show_bug.cgi?id=31035 > > Or Clang up to 3.8 won't be able to compile compiler-rt 3.9.1 (as it can't > compile 3.9.0 already). > > We'll get a commit to trunk as soon as possible, then I'll create a "merge > bug" for the rev, but I'm assing that bug as a dep here to make sure we > don't forget about it. > > cheers, > --renato This has been merged.
(In reply to comment #18) > I would like to nominate bug 29163, which hurts Rust pretty badly: > https://github.com/rust-lang/rust/issues/36023 This is request is being tracked by: https://llvm.org/bugs/show_bug.cgi?id=31113