Page MenuHomePhabricator

phosek (Petr Hosek)
User

Projects

User does not belong to any projects.

User Details

User Since
Mar 4 2015, 5:44 PM (359 w, 22 h)

Recent Activity

Today

phosek accepted D117537: [cmake] Duplicate `{llvm,compiler_rt}_check_linker_flag` for runtime libs and llvm.

@phosek I agree with your fixing what language to do each flag, but do note I was purposely matching the existing dubious categorizations. Might it be better to first land this as-is, and then make a separate patch to clean up the languages?

Thu, Jan 20, 12:32 PM · Restricted Project, Restricted Project, Restricted Project, Restricted Project, Restricted Project
phosek accepted D117815: [compiler-rt][cmake] Use HandleOutOfTreeLLVM like libcxx and friends.

LGTM

Thu, Jan 20, 12:32 PM · Restricted Project
phosek accepted D117589: [Support] [DebugInfo] Lazily create cache dir..

LGTM

Thu, Jan 20, 12:51 AM · Restricted Project

Yesterday

phosek added inline comments to D117537: [cmake] Duplicate `{llvm,compiler_rt}_check_linker_flag` for runtime libs and llvm.
Wed, Jan 19, 11:56 PM · Restricted Project, Restricted Project, Restricted Project, Restricted Project, Restricted Project
phosek accepted D117752: [MLGO] Improved support for AOT cross-targeting scenarios.

LGTM

Wed, Jan 19, 11:46 PM · Restricted Project
phosek accepted D117639: [cmake] Make include(GNUInstallDirs) always below project(..).

LGTM

Wed, Jan 19, 7:48 PM · Restricted Project, Restricted Project, Restricted Project, Restricted Project, Restricted Project, Restricted Project, Restricted Project, Restricted Project
phosek updated subscribers of D117611: [Sema] Warn about printf %n on Android and Fuchsia.

@srhines can you also take a look at the Android changes or suggest other reviewers?

Wed, Jan 19, 1:20 PM
phosek accepted D117617: [cmake] Move HandleOutOfTreeLLVM to common cmake utils.

LGTM we should see if we can also use HandleOutOfTreeLLVM.cmake in compiler-rt in a follow-up change.

Wed, Jan 19, 1:19 PM · Restricted Project, Restricted Project, Restricted Project
phosek added inline comments to D117589: [Support] [DebugInfo] Lazily create cache dir..
Wed, Jan 19, 10:46 AM · Restricted Project
phosek committed rGfd0782a37bbf: [DebugInfod] Prefix debuginfod cache directory (authored by phosek).
[DebugInfod] Prefix debuginfod cache directory
Wed, Jan 19, 10:09 AM
phosek closed D117619: [DebugInfod] Prefix debuginfod cache directory..
Wed, Jan 19, 10:09 AM · Restricted Project

Tue, Jan 18

phosek added a comment to D116179: [InstrProf][NFC] Do not assume size of counter type.

Is it really necessary to rename the fields in InstrProfData.inc? That header is shipped with Clang and may have other uses outside of compiler-rt. For example, in Fuchsia we use InstrProfData.inc in the kernel where we cannot use the existing runtime and the name change broke our build. The original names may not be perfect, but that shouldn't justify breaking source compatibility.

I've created D117631 to restore InstrProfData.inc and hopefully fix your build. In retrospect I guess this wasn't NFC because this file is being used elsewhere, good to know.

Tue, Jan 18, 6:03 PM · Restricted Project, Restricted Project
phosek accepted D117631: [InstrProf] Restore InstrProfData.inc to fix Fuchsia builds.

LGTM thanks!

Tue, Jan 18, 5:57 PM · Restricted Project, Restricted Project
phosek added a comment to D116179: [InstrProf][NFC] Do not assume size of counter type.

Is it really necessary to rename the fields in InstrProfData.inc? That header is shipped with Clang and may have other uses outside of compiler-rt. For example, in Fuchsia we use InstrProfData.inc in the kernel where we cannot use the existing runtime and the name change broke our build. The original names may not be perfect, but that shouldn't justify breaking source compatibility.

Tue, Jan 18, 4:59 PM · Restricted Project, Restricted Project
phosek added a comment to D117619: [DebugInfod] Prefix debuginfod cache directory..

Note: I don't yet have commit access; I'll shoot of an email about it later this afternoon.

Tue, Jan 18, 4:12 PM · Restricted Project
phosek accepted D117619: [DebugInfod] Prefix debuginfod cache directory..

LGTM

Tue, Jan 18, 4:04 PM · Restricted Project
phosek added a comment to D117537: [cmake] Duplicate `{llvm,compiler_rt}_check_linker_flag` for runtime libs and llvm.

I'd prefer going in a different direction, that is figure out how to deduplicate llvm_check_linker_flag rather than duplicating it under a different name.

Tue, Jan 18, 9:48 AM · Restricted Project, Restricted Project, Restricted Project, Restricted Project, Restricted Project

Thu, Jan 13

phosek added a comment to D117263: [CMake] Support runtimes targets without specifying triple.

Another point I'd also add is that using CMAKE_{ASM,C,CXX}_COMPILER_TARGET as we do today isn't always sufficient. We ran into this recently on Windows. The problem is that CMake only processes CMAKE_{ASM,C,CXX}_COMPILER_TARGET after it does compiler detection, but in the case of Clang, --target= option changes the compiler behavior which can throw CMake compiler detection off. Specifically on Windows, CMake would detect Clang as "clang-cl in GNU compatibility mode" and to workaround it, we had to start passing --target= manually via CMAKE_{ASM,C,CXX}_FLAGS so the target triple handling runtimes build does right now looses some of its value. I'd more inclined to just deprecate it and require users to set those flags manually, but I'm not sure if that's feasible at this point since it'd be a breaking change.

Thu, Jan 13, 6:50 PM · Restricted Project
phosek added a comment to D117263: [CMake] Support runtimes targets without specifying triple.

I don't understand the motivation for this change.

The Darwin compiler-rt support works by overriding the OSX_ARCHITECTURES target property (which makes CMake pass -arch flags), and emptying CMAKE_OSX_DEPLOYMENT_TARGET and passing its own -m*-version-min flags. The Clang driver seems to have some special support where if your triple is x86_64-apple-darwin (I assume other architectures would work too, but I haven't checked), it'll override the architecture and the deployment target based on the -arch and -m*-version-min flags, which means the -target is effectively a no-op in this case (besides enabling the overwriting based on those other flags). The result is that you can add x86_64-apple-darwin to your LLVM_BUILTIN_TARGETS and LLVM_RUNTIME_TARGETS, and then control the actual platforms and architectures via the appropriate CMake variables for that target (COMPILER_RT_ENABLE_<PLATFORM>, DARWIN_<PLATFORM>_BUILTIN_ARCHS, etc.), which is what I set up in D86313, and what we use internally. Does that not work for your use case?

Thu, Jan 13, 6:39 PM · Restricted Project
phosek added a comment to D116976: [CMake] Passthrough OSX CMake options to builtins and runtimes.

Actually, could you give an example of the type of variables you want to pass in? I thought the CMAKE_OSX_* options were mostly for setting OS and architecture, which the Darwin compiler-rt build does itself anyway.

Thu, Jan 13, 6:33 PM · Restricted Project
phosek added a comment to D116976: [CMake] Passthrough OSX CMake options to builtins and runtimes.

Ping?

Thu, Jan 13, 5:36 PM · Restricted Project
phosek updated the summary of D117263: [CMake] Support runtimes targets without specifying triple.
Thu, Jan 13, 4:53 PM · Restricted Project
phosek requested review of D117263: [CMake] Support runtimes targets without specifying triple.
Thu, Jan 13, 4:53 PM · Restricted Project
phosek accepted D116844: [Driver][Fuchsia] -r: imply -nostdlib like GCC.

Thank you! I was going to send the same change but you beat me to it.

Thu, Jan 13, 3:40 PM · Restricted Project
phosek accepted D116843: [Driver][Gnu] -r: imply -nostdlib like GCC.

LGTM

Thu, Jan 13, 1:52 AM · Restricted Project

Tue, Jan 11

phosek added a comment to rG9e634b35ff51: [libc++] Temporarily disable the in_out_result test on Fuchsia..

I think this change should be reverted, we aren't yet running libc++ tests on Fuchsia so this change doesn't take any effect.

Tue, Jan 11, 10:54 PM
phosek added a comment to D116023: [libc++][ranges] Implement `uninitialized_copy{,_n}` and `uninitialized_move{,_n}`..

@vvereschaka, @uabelho Thanks for the error messages! I'll prepare a fix ASAP.

Tue, Jan 11, 10:51 PM · Restricted Project
phosek accepted D117058: [ifs] Added missing DT_STRSZ to the .dynamic section.

LGTM

Tue, Jan 11, 8:22 PM · Restricted Project
phosek accepted D117052: [llvm] Fix typo for libxml2 detection.

LGTM

Tue, Jan 11, 2:06 PM · Restricted Project
phosek added a reviewer for D116988: [libc++][libc++abi][libunwind] Dedup install path var definitions: ldionne.
Tue, Jan 11, 12:23 AM · Restricted Project, Restricted Project, Restricted Project
phosek accepted D116988: [libc++][libc++abi][libunwind] Dedup install path var definitions.

LGTM

Tue, Jan 11, 12:23 AM · Restricted Project, Restricted Project, Restricted Project

Mon, Jan 10

phosek requested review of D116976: [CMake] Passthrough OSX CMake options to builtins and runtimes.
Mon, Jan 10, 2:55 PM · Restricted Project
phosek accepted D116524: [doc][cmake] Convert read-me for the common CMake utils to reST.
Mon, Jan 10, 1:33 AM · Restricted Project

Fri, Jan 7

phosek reopened D116524: [doc][cmake] Convert read-me for the common CMake utils to reST.

This looks reasonable, thank you for writing it. The only thing I'm not sure about is the format. We use reStructuredText for most documentation in LLVM although we do have a handful of Markdown documents in the tree. Personally, I'm fine with either format but others may have stronger opinion on this.

Fri, Jan 7, 12:45 PM · Restricted Project
phosek added a reviewer for D116524: [doc][cmake] Convert read-me for the common CMake utils to reST: smeenai.
Fri, Jan 7, 12:43 PM · Restricted Project

Thu, Jan 6

phosek added a comment to D115852: Passthrough LLVM_USE_LLD LLVM_USE_LINKER into runtimes.

I don't think this is correct when cross-compiling. For example, if you're cross-compiling for Windows on a Linux host and using gold as your Linux host linker, it doesn't make sense to also use gold for your Windows links. PASSTHROUGH_PREFIXES is also annoying because it's added last in the external project invocation (https://github.com/apple/llvm-project/blob/7ca7a2547f00e34f5ec91be776a1d0bbca74b7a9/llvm/cmake/modules/LLVMExternalProjectUtils.cmake#L308-L309), so there's no way to override it. Is there an alternative solution that would work here while not breaking the cross-compilation case?

Thu, Jan 6, 1:45 PM · Restricted Project

Wed, Jan 5

phosek committed rG9eeb2c98f452: [msan] Check for AVX regs using offset (authored by phosek).
[msan] Check for AVX regs using offset
Wed, Jan 5, 7:26 PM
phosek closed D116695: [msan] Check for AVX regs using offset.
Wed, Jan 5, 7:25 PM · Restricted Project
phosek added a comment to D116695: [msan] Check for AVX regs using offset.

Add Missing Header

Wed, Jan 5, 6:12 PM · Restricted Project
phosek added inline comments to D116695: [msan] Check for AVX regs using offset.
Wed, Jan 5, 6:11 PM · Restricted Project
phosek accepted D116695: [msan] Check for AVX regs using offset.

LGTM

Wed, Jan 5, 2:36 PM · Restricted Project
phosek accepted D116503: [clang] Add --start-no-unused-arguments/--end-no-unused-arguments to silence some unused argument warnings.

LGTM this is something we have a use case for this in Fuchsia as well.

Wed, Jan 5, 1:05 PM · Restricted Project
phosek requested review of D116689: [libunwind][libcxxabi] Use object libraries in the build.
Wed, Jan 5, 12:37 PM · Restricted Project, Restricted Project, Restricted Project
phosek committed rG96e36048fddf: [CMake] Move the AIX archiver settings to a module (authored by phosek).
[CMake] Move the AIX archiver settings to a module
Wed, Jan 5, 11:47 AM
phosek closed D115276: [CMake] Move the AIX archiver settings to a module.
Wed, Jan 5, 11:46 AM · Restricted Project
phosek updated the diff for D115276: [CMake] Move the AIX archiver settings to a module.
Wed, Jan 5, 11:46 AM · Restricted Project
phosek added a comment to D115276: [CMake] Move the AIX archiver settings to a module.

Renamed the module to SetPlatformToolchainTools.cmake as suggested.

Wed, Jan 5, 11:18 AM · Restricted Project
phosek updated the diff for D115276: [CMake] Move the AIX archiver settings to a module.
Wed, Jan 5, 11:18 AM · Restricted Project
phosek accepted D115815: [CMake] Support passing arguments to build tool for external projects.

LGTM

Wed, Jan 5, 10:24 AM · Restricted Project

Tue, Jan 4

phosek accepted D116050: [libc++] Disable coverage with sanitize-coverage=0.

LGTM

Tue, Jan 4, 2:18 PM · Restricted Project, Restricted Project, Restricted Project, Restricted Project
phosek added a comment to D115847: [mlgo][inline] Improve global state tracking.

This looks reasonable although I'm not familiar with this part of the codebase so I'd like for @aeubanks to take a look as well.

Tue, Jan 4, 11:31 AM · Restricted Project
phosek accepted D116606: [libcxx] Use Fuchsia-native monotonic clock for std::chrono::steady_clock.

LGTM

Tue, Jan 4, 10:29 AM · Restricted Project
phosek added a comment to D116498: [libcxx] Use Fuchsia-native CPRNG for std::random_device.

LGTM

Tue, Jan 4, 10:27 AM · Restricted Project

Tue, Dec 28

phosek added a comment to rG21dbfd430021: [ELF] --gc-sections: Change startwith(".init") (and ".fini") to exact match.

Perhaps fundamentally it's because I do not trust that you root caused to startswith(".jcr") or the == ".init" change.
Hope that you can show evidence that matching startswith(".init") really helps your situation.

Tue, Dec 28, 3:50 PM
phosek added a comment to rG21dbfd430021: [ELF] --gc-sections: Change startwith(".init") (and ".fini") to exact match.

Hi, this change as well as 7c0881a38f5e0f50375a089e5a3c9b2d215be0e9 broke Fuchsia rust build and we are seeing crashes from some build tools. Both of your changes are not NFC but they landed without proper review. Could you revert them now so we can restore the build please?

No.

Rust has broken usage https://github.com/rust-lang/rust/issues/92181 and I have worked around it in 48161b7490e4aa174faf2b0e840e743a3ee56a4a

Tue, Dec 28, 3:17 PM

Dec 16 2021

phosek accepted D115852: Passthrough LLVM_USE_LLD LLVM_USE_LINKER into runtimes.

LGTM

Dec 16 2021, 2:45 PM · Restricted Project

Dec 15 2021

phosek accepted D115814: [Debuginfod] Remove f-string from llvm-debuginfod-find lit test..

LGTM

Dec 15 2021, 11:53 PM · Restricted Project
phosek accepted D113462: [llvm] Fix small typos in Programmer's Manual and Chrono.h..

LGTM

Dec 15 2021, 11:52 PM · Restricted Project
phosek added inline comments to D115852: Passthrough LLVM_USE_LLD LLVM_USE_LINKER into runtimes.
Dec 15 2021, 11:24 PM · Restricted Project

Dec 13 2021

phosek added inline comments to D115568: [lld][CMake] Use `GNUInstallDirs` to support custom installation dirs.
Dec 13 2021, 11:44 PM · lld, Restricted Project
phosek updated the diff for D115674: [runtime] Build compiler-rt with --unwindlib=none.
Dec 13 2021, 1:57 PM · Restricted Project
phosek requested review of D115674: [runtime] Build compiler-rt with --unwindlib=none.
Dec 13 2021, 1:46 PM · Restricted Project
phosek accepted D115659: [lld] Replace Symbolize.h with DIContext.h in lld's COFF lib.

LGTM

Dec 13 2021, 1:22 PM · Restricted Project
phosek added a reverting change for rGe286fefa8caa: [CMake] Update internal libc++ configuration: rGa65389849d61: Revert "[CMake] Update internal libc++ configuration".
Dec 13 2021, 11:14 AM
phosek committed rGa65389849d61: Revert "[CMake] Update internal libc++ configuration" (authored by phosek).
Revert "[CMake] Update internal libc++ configuration"
Dec 13 2021, 11:14 AM
phosek committed rGe286fefa8caa: [CMake] Update internal libc++ configuration (authored by phosek).
[CMake] Update internal libc++ configuration
Dec 13 2021, 11:08 AM
phosek accepted D115601: Simplify AddSphinxTarget.cmake by just including `GNUInstallDirs`.

LGTM

Dec 13 2021, 2:11 AM · Restricted Project
phosek added a comment to D115554: [gn build] Remove unnecessary lld dependency on DebugInfo/Symbolize..

Thanks for the review @phosek . Just to avoid any possible confusion I've also corrected SymbolTable.cpp to include "DIContext.h" instead of "Symbolize.h". Does this still LGTY?

Dec 13 2021, 2:07 AM · Restricted Project

Dec 10 2021

phosek accepted D115554: [gn build] Remove unnecessary lld dependency on DebugInfo/Symbolize..

LGTM

Dec 10 2021, 3:09 PM · Restricted Project

Dec 9 2021

phosek added a comment to D115131: [Support] [Debuginfod] Move HTTPClient to Debuginfod library..

Moving the dependency to another library doesn't help in the case of DLLVM_BUILD_LLVM_DYLIB=ON, because libLLVMDebuginfod is included in libLLVM.so. Are there still plans to make the libcurl dependency off by default?

Dec 9 2021, 11:55 PM · Restricted Project
phosek added a comment to D114782: [X86][clang] Emit diagnostic for float and double when we have features -x87 and -sse on 64-bits.

Hi @phosek , thanks for reporting it.
I would consider it as a misuse of the header file when SSE is not usable, especially we now have the specific header "x86gprintrin.h" for such scenarios. Can you help to try if changing to "x86gprintrin.h" works?

It's a normal practice to compile with -mno-{sse,avx} and then use SSE/AVX intrinsics explicitly inside runtime cpuid conditionals. The function calling the intrinsic might be marked with __attribute__((target("avx"))), but the source file's ambient -mo-{sse,avx} will be in force when #include <immintrin.h> is done. This change breaks that practice which is likely going to affect existing codebases.

Dec 9 2021, 5:28 PM · Restricted Project
phosek added a comment to D114782: [X86][clang] Emit diagnostic for float and double when we have features -x87 and -sse on 64-bits.

Hi @phosek , thanks for reporting it.
I would consider it as a misuse of the header file when SSE is not usable, especially we now have the specific header "x86gprintrin.h" for such scenarios. Can you help to try if changing to "x86gprintrin.h" works?

Dec 9 2021, 5:25 PM · Restricted Project
phosek added a comment to D115024: [ifs] Patch llvm-ifs to allow output multiple types of stub file at the same time.

LGTM

Dec 9 2021, 4:52 PM · Restricted Project
phosek added a comment to D114782: [X86][clang] Emit diagnostic for float and double when we have features -x87 and -sse on 64-bits.

Note that GCC had the same issue in the past https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80298 which was later addressed.

Dec 9 2021, 4:47 PM · Restricted Project
phosek added a comment to D114782: [X86][clang] Emit diagnostic for float and double when we have features -x87 and -sse on 64-bits.

This change broke our kernel build. I believe that the issue is that the compiler doesn't consider function attributes.

Dec 9 2021, 4:41 PM · Restricted Project
phosek added a comment to rG458ead66dc37: [compiler-rt] Use the runtimes build for custom libc++.

Can this bot be broken by this, or related patches?
https://lab.llvm.org/buildbot/#/builders/19

Dec 9 2021, 3:31 PM
phosek added a comment to D114498: [sanitizer] Run Stack compression in background thread.

Windows bot accepted the updated revision. I don't know where to track Fuchsia.

Dec 9 2021, 12:17 PM · Restricted Project
phosek added a comment to D113253: [runtimes] Fix building initial libunwind+libcxxabi+libcxx with compiler implied -lunwind.

I managed to reproduce the issue. @phosek, can you advise how to proceed from here?

One part of the issue, is that setting CMAKE_TRY_COMPILE_TARGET_TYPE to STATIC_LIBRARY makes the initial detection step detect less things. In particular, if doing full link tests, cmake sets the CMAKE_LIBRARY_ARCHITECTURE variable to x86_64-linux-gnu. This seems to be vital for some parts of library detection. To show off the issue, one can execute this minimal standalone cmake snippet:

set(CMAKE_TRY_COMPILE_TARGET_TYPE STATIC_LIBRARY)
project(Runtimes)
cmake_minimum_required(VERSION 3.10)
find_package(ZLIB)

On Ubuntu, it fails to find zlib, but if the set(CMAKE_TRY_COMPILE_TARGET_TYPE STATIC_LIBRARY) line is uncommented, it finds it as expected.

Dec 9 2021, 11:11 AM · Restricted Project, Restricted Project
phosek added inline comments to D115024: [ifs] Patch llvm-ifs to allow output multiple types of stub file at the same time.
Dec 9 2021, 10:36 AM · Restricted Project
phosek accepted D115024: [ifs] Patch llvm-ifs to allow output multiple types of stub file at the same time.

LGTM

Dec 9 2021, 10:19 AM · Restricted Project
phosek updated subscribers of D103947: [libcxx] Fix using the vcruntime ABI with _HAS_EXCEPTIONS=0 defined.
Dec 9 2021, 10:06 AM · Restricted Project
phosek added a comment to D114498: [sanitizer] Run Stack compression in background thread.

We started seeing build failure on our builders after this change landed:

ld.lld: error: undefined symbol: __sanitizer::internal_start_thread(void* (*)(void*), void*)
>>> referenced by sanitizer_stackdepot.cpp:133 (compiler-rt/lib/sanitizer_common/sanitizer_stackdepot.cpp:133)
>>>               compiler-rt/lib/sanitizer_common/CMakeFiles/RTSanitizerCommonSymbolizer.x86_64.dir/sanitizer_stackdepot.cpp.obj:(__sanitizer::(anonymous namespace)::CompressThread::NewWorkNotify())

Would it be possible to take a look and revert the change if it cannot be resolved quickly?

+1, this is also visible on https://lab.llvm.org/buildbot/#/builders/127.

I see an attempt to revert something to fix this in rG02997febe61597b7c3f215534c5adce851ba8963 by @browneee, but that seems to have reverted the wrong commit.

Dec 9 2021, 9:58 AM · Restricted Project
phosek added a reverting change for D114498: [sanitizer] Run Stack compression in background thread: rGa1e7f6280104: Revert "[sanitizer] Run Stack compression in background thread".
Dec 9 2021, 9:57 AM · Restricted Project
phosek added a reverting change for rGe5c2a46c5e8f: [sanitizer] Run Stack compression in background thread: rGa1e7f6280104: Revert "[sanitizer] Run Stack compression in background thread".
Dec 9 2021, 9:57 AM
phosek committed rGa1e7f6280104: Revert "[sanitizer] Run Stack compression in background thread" (authored by phosek).
Revert "[sanitizer] Run Stack compression in background thread"
Dec 9 2021, 9:57 AM
phosek added a comment to D114498: [sanitizer] Run Stack compression in background thread.

We started seeing build failure on our builders after this change landed:

ld.lld: error: undefined symbol: __sanitizer::internal_start_thread(void* (*)(void*), void*)
>>> referenced by sanitizer_stackdepot.cpp:133 (compiler-rt/lib/sanitizer_common/sanitizer_stackdepot.cpp:133)
>>>               compiler-rt/lib/sanitizer_common/CMakeFiles/RTSanitizerCommonSymbolizer.x86_64.dir/sanitizer_stackdepot.cpp.obj:(__sanitizer::(anonymous namespace)::CompressThread::NewWorkNotify())

Would it be possible to take a look and revert the change if it cannot be resolved quickly?

Dec 9 2021, 12:09 AM · Restricted Project

Dec 8 2021

phosek committed rG458ead66dc37: [compiler-rt] Use the runtimes build for custom libc++ (authored by phosek).
[compiler-rt] Use the runtimes build for custom libc++
Dec 8 2021, 12:53 PM
phosek added a comment to D112748: [release] Use the Bootstrapping build for building LLVM releases.

@phosek With this patch as-is, we end up not building the runtimes at all because install-runtimes isn't executed when we run ninja -C build. I *think* that's because install-runtimes is not added to the ALL target in CMake. Is that voluntary or is it an oversight?

That's because nobody ever asked for it. From what I know, all users of the bootstrapping build use the distribution support.

I'm fine including install-runtimes in the ALL target, but I also think that releases should ideally be using distribution support.

Oh, well, in that case, I think we should definitely be using the distribution support. @phosek any appetite for commandeering this to migrate test-release.sh to a proper distribution build?

Dec 8 2021, 9:29 AM · Restricted Project
phosek added a reverting change for rGbda3f2dd763b: [compiler-rt] Use the runtimes build for custom libc++: rG824ddeb994ab: Revert "[compiler-rt] Use the runtimes build for custom libc++".
Dec 8 2021, 9:14 AM
phosek committed rG824ddeb994ab: Revert "[compiler-rt] Use the runtimes build for custom libc++" (authored by phosek).
Revert "[compiler-rt] Use the runtimes build for custom libc++"
Dec 8 2021, 9:14 AM
phosek added a reverting change for D114922: [compiler-rt] Use the runtimes build for custom libc++: rG824ddeb994ab: Revert "[compiler-rt] Use the runtimes build for custom libc++".
Dec 8 2021, 9:14 AM · Restricted Project

Dec 7 2021

phosek committed rG40028eaf703a: [CMake] Specify byproducts for custom libcxx builds (authored by phosek).
[CMake] Specify byproducts for custom libcxx builds
Dec 7 2021, 11:15 PM
phosek committed rGbda3f2dd763b: [compiler-rt] Use the runtimes build for custom libc++ (authored by phosek).
[compiler-rt] Use the runtimes build for custom libc++
Dec 7 2021, 4:08 PM
phosek closed D114922: [compiler-rt] Use the runtimes build for custom libc++.
Dec 7 2021, 4:08 PM · Restricted Project
phosek added inline comments to D114922: [compiler-rt] Use the runtimes build for custom libc++.
Dec 7 2021, 4:07 PM · Restricted Project
phosek updated the diff for D114922: [compiler-rt] Use the runtimes build for custom libc++.
Dec 7 2021, 4:07 PM · Restricted Project
phosek added a comment to D115276: [CMake] Move the AIX archiver settings to a module.

I'm thinking about moving https://github.com/llvm/llvm-project/blob/cfe5d768be95ae0a62cf430e56e7762cebf81a65/llvm/cmake/modules/HandleLLVMOptions.cmake#L174 and https://github.com/llvm/llvm-project/blob/cfe5d768be95ae0a62cf430e56e7762cebf81a65/compiler-rt/cmake/Modules/UseLibtool.cmake into CheckPlatform.cmake since this is logic we want to reuse across projects. Let me know if you have a better suggestion for the name of this module.

Dec 7 2021, 12:21 PM · Restricted Project
phosek added a reviewer for D115276: [CMake] Move the AIX archiver settings to a module: ldionne.
Dec 7 2021, 12:17 PM · Restricted Project
phosek requested review of D115276: [CMake] Move the AIX archiver settings to a module.
Dec 7 2021, 12:17 PM · Restricted Project
phosek added a comment to D115045: [Clang] Ignore CLANG_DEFAULT_LINKER for custom-linker toolchains.

Rather, we shouldn't consider -DCLANG_DEFAULT_LINKER=lld as a valid value. Instead AMDGPU bot should use -DCLANG_DEFAULT_LINKER=ld.lld.

I think this would be a kinda disruptive change; I (and others) specify it as -DCLANG_DEFAULT_LINKER=lld so far: https://github.com/mstorsjo/llvm-mingw/blob/master/build-llvm.sh#L162 And other projects too: https://github.com/msys2/MINGW-packages/blob/master/mingw-w64-clang/PKGBUILD#L204

Dec 7 2021, 1:51 AM · Restricted Project