valgrind: fix building with llvm by RossComputerGuy · Pull Request #329995 · NixOS/nixpkgs

RossComputerGuy · 2024-07-25T20:36:23Z

Description of changes

Discovered in https://github.com/RossComputerGuy/nixpkgs-llvm-ws/runs/27931320260

Things done

Add a 👍 reaction to pull requests you find important.

RossComputerGuy · 2024-07-25T20:46:44Z

New issue found with x86_64, works on aarch64 though:

valgrind-x86_64-unknown-linux-gnu-3.23.0> x86_64-unknown-linux-gnu-ld: error: relocation R_X86_64_64 cannot be used against local symbol; recompile with -fPIC
valgrind-x86_64-unknown-linux-gnu-3.23.0> >>> defined in int3-amd64.o
valgrind-x86_64-unknown-linux-gnu-3.23.0> >>> referenced by int3-amd64.c:34 (/build/valgrind-3.23.0/memcheck/tests/amd64-linux/int3-amd64.c:34)
valgrind-x86_64-unknown-linux-gnu-3.23.0> >>>               int3-amd64.o:(main)
valgrind-x86_64-unknown-linux-gnu-3.23.0> clang: error: linker command failed with exit code 1 (use -v to see invocation

RossComputerGuy · 2024-08-02T18:59:58Z

Now valgrind is failing because of this:

done
aarch64-unknown-linux-gnu-clang -m64 -O2 -g -Wall -Wmissing-prototypes -Wshadow -Wpointer-arith -Wstrict-prototypes -Wmissing-declarations -Wcast-align -Wcast-qual -Wwrite-strings -Wempty-body -Wformat -Wformat-security -Wignored-qualifiers -Wenum-conversion -finline-functions -fno-stack-protector -fno-strict-aliasing -fno-builtin -Wno-cast-align -Wno-self-assign -Wno-tautological-compare  -O -g -fno-omit-frame-pointer -fno-strict-aliasing -fpic -fno-builtin  -O2  -nodefaultlibs -shared -Wl,-z,interpose,-z,initfirst  -m64 -Wl,--whole-archive ../coregrind/libreplacemalloc_toolpreload-arm64-linux.a -Wl,--no-whole-archive  -o vgpreload_memcheck-arm64-linux.so vgpreload_memcheck_arm64_linux_so-mc_replace_strmem.o  
../coregrind/link_tool_exe_linux 0x58000000 aarch64-unknown-linux-gnu-clang     -o memcheck-arm64-linux  -m64 -O2 -g -Wall -Wmissing-prototypes -Wshadow -Wpointer-arith -Wstrict-prototypes -Wmissing-declarations -Wcast-align -Wcast-qual -Wwrite-strings -Wempty-body -Wformat -Wformat-security -Wignored-qualifiers -Wenum-conversion -finline-functions -fno-stack-protector -fno-strict-aliasing -fno-builtin -Wno-cast-align -Wno-self-assign -Wno-tautological-compare  -O2 -static -nodefaultlibs -nostartfiles -u _start  -m64 memcheck_arm64_linux-mc_leakcheck.o memcheck_arm64_linux-mc_malloc_wrappers.o memcheck_arm64_linux-mc_main.o memcheck_arm64_linux-mc_main_asm.o memcheck_arm64_linux-mc_translate.o memcheck_arm64_linux-mc_machine.o memcheck_arm64_linux-mc_errors.o ../coregrind/libcoregrind-arm64-linux.a ../VEX/libvex-arm64-linux.a -lgcc  ../coregrind/libgcc-sup-arm64-linux.a 
aarch64-unknown-linux-gnu-ld: error: undefined symbol: getauxval
>>> referenced by aarch64.c.o:(init_have_lse_atomics) in archive /nix/store/lfypx41m7m76fxmcazy4fvh9d6pqja48-compiler-rt-libc-aarch64-unknown-linux-gnu-18.1.8/lib/libgcc.a
>>> referenced by aarch64.c.o:(__init_cpu_features) in archive /nix/store/lfypx41m7m76fxmcazy4fvh9d6pqja48-compiler-rt-libc-aarch64-unknown-linux-gnu-18.1.8/lib/libgcc.a
>>> referenced by aarch64.c.o:(__init_cpu_features) in archive /nix/store/lfypx41m7m76fxmcazy4fvh9d6pqja48-compiler-rt-libc-aarch64-unknown-linux-gnu-18.1.8/lib/libgcc.a
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[3]: *** [Makefile:1155: memcheck-arm64-linux] Error 1

RossComputerGuy · 2024-08-02T19:34:43Z

This PR now builds again.

Ericson2314 · 2024-08-02T21:03:46Z

Sorry about the conflict with the other PR @RossComputerGuy.

do you think we could instead add a withSantizers ? haveLibc configuration flag to compiler-rt? I think that would be cleaner.

RossComputerGuy · 2024-08-02T21:05:08Z

do you think we could instead add a withSantizers ? haveLibc configuration flag to compiler-rt? I think that would be cleaner.

Yeah, having the ability to turn off sanitizers or having a llvmPackages.compiler-rt-no-libc-no-sanitizers drv might help here.

Ericson2314 · 2024-08-02T21:10:45Z

@RossComputerGuy Yeah I would prefer to always building the sanitizers in a separate derivation. Ideally that would not be another compiler-rt but just be the santizers, linking compiler-rt/libc/etc. as needed, but that may require some LLVM-side CMake work to do well.

RossComputerGuy · 2024-08-03T01:14:28Z

Ideally that would not be another compiler-rt but just be the santizers, linking compiler-rt/libc/etc. as needed,

How exactly would that look like? Would we have a $sanitized output which has lib with several libs that are just the sanitizer built versions of what's in the main output?

Ericson2314 · 2024-08-04T14:24:30Z

@RossComputerGuy See https://github.com/llvm/llvm-project/blob/0edafc461f5f98b2ed5d2d621e1d9de70ccbd4e5/compiler-rt/CMakeLists.txt#L42-L64 We can enable/disable things fairly independently. Perhaps easier than I thought.

(What I would like to advocate they do however is break the stuff into proper separate packages that can be built together with the stuff that libc++ libc++abi and libunwind use, rather than one mega-package that we can split up with enable/disable flags.)

RossComputerGuy · 2024-09-23T03:34:09Z

@Ericson2314 So based on your previous comments, is there anything blocking this PR or what should we do exactly? I would like to see this PR come in so Mesa can build under LLVM.

nixos-discourse · 2025-04-18T04:38:33Z

This pull request has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/prs-ready-for-review/3032/5395

RossComputerGuy · 2025-05-13T03:54:40Z

Updating this PR, for whatever reason resetting back to 52ee4b6141beeab91dea471b09ec4036465bbc73 doesn't work.

RossComputerGuy · 2025-05-13T03:56:11Z

Fixed now

RossComputerGuy · 2025-09-12T02:31:21Z

Rebased to hopefully fix CI

alyssais · 2025-09-13T12:05:18Z

pkgs/by-name/va/valgrind/package.nix

+            inherit (llvmPackages.compiler-rt-no-libc.stdenv.hostPlatform.parsed) cpu;
+          };
+          useLLVM = false;
+        };


Why do we need a custom hostPlatform? To disable sanitizers? Surely there's a better way to do that.

I think so, I don't remember since this PR comes from forever ago.

alyssais · 2025-09-13T12:05:57Z

pkgs/by-name/va/valgrind/package.nix

+      (
+        f: p: {
+          hardeningDisable = p.hardeningDisable or [ ] ++ [ "stackprotector" ];
+        }


If we have to disable stackprotector based on either the hostPlatform or doFakeLibgcc, that should happen centrally in the compiler-rt expression so it doesn't need to be repeated every time somebody wants to override one of those.

CC @RossComputerGuy can we get this going?

No, this is blocked by source-highlight and LLVM's unwind being broken. I haven't had time to jump into it. source-highlight will require a new version upstream since it needs a newer libtool and we cant inject one without using an already dropped autotools version.

alyssais · 2025-09-13T12:06:31Z

pkgs/by-name/va/valgrind/package.nix

  configureFlags = lib.optional stdenv.hostPlatform.isx86_64 "--enable-only64bit";

-  doCheck = true;
+  # Some tests fail on aarch64 and x86_64 under LLVM.


Why? For reasons that should be reported upstream, or not? Can we disable just those tests? And why are we listing platforms? Are there platforms it doesn't fail on with LLVM?

I remember some tests just breaking entirely. I tried looking on Bugzilla but I don't see anything reported for it. I'm trying to build things right now to rediscover what I did.

RossComputerGuy force-pushed the fix/pkgsllvm/valgrind branch from 5979f31 to 414a45d Compare July 25, 2024 20:53

ofborg bot added 10.rebuild-darwin: 0 This PR does not cause any packages to rebuild on Darwin. 10.rebuild-linux: 0 This PR does not cause any packages to rebuild on Linux. labels Jul 25, 2024

RossComputerGuy force-pushed the fix/pkgsllvm/valgrind branch 2 times, most recently from 64c07c2 to 69e7ad1 Compare July 31, 2024 05:57

RossComputerGuy mentioned this pull request Aug 2, 2024

llvm: Fix compiler-rt missing sanitizers when using useLLVM #297144

Merged

13 tasks

RossComputerGuy force-pushed the fix/pkgsllvm/valgrind branch from 69e7ad1 to 64b61ce Compare August 2, 2024 19:34

wegank added the 2.status: merge conflict This PR has merge conflicts with the target branch label Sep 27, 2024

RossComputerGuy force-pushed the fix/pkgsllvm/valgrind branch from 64b61ce to ec5f27d Compare October 12, 2024 00:48

ofborg bot removed the 2.status: merge conflict This PR has merge conflicts with the target branch label Oct 12, 2024

wegank added the 2.status: merge conflict This PR has merge conflicts with the target branch label Apr 2, 2025

RossComputerGuy force-pushed the fix/pkgsllvm/valgrind branch from ec5f27d to 52ee4b6 Compare April 18, 2025 04:38

ofborg bot removed the 2.status: merge conflict This PR has merge conflicts with the target branch label Apr 18, 2025

RossComputerGuy mentioned this pull request Apr 18, 2025

llvmPackages.libunwind: add fake libgcc #330037

Merged

13 tasks

RossComputerGuy closed this May 13, 2025

RossComputerGuy force-pushed the fix/pkgsllvm/valgrind branch from 52ee4b6 to 748fab9 Compare May 13, 2025 03:49

RossComputerGuy reopened this May 13, 2025

RossComputerGuy mentioned this pull request Sep 5, 2025

pkgsLLVM.linuxPackages: kernel module fixes #440075

Merged

13 tasks

JohnRTitor closed this Sep 6, 2025

JohnRTitor reopened this Sep 6, 2025

JohnRTitor requested review from a team and emilazy September 6, 2025 21:19

RossComputerGuy force-pushed the fix/pkgsllvm/valgrind branch from 419271d to 832ed2a Compare September 12, 2025 02:31

valgrind: fix building with llvm

3552ede

RossComputerGuy force-pushed the fix/pkgsllvm/valgrind branch from 832ed2a to 3552ede Compare September 12, 2025 02:59

alyssais requested changes Sep 13, 2025

View reviewed changes

nixpkgs-ci bot added the 2.status: merge-bot eligible This PR can be merged by commenting "@NixOS/nixpkgs-merge-bot merge". label Nov 2, 2025

nixpkgs-ci bot removed the 2.status: merge-bot eligible This PR can be merged by commenting "@NixOS/nixpkgs-merge-bot merge". label Jan 11, 2026

Uh oh!

Conversation

RossComputerGuy commented Jul 25, 2024

Description of changes

Things done

Uh oh!

RossComputerGuy commented Jul 25, 2024

Uh oh!

RossComputerGuy commented Aug 2, 2024

Uh oh!

RossComputerGuy commented Aug 2, 2024

Uh oh!

Ericson2314 commented Aug 2, 2024

Uh oh!

RossComputerGuy commented Aug 2, 2024

Uh oh!

Ericson2314 commented Aug 2, 2024

Uh oh!

RossComputerGuy commented Aug 3, 2024

Uh oh!

Ericson2314 commented Aug 4, 2024

Uh oh!

RossComputerGuy commented Sep 23, 2024

Uh oh!

nixos-discourse commented Apr 18, 2025

Uh oh!

RossComputerGuy commented May 13, 2025

Uh oh!

RossComputerGuy commented May 13, 2025

Uh oh!

RossComputerGuy commented Sep 12, 2025

Uh oh!

alyssais Sep 13, 2025

Choose a reason for hiding this comment

Uh oh!

RossComputerGuy Sep 13, 2025

Choose a reason for hiding this comment

Uh oh!

alyssais Sep 13, 2025

Choose a reason for hiding this comment

Uh oh!

JohnRTitor Oct 7, 2025

Choose a reason for hiding this comment

Uh oh!

RossComputerGuy Oct 7, 2025

Choose a reason for hiding this comment

Uh oh!

alyssais Sep 13, 2025

Choose a reason for hiding this comment

Uh oh!

RossComputerGuy Sep 13, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants