Bug 708430 – sys-libs/compiler-rt-sanitizers-9.0.1 with sys-libs/glibc-2.31 - .../work/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.cc:1131:1: error: 'assertion_failed__1131' declared as an array with a negative size

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Mike Lothian 2020-02-05 23:58:06 UTC]

This has been fixed upstream with:

https://github.com/llvm/llvm-project/commit/947f9692440836dcb8d88b74b69dd379d85974ce.patch

I've reworked the patch to apply on 9.0.1

Reproducible: Always

[Comment 1 Mike Lothian 2020-02-05 23:58:40 UTC]

Created attachment 612052 [details]
Build log

[Comment 2 Mike Lothian 2020-02-05 23:59:25 UTC]

Created attachment 612054 [details, diff]
Patch for glibc-2.31

[Comment 3 Mike Lothian 2020-02-06 00:00:18 UTC]

FAILED: lib/sanitizer_common/CMakeFiles/RTSanitizerCommonNoTermination.i386.dir/sanitizer_platform_limits_posix.cc.o 
/usr/lib/llvm/9/bin/x86_64-pc-linux-gnu-clang++ -DHAVE_RPC_XDR_H=0 -I/var/tmp/portage/sys-libs/compiler-rt-sanitizers-9.0.1/work/compiler-rt/lib/sanitizer_common/..  -O3 -march=native -pipe -Wall -std=c++11 -Wno-unused-parameter    -m32 -fPIC -fno-builtin -fno-exceptions
 -fomit-frame-pointer -funwind-tables -fno-stack-protector -fno-sanitize=safe-stack -fvisibility=hidden -fno-lto -O3 -gline-tables-only -Wno-gnu -Wno-variadic-macros -Wno-c99-extensions -Wno-non-virtual-dtor -fno-rtti -Wframe-larger-than=570 -Wglobal-constructors -MD -MT
 lib/sanitizer_common/CMakeFiles/RTSanitizerCommonNoTermination.i386.dir/sanitizer_platform_limits_posix.cc.o -MF lib/sanitizer_common/CMakeFiles/RTSanitizerCommonNoTermination.i386.dir/sanitizer_platform_limits_posix.cc.o.d -o lib/sanitizer_common/CMakeFiles/RTSanitizer
CommonNoTermination.i386.dir/sanitizer_platform_limits_posix.cc.o -c /var/tmp/portage/sys-libs/compiler-rt-sanitizers-9.0.1/work/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.cc
/var/tmp/portage/sys-libs/compiler-rt-sanitizers-9.0.1/work/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.cc:1131:1: error: 'assertion_failed__1131' declared as an array with a negative size
CHECK_SIZE_AND_OFFSET(ipc_perm, mode);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/var/tmp/portage/sys-libs/compiler-rt-sanitizers-9.0.1/work/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.h:1437:3: note: expanded from macro 'CHECK_SIZE_AND_OFFSET'
  COMPILER_CHECK(sizeof(((__sanitizer_##CLASS *) NULL)->MEMBER) == \
  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/var/tmp/portage/sys-libs/compiler-rt-sanitizers-9.0.1/work/compiler-rt/lib/sanitizer_common/sanitizer_internal_defs.h:336:30: note: expanded from macro 'COMPILER_CHECK'
#define COMPILER_CHECK(pred) IMPL_COMPILER_ASSERT(pred, __LINE__)
                             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/var/tmp/portage/sys-libs/compiler-rt-sanitizers-9.0.1/work/compiler-rt/lib/sanitizer_common/sanitizer_internal_defs.h:342:57: note: expanded from macro 'IMPL_COMPILER_ASSERT'
    typedef char IMPL_PASTE(assertion_failed_##_, line)[2*(int)(pred)-1]
                                                        ^~~~~~~~~~~~~~~
1 error generated.

[Comment 4 Michał Górny 2020-02-23 18:06:20 UTC]

Thanks, I'll backport it now.

[Comment 5 Michał Górny 2020-02-23 18:19:07 UTC]

Actually, it doesn't apply cleanly and it seems that getting full set of patches is not worth the effort.  Just use 10.x if you need bleeding edge glibc.

[Comment 6 Michał Górny 2020-02-25 18:59:08 UTC]

*** Bug 710800 has been marked as a duplicate of this bug. ***

[Comment 7 Frits Letteboer 2020-03-11 19:49:35 UTC]

Created attachment 618030 [details, diff]
Fixes the compilation issue with glibc 2.31

As a quick workaround

Taken from: http://www.linuxfromscratch.org/blfs/view/svn/general/llvm.html

[Comment 8 Craig Andrews 2020-05-01 02:34:06 UTC]

(In reply to Frits Letteboer from comment #7)
> Created attachment 618030 [details, diff] [details, diff]
> Fixes the compilation issue with glibc 2.31
> 
> As a quick workaround
> 
> Taken from: http://www.linuxfromscratch.org/blfs/view/svn/general/llvm.html

I confirm that this patch solves the problem.

[Comment 9 Sam James 2020-05-01 03:36:35 UTC]

*** Bug 720222 has been marked as a duplicate of this bug. ***

[Comment 10 Pavel Volkov 2020-05-08 06:25:43 UTC]

(In reply to Michał Górny from comment #5)
> Actually, it doesn't apply cleanly and it seems that getting full set of
> patches is not worth the effort.  Just use 10.x if you need bleeding edge
> glibc.

It's said that some packages are broken with LLVM 10, in other distros as well — bug 711788.

[Comment 11 Victor Mataré 2020-05-13 22:26:21 UTC]

app-doc/doxygen has RDEPEND="clang? ( <sys-devel/clang-10:= )", so we should either get rid of that or continue to support clang:9.

[Comment 12 Bernd 2020-05-19 09:45:58 UTC]

This issue also happens to compiler-rt-sanitizers:8.0.1

[Comment 13 Arfrever Frehtes Taifersar Arahesis 2020-07-21 21:57:36 UTC]

*** Bug 733378 has been marked as a duplicate of this bug. ***

[Comment 14 Andrew 2020-07-27 14:21:02 UTC]

I also confirm that the patch from  Frits Letteboer  solves the problem

[Comment 15 Viktar Patotski 2020-07-31 14:53:37 UTC]

I Confirm that patch from Frits Letteboer solves compilation problem and I'm able to build sys-libs/compiler-rt-sanitizers-9.0.1. However I believe disabling check is not the best way fixing it. We need to wait for upstream to fix it.

[Comment 16 Herbert Wantesh 2020-09-12 10:12:11 UTC]

the patch from Frits Letteboer fixes the compile for me too

[Comment 17 Luke-Jr 2020-09-27 13:31:14 UTC]

*** Bug 744877 has been marked as a duplicate of this bug. ***

[Comment 18 Luke-Jr 2020-09-27 13:32:30 UTC]

glibc 2.31 is stable now... would be nice to have a path forward?

(In reply to Michał Górny from comment #5)
> Just use 10.x if you need bleeding edge glibc.

10.x what? This build still fails with LLVM/Clang 10 installed.

[Comment 19 tt_1 2020-09-27 15:18:39 UTC]

llvm/clang is a whole toolchain and some of the ebuilds have slots, others don't. compiler-rt-sanitizers has slots, which means you have to emerge =compiler-rt-sanitizers-10.0.x to match your llvm/clang-10 toolchain. I thought this was obvious? 

personally I whished the provided patch was backported, but I can see that ressources for testing are limited. I thus removed USE="sanitize" from my llvm:9 and wait a bit longer for llvm-10.0.1 to be stable on all relevant arches

[Comment 20 Luke-Jr 2020-09-27 17:43:38 UTC]

(In reply to Craig Andrews from comment #8)
> I confirm that this patch solves the problem.

Are you sure it doesn't simply remove compile-time sanity checks? Looks like with this patch, big endian platforms (other than MIPS, aarch64, and s390x) will break at runtime (ie, when performing the actual sanitizer checks)...

[Comment 21 Luke-Jr 2020-09-27 17:49:13 UTC]

Created attachment 662686 [details, diff]
Patch to adjust sys-libs/compiler-rt-sanitizers-9.0.1 for glibc 2.31+

This patch is an actual backport of the relevant changes in version 10.

Disclaimer: This does remove functionality from (and perhaps breaks sanitizer use with) systems with older glibc versions (2.30 and earlier); but with a current glibc, I think it should work. Untested.

[Comment 22 Alex Efros 2020-12-05 10:39:20 UTC]

(In reply to Luke-Jr from comment #21)
> Created attachment 662686 [details, diff] [details, diff]
> Patch to adjust sys-libs/compiler-rt-sanitizers-9.0.1 for glibc 2.31+
> 
> This patch is an actual backport of the relevant changes in version 10.

Works for me, thanks!

[Comment 23 Guilherme Amadio 2020-12-19 14:02:32 UTC]

Many packages still have not adapted t

[Comment 24 Guilherme Amadio 2020-12-19 14:04:57 UTC]

Many packages have not yet adapted to LLVM/Clang 10, so we still need to keep 9 around in the mean time. If there are no objections, I will apply the patch by Luke Jr to 9.0.1 in a new revision. Michał, what do you think?

[Comment 25 Michał Górny 2020-12-19 14:13:07 UTC]

(In reply to Guilherme Amadio from comment #24)
> Many packages have not yet adapted to LLVM/Clang 10, so we still need to
> keep 9 around in the mean time. If there are no objections, I will apply the
> patch by Luke Jr to 9.0.1 in a new revision. Michał, what do you think?

Just make sure to put it into files/9.0.1/ directory.

[Comment 26 Larry the Git Cow 2021-01-07 13:59:57 UTC]

The bug has been closed via the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=2741567d6f7fb139c8ad37137fb581987a779e97

commit 2741567d6f7fb139c8ad37137fb581987a779e97
Author:     Guilherme Amadio <[e-mail]>
AuthorDate: 2020-12-19 14:07:38 +0000
Commit:     Guilherme Amadio <[e-mail]>
CommitDate: 2021-01-07 13:59:40 +0000

    sys-libs/compiler-rt-sanitizers: fix for glibc 2.31+, bug #708430
    
    Closes: https://bugs.gentoo.org/708430
    Package-Manager: Portage-3.0.12, Repoman-3.0.2
    Signed-off-by: Guilherme Amadio <[e-mail]>

 .../compiler-rt-sanitizers-9.0.1-r1.ebuild         | 160 +++++++++++++++++++++
 .../files/9.0.1/glibc-2.31.patch                   |  35 +++++
 2 files changed, 195 insertions(+)