This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: binutils 2.28 x86_64-linux 'make check' failures when built with gcc-5.4.0 + glibc-2.25
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: Jason Vas Dias <jason dot vas dot dias at gmail dot com>
- Cc: Binutils <binutils at sourceware dot org>
- Date: Tue, 7 Mar 2017 11:04:49 -0800
- Subject: Re: binutils 2.28 x86_64-linux 'make check' failures when built with gcc-5.4.0 + glibc-2.25
- Authentication-results: sourceware.org; auth=none
- References: <CALyZvKwScVUn5x6m4jPLm1GtwdL8-5iCe=Bn47_aUgRFmZ1fgA@mail.gmail.com>
On Tue, Mar 7, 2017 at 10:48 AM, Jason Vas Dias
<jason.vas.dias@gmail.com> wrote:
> Good day binutils wizards -
>
> I just built binutils-2.28 + gdb-7.12.1 on an x86_64 corei7 Haswell CPU,
> (Intel i4910MQ 2.9-3.9Ghz 4-CPU 8-core)
> with GCC 5.4.0 & glibc-2.25 installed , with configure args:
>
> $ /usr/os_src/binutils-2.28/configure --prefix=/usr --libdir=/usr/lib64 \
> --sysconfdir=/etc --localstatedir=/var --enable-targets=all --enable-shared \
> --with-system-zlib --build=x86_64-linux-gnu --host=x86_64-linux-gnu
>
> The build of binutils & gdb succeeded with no errors / warnings.
>
> I got some new test failures running 'make check', and was wondering if they
> are serious or harmless known issues :
>
> $ make check 2>&1 | tee make.check.log
> ...
> === ld Summary ===
>
> # of expected passes 1732
> # of unexpected failures 31
> # of unexpected successes 33
> # of expected failures 24
> # of untested testcases 1
> # of unsupported tests 12
> /usr/build/linux/binutils-2.28/ld/ld-new 2.28
> ...
>
> $ grep ^FAIL make.check.log
> FAIL: TLS -fpic and -fno-pic exec transitions
> FAIL: TLS -fpic and -fno-pic exec transitions without PLT
> FAIL: TLS descriptor -fpic and -fno-pic exec transitions
> FAIL: TLS -fno-pic -shared
> FAIL: ld-ifunc/pr17154-x86-64
> FAIL: NOCROSSREFS_TO 3
> FAIL: NOCROSSREFS_TO 4
> FAIL: Build size-7
> FAIL: Build size-8
> FAIL: S-records
> FAIL: S-records with constructors
> FAIL: Build mpx4
> FAIL: ld-x86-64/bnd-ifunc-2
> FAIL: ld-x86-64/bnd-plt-1
> FAIL: TLS -fpic -shared transitions
> FAIL: TLS -fpic -shared transitions with r15 as GOT base
> FAIL: TLS descriptor -fpic -shared transitions
> FAIL: TLS -fpic and -fno-pic exec transitions
> FAIL: TLS -fpic and -fno-pic exec transitions without PLT
> FAIL: TLS descriptor -fpic and -fno-pic exec transitions
> FAIL: PR ld/14207
> FAIL: ld-x86-64/ilp32-4
> FAIL: ld-x86-64/load1c
> FAIL: ld-x86-64/load1d
> FAIL: ld-x86-64/pr19162
> FAIL: ld-x86-64/pr19636-2d
> FAIL: ld-x86-64/pr20253-1d
> FAIL: ld-x86-64/pr20253-1f
> FAIL: ld-x86-64/pr20253-1j
> FAIL: ld-x86-64/pr20253-1l
> FAIL: Undefined weak symbol
>
> The S-Record failures maybe I can understand (not on embedded JTAG
> controlled platform) but the TLS failures worry me, as I use TLS alot
> in my code .
>
> Should there be TLS failures on this platform ? Should I not use this build ?
> I have been using a binutils-2.27 version, which had fewer test failures :
>
> $ grep '^FAIL' ${my_2_27_build}/make.check.log
> FAIL: NOCROSSREFS_TO 3
> FAIL: NOCROSSREFS_TO 4
> FAIL: Build size-7
> FAIL: Build size-7
> FAIL: Build size-8
> FAIL: Build size-8
> FAIL: S-records
> FAIL: S-records with constructors
> FAIL: Undefined weak symbol
> FAIL: PLT PC-relative offset overflow check
>
> I am inclined to continue using 2.27 until I can verify if the new
> test failures are
> serious issues or not - any help / advice in making this determination would be
> much appreciated - has anyone seen such failures on linux-x86_64 with
> 2.28 before?
>
> Thanks & Regards,
> Jason
I got no failures with glibc 2.24 and gcc 5.3.1. I guess your compiler default
may not be compatible with linker tests.
--
H.J.