Request 455481 (accepted)

configure:19675: checking for res_gethostbyaddr in -lresolv
configure:19700: g++ -o conftest -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -g -Wall -Wextra -Winit-self -Wswitch-default -Wswitch-enum -Wstrict-aliasing=2 -Wfloat-equal -Wshadow   conftest.cpp -lresolv -lresolv  >&5
/tmp/ccej4g9H.o: In function `main':
/home/abuild/rpmbuild/BUILD/ekiga-4.0.1/conftest.cpp:45: undefined reference to `res_gethostbyaddr'
collect2: error: ld returned 1 exit status

Apparently that disappeared from glibc 2.25?

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

@Andreas_Schwab just wondering; is this change expected? (it's what makes ekiga fail)

>readelf -a /lib64/libresolv-2.24.so | grep gethostbyaddr
   148: 0000000000004e60  1143 FUNC    GLOBAL DEFAULT   14 res_gethostbyaddr@[@GLIBC_2.2.5](https://build.opensuse.org/users/GLIBC_2.2.5)
   281: 0000000000004e60  1143 FUNC    GLOBAL DEFAULT   14 res_gethostbyaddr

> readelf -a /lib64/libresolv-2.25.so | grep gethostbyaddr
   147: 000000000000e8b0   918 FUNC    GLOBAL DEFAULT   14 res_gethostbyaddr@GLIBC_2.2.5
   281: 000000000000e8b0   918 FUNC    GLOBAL DEFAULT   14 res_gethostbyaddr@GLIBC_2

So with glibc 2.24, there is a res_gethostbyaddr DEFAULT symbol; on glibc 2.25 there is no longer a default, just the two symbols for backwards compat.

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

Repetitive build failures of libzio on ppc64le observed:

[   13s] gcc -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -g -pipe -Wall -D_GNU_SOURCE -D_REENTRANT  -DHAVE_LIBIO_H -DHAVE_FOPENCOOKIE  -DHAS_ZLIB_H -DHAS_BZLIB_H  -DHAS_LZMA_H -DNO_WEAK -fPIC -o obs/unlzw.o -c unlzw.c
[   13s] In file included from lzw.h:48:0,
[   13s]                  from unlzw.c:34:
[   13s] lzw.h:52:13: error: 'pthread_sigmask' undeclared here (not in a function)
[   13s]  weak_symbol(pthread_sigmask);

@WernerFink - glibc 2.25 update is involved here

Dr. Werner Fink (WernerFink) - about 7 years ago

source maintainer

Does this only happen on ppc64le?

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

From what I have seen in Staging so far yes - the other archs seem to build

Dr. Werner Fink (WernerFink) - about 7 years ago

source maintainer

I had yesterday add the -D_DEFAULT_SOURCE switch as this is more up to date than the old -D_REENTRANT switch. I'd like to know if this might help on ppc64le to get /usr/include/bits/sigthread.h loaded by /usr/include/signal.h ... that is that __USE_POSIX199506 becomes a defined switch by /usr/include/features.h

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

strace fails test suite on x86_64 with glibc 2.25

https://build.opensuse.org/package/live_build_log/openSUSE:Factory:Staging:B/strace/standard/x86_64

@AndreasSchwab

Andreas Schwab (AndreasSchwab) - about 7 years ago

Worksforme. https://build.opensuse.org/package/show/home:Andreas_Schwab:Factory/strace

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

Thanks for your verification run - seems strace has seen an update on a parallel staging process - which did not make it through :B

r59 | dimstar_suse | 2017-02-20 13:27:15 | f301a16e90ce036a83aada8800755127 | 4.16 | rq457371

:B did still test strace 4.15.. will fix this up

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

gcc5 fails constantly to build on ppc64le using glibc 2.25

CC @rguenther and matz2

https://build.opensuse.org/package/live_build_log/openSUSE:Factory:Staging:B/gcc5/standard/ppc64le

Richard Biener (rguenther) - about 7 years ago

../../../../libsanitizer/asan/asan_linux.cc: In f unction 'bool __asan::AsanInterceptsSignal(int)': [ 3365s] ../../../../libsanitizer/asan/asan_linux.cc:222:20: error: 'SIGSEGV' wa s not declared in this scope [ 3365s] return signum == SIGSEGV && common_flags()->handle_segv; [ 3365s] ^

missing include. But gcc5 can be dropped once libffi is in... (openSUSE:Factory:Staging:adi:24)

You really want me to fix gcc5?

Richard Biener (rguenther) - about 7 years ago

gcc5 can move out of ring0 to DVD, only libffi-gcc5 needs to be retained for now (in -MinimalX)

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

wouldn't change the fact that I need a fix for gcc5 - ring0, ring1 and ring2 are tested and guaranteed to build; but moving it out of Ring0 to a higher ring is certainly something we will have a look at then. Thanks for the pointer

Richard Biener (rguenther) - about 7 years ago

Fix is in devel:gcc/gcc5

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

Hmm. you mean gcc5 could be completley dropped with libffi in; that sounds actually interesting and would indeed mean we should not invest into fixing this.

I'll try to add a delete request for gcc5 and temporarily add libffi from adi:24 over here to Staging:C - then we also get the confirmation that this is indeed all as good as we expect

Richard Biener (rguenther) - about 7 years ago

Yes. At least gcc5 can be moved from Ring0 to DVD (so out of stagings). Without libffi in we only need libffi-gcc5 in Ring1.

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

That part is unfortunatley not possible, as libffi-gcc5 is not the 'leading' package, it requires at least that gcc5 to be in the rings too (ring0, 1 or 2) and in this case the build would need to be fixed.

But the current setup I have is with libffi 3.2.1 added to ring1 with gcc5 and libff-gcc5 removed from the rings. This seems to work and is a solution I like (especially as I can get a compiler out of the rings)

Richard Biener (rguenther) - about 7 years ago

Ah, that's unfortunate. Let's hope we can get the pending legal review for libffi quickly then.

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

gdb fails to build with glibc 2.25 (x86_64 and ppc64le)

https://build.opensuse.org/package/live_build_log/openSUSE:Factory:Staging:B/gdb/standard/x86_64

CC @matz2

Richard Biener (rguenther) - about 7 years ago

SR#460722

Dominique Leuenberger (dimstar) - about 7 years ago

reviewer

Xen fails to build with glibc 2.25