Hi All,
I was wondering if there were any plans to do a new 7.6 release in the near future. There's a bug that's critical for me in the .112 Mysql module that causes a segfault, that's been fixed in 7.6 CVS. Because of this, I was thinking about holding off on a release on my end if a new pike version was looming on the horizon. If not, that's okay, too, I can try to come up with a workaround.
Bill
Maybe I should say it failed. There is a problem with compilation of Gdbm on Solaris. Configure tries and succeds to link with gcc, but it then uses /usr/ccs/bin/ld to fail at linking during the actual compilation. It's not a recent regression. I'm not even sure it's a regression, but I'd like that sorted out before we do a release. It's just wrong to use diffrent tools when configuring and compileing.
I don't think I will get it fixed this weekend, so anyone who wants a release quickly should help out. :)
For reference:
OK:
configure:3683: /home/peter/hack/Pike/7.6-distmaker/build/sunos-5.11-i86pc/smartlink gcc -o conftest -g -Wa,--execstack -mcpu=i686 -I/home/peter/hack/Pike/7.6-distmaker/build/sunos-5.11-i86pc/bundles/include -I/usr/local/include -I/usr/X11/include -L/home/peter/hack/Pike/7.6-distmaker/build/sunos-5.11-i86pc/bundles/lib -R/usr/local/lib/. -L/usr/local/lib/. -R/usr/X11/lib/. -L/usr/X11/lib/. -R/usr/X11R6/lib/. -L/usr/X11R6/lib/. conftest.c -lgdbm -ldl -lrt -lnsl -lsocket -lm -lpthread -lthread >&5
FAIL:
/home/peter/hack/Pike/7.6-distmaker/bin/smartlink /usr/ccs/bin/ld -G -o module.so gdbmmod.o -L/home/peter/hack/Pike/7.6-distmaker/build/sunos-5.11-i86pc/bundles/lib -R/usr/local/lib/. -L/usr/local/lib/. -R/usr/X11/lib/. -L/usr/X11/lib/. -R/usr/X11R6/lib/. -L/usr/X11R6/lib/. -lgdbm -ldl -lrt -lnsl -lsocket -lm -lpthread -lthread /opt/csw/gcc3/lib/gcc/i386-pc-solaris2.8/3.4.5/libgcc.a -lc /opt/csw/gcc3/lib/gcc/i386-pc-solaris2.8/3.4.5/libgcc.a
Compilation failure:
Linking Gdbm ld: fatal: library -lgdbm: not found ld: fatal: File processing errors. No output written to module.so Linking failed: /home/peter/hack/Pike/7.6-distmaker/bin/smartlink /usr/ccs/bin/ld -G -o module.so gdbmmod.o -L/home/peter/hack/Pike/7.6-distmaker/build/sunos-5.11-i86pc/bundles/lib -R/usr/local/lib/. -L/usr/local/lib/. -R/usr/X11/lib/. -L/usr/X11/lib/. -R/usr/X11R6/lib/. -L/usr/X11R6/lib/. -lgdbm -ldl -lrt -lnsl -lsocket -lm -lpthread -lthread /opt/csw/gcc3/lib/gcc/i386-pc-solaris2.8/3.4.5/libgcc.a -lc /opt/csw/gcc3/lib/gcc/i386-pc-solaris2.8/3.4.5/libgcc.a *** Error code 1
Odd... there's nothing unusual about the Makefile or configure script, and they don't seem to have changed in years. I assume that libgdbm is in one of the specified directories? Perhaps a command line order problem? To the bigger problem, I'm guessing that it's using LDSHARED or such during compilation? Don't know if there's a way to get around that, as I'd guess the configure scripts are somewhat hard coded to use CC rather than LDSHARED.
Wish I still had a solaris box laying around :/
Bill
On Sun, 9 Dec 2007, Peter Bortas @ Pike developers forum wrote:
Maybe I should say it failed. There is a problem with compilation of Gdbm on Solaris. Configure tries and succeds to link with gcc, but it then uses /usr/ccs/bin/ld to fail at linking during the actual compilation. It's not a recent regression. I'm not even sure it's a regression, but I'd like that sorted out before we do a release. It's just wrong to use diffrent tools when configuring and compileing.
I don't think I will get it fixed this weekend, so anyone who wants a release quickly should help out. :)
The problem is that gcc har built in linkpaths, in this case gcc comes from the Blastwave archives, and automaticly searches for libs in /opt/csw. The system ld does not.
gcc is obviously installed somewhere in /opt/csw/, and gcc tends to add ${prefix}/lib to the libpath. One possible solution would be to modify AC_CHECK_LIB to use the -print-file-name option to locate the library if the compiler is gcc, and add the path to LDFLAGS if it isn't already there.
Current status is that things work well on the Unix end, but I've let the Windows port fall into disrepair. You can expect a new build when this turns green:
http://pike.ida.liu.se/development/pikefarm/result.xml?id=733_341&pike=7...
I actually have experienced the same error on my end (I was trying to do a windows build again, and it failed at the exact same test during configure). I was using WinXP/iX86 + VC6 + rntcl.
Bill
On Jan 7, 2008, at 3:00 AM, Peter Bortas @ Pike developers forum wrote:
Current status is that things work well on the Unix end, but I've let the Windows port fall into disrepair. You can expect a new build when this turns green:
http://pike.ida.liu.se/development/pikefarm/result.xml? id=733_341&pike=7_6
!DSPAM:4781dcfc274761006682466!
pike-devel@lists.lysator.liu.se
-
H. William Welliver III -
Henrik Grubbstr�m (Lysator) ＠ Pike (-) developers forum -
Marcus Comstedt (ACROSS) (Hail Ilpalazzo!) ＠ Pike (-) developers forum
-
Peter Bortas ＠ Pike developers forum