Sage build fails. But restarts OK after typing "make" again ???

[David Kirkby]

I tried to build Sage 5.8.beta2 on a Sun Blade 2000 SPARC. The build
fails, but install.log does not indicate why. Restarting the build
appears to solve the problem.

Writing /export/home/drkirkby/sage-5.8.beta2/local/lib/python2.7/site-packages/scipy-0.11.0-py2.7.egg-info
running install_clib

real 51m11.101s
user 30m27.387s
sys 1m9.826s
Successfully installed scipy-0.11.0.p1
Deleting temporary build directory
/export/home/drkirkby/sage-5.8.beta2/spkg/build/scipy-0.11.0.p1
Finished installing scipy-0.11.0.p1.spkg
make[2]: Leaving directory `/export/home/drkirkby/sage-5.8.beta2/spkg'
make[1]: *** [all] Error 2
make[1]: Leaving directory `/export/home/drkirkby/sage-5.8.beta2/spkg'

real 304m56.173s
user 486m56.945s
sys 40m15.698s
Error building Sage.

So the build failed. Typing "make" again restarts the build, and as I
write fflas_ffpack, iml and cvxopt have all built after scipy.

A bit odd. I've never seen this happen before.

I built another copy of Sage in parallel with this one on the same
machine. That built without any problems at all.

Dave

[Jeroen Demeyer]

On 2013-03-02 12:28, David Kirkby wrote:
> I tried to build Sage 5.8.beta2 on a Sun Blade 2000 SPARC. The build
> fails, but install.log does not indicate why.

I assume this was a parallel build?

I'm sure it does indicate why, just not at the end of the file. When
make encounters an errors, it continues building the stuff it was
already building. This means that there can be a large gap between the
point the error occurs and the end of the build.

Try grepping the logs in spkg/logs for a message like "Error installing"

[kcrisman]

Or, when it happens, check out which package is still in spkg/build/ and check the log for THAT spkg.  That's what I have to do. 

[David Kirkby]

On 2 March 2013 13:04, Jeroen Demeyer <jdem...@cage.ugent.be> wrote:
> On 2013-03-02 12:28, David Kirkby wrote:
>> I tried to build Sage 5.8.beta2 on a Sun Blade 2000 SPARC. The build
>> fails, but install.log does not indicate why.
> I assume this was a parallel build?
>
> I'm sure it does indicate why, just not at the end of the file. When
> make encounters an errors, it continues building the stuff it was
> already building. This means that there can be a large gap between the
> point the error occurs and the end of the build.

Thank you. Checking I see you are right. It was the build of "R" that
failed. Odd this.

mkdir ../../../../library/tools/libs
make[9]: Leaving directory
`/export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/src/library/tools/src'
make[8]: Leaving directory
`/export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/src/library/tools/src'
make[7]: Leaving directory
`/export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/src/library/tools'
ld.so.1: /export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/bin/exec/R:
fatal: libgcc_s.so.1: version `GCC_4.0.0' not found (required by file
/export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/lib/libR.so)
ld.so.1: /export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/bin/exec/R:
fatal: libgcc_s.so.1: open failed: No such file or directory
/bin/bash: line 1: 5021 Done echo
"tools:::.install_package_description('.',
'"../../../library/tools"')"
5022 Killed | R_DEFAULT_PACKAGES=NULL
../../../bin/R --vanilla --slave >/dev/null
make[6]: *** [all] Error 137
make[6]: Leaving directory
`/export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/src/library/tools'
make[5]: *** [R] Error 1
make[5]: Leaving directory
`/export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/src/library'
make[4]: *** [R] Error 1
make[4]: Leaving directory
`/export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src/src'
make[3]: *** [R] Error 1
make[3]: Leaving directory
`/export/home/drkirkby/sage-5.8.beta2/spkg/build/r-2.15.2.p1/src'
Error building R.

real 39m50.830s
user 13m53.030s
sys 1m39.435s
************************************************************************
Error installing package r-2.15.2.p1
************************************************************************
Please email sage-devel (http://groups.google.com/group/sage-devel)

I would note this failure was with the gcc on my computer (4.5.0), not
the one in Sage tarball. Perhaps I have a compiler issue.

I was at the same time building another copy of the same version of
Sage, where I'd purposely changed the path so only the old 3.4.3
version of gcc could be found. So to build Sage it was necessary for
the gcc 4.6.3 package in Sage to be built. That build of Sage
completed OK and I'm just running the doctests on that now.

I'm a bit puzzled by this, as I've built Sage with the 4.5.0 before.
But the 4.6.3 package in Sage is saving the day!

Dave

[leif]

Good point. (Unless you've set SAGE_KEEP_BUILT_SPKGS=yes, or previously
kept the build dirs of some spkgs.)

We should probably give hints on that (at least if
SAGE_PARALLEL_SPKG_BUILD=yes) in the error message at the end as well;
how to grep for the failing spkg(s) is currently documented in some
installation instructions (Installation Guide, README?) only.


No idea what went wrong in Dave's first build of R, since libgcc_s.*
should be located outside the Sage tree unless Sage's GCC got installed.


-leif

--
() The ASCII Ribbon Campaign
/\ Help Cure HTML E-Mail