Le vendredi 15 août 2014 à 11:29 +0200, Gilles Filippini a écrit : > Source: octave > Version: 3.8.2-1 > Severity: serious > Justification: FTBFS on s390x > User: pini@debian.org > Usertags: HDF5-transition > The last upload of octave FTBFS on s390x with a segfault during the > test step: > libinterp/dldfcn/__delaunayn__.cc-tst .................. PASS 1/1 > libinterp/dldfcn/__dsearchn__.cc-tst ................... PASS 1/1 > libinterp/dldfcn/__fltk_uigetfile__.cc-tst ............. PASS 1/1 > libinterp/dldfcn/__glpk__.cc-tst ....................... PASS 1/1 > libinterp/dldfcn/__magick_read__.cc-tst ................ PASS 4/4 > libinterp/dldfcn/__voronoi__.cc-tst .................... PASS 1/1 > libinterp/dldfcn/chol.cc-tst ........................... PASS 29/30 FAIL 1 > libinterp/dldfcn/convhulln.cc-tst ......................panic: Segmentation fault -- stopping myself... > Makefile:2565: recipe for target 'check' failed > make[3]: *** [check] Segmentation fault > make[3]: Leaving directory '/«PKGBUILDDIR»/test' > Makefile:2849: recipe for target 'check' failed > > Unfortunately this FTBFS is in the way of the upcoming HDF5 transition > and I really don't know how to help in solving it. My current hypothesis is that this FTBFS is due to the glibc ABI break (http://lwn.net/Articles/605607/), in relation with the qhull library, which was compiled with the old ABI and which is used in convhulln.cc. The ABI break was reverted in glibc 2.19-9, which was uploaded yesterday. I am currently testing a rebuild of octave against the new glibc. If that works, I'll just request a give back. -- .''`. Sébastien Villemot : :' : Debian Developer `. `' http://www.dynare.org/sebastien `- GPG Key: 4096R/381A7594
Attachment:
signature.asc
Description: This is a digitally signed message part