[gnuastro-devel] [bug #51476] Crashes on 32-bit and big-endian systems

[Mohammad Akhlaghi]

URL:
  <http://savannah.gnu.org/bugs/?51476>

                 Summary: Crashes on 32-bit and big-endian systems
                 Project: GNU Astronomy Utilities
            Submitted by: makhlaghi
            Submitted on: Sat 15 Jul 2017 11:38:12 PM CEST
                Category: MakeProfiles
                Severity: 5 - Blocker
              Item Group: Crash
                  Status: In Progress
                 Privacy: Public
             Assigned to: makhlaghi
             Open/Closed: Open
         Discussion Lock: Any

    _______________________________________________________

Details:

Gnuastro 0.3.31 (fix to bug #51467) was approved and uploaded onto Debian a
few hours ago and almost immediately several crashes were reported in the
tests (they are still ongoing, but already 12 architectures have failed!):

https://buildd.debian.org/status/package.php?p=gnuastro

A summary of the results is:


armel:   radeccat.sh segfault
armhf:   radeccat.sh segfault
i386:    radeccat.sh segfault
mips:    all mkprof segfault
mipsel:  radeccat.sh segfault
s390x:   all mkprof no mode option
hppa:    all mkprof segfault
powerpc: all mkprof segfault
ppc64:   all mkprof no mode option
sh4:     memory problem in arithmetic library.
sparc64: all mkprof Bus error
x32:     radeccat.sh segfault


Except for the `sh4' architecture, the rest are all in MakeProfiles and appear
related.

After trying the build on a 32-bit virtual machine, I was able to find the
cause of the problem: the `mode' element of the `mkprofparams' (defined in
MakeProfile's `main.h') was defined as uint8_t, but in the
`ui_parse_coordinate_mode', it was being read as an `int'! Francly it is
strange for me why this didn't cause a crash on my own system!?! 

This maybe one of the problems with compilers becoming "smarter" (and hiding
problems)!





    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?51476>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/