[Ecls-list] uninitialized ihs_frame object in call to ecl_cs_check
Gabriel Dos Reis
gdr at cs.tamu.edu
Thu Nov 6 10:03:30 UTC 2008
"Juan Jose Garcia-Ripoll" <juanjose.garciaripoll at googlemail.com> writes:
| > But, I'm still puzzled by this curious bug (which I think Waldek
| > also alluded to at some point in the past).
| > If I build OpenAxiom with debug=3, then the bug disappears.
| > If I set debug=2 or less, I can reproduce it. If I trace some specific
| > functions, I can no longer reproduce the bug for some cases.
| > And the crash happen only when executing a THROW at the
| > nearly end of an ECL-based OpenAxiom seession.
| Is there an easy (or not so easy way) I can reproduce this?
| Downloading, checking out or building Axiom is ok, AFA you give me a
| finite set of simple steps. Thanks in advance,
(1) Get a copy of OpenAxiom (trunk)
mkdir sandbox && cd sandbox
svn co https://open-axiom.svn.sf.net/svnroot/open-axiom/trunk \
(2) Apply the patch below (conf.diff) to disable the debug workaround
patch -p0 < /path/to/conf.diff
(3) Configure and build OpenAxiom
mkdir oa-build && oa-build
After a while, the build should fail at the end of compiling
OrderedAbelianMonoid (OAMON.spad), with `segmentation fault' on MAC
OS, and `Cannot grow stack size' on x86_64.
If I trace the function Join, by inserting
at the end of the file oa.trunk/src/interp/category.boot, and rebuild
cd src/algebra && make clean && cd -
then the build does not fail on OAMON anymore. It progresses till
failure happens at the end of compiling PrincipalIdealDomain (PID.spad).
If I try to run the failing compilation under GDB, then no failure
happens (at least, I cannot see it in the millions of SIGSEV signals
generated for the GC)
Note that without the patch conf.diff (so debug=3), everything builds
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 2292 bytes
Desc: not available
More information about the ecl-devel