[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: segfault in bsxfun during "make check"
From: |
PhilipNienhuis |
Subject: |
Re: segfault in bsxfun during "make check" |
Date: |
Thu, 7 Mar 2019 12:53:07 -0600 (CST) |
John W. Eaton wrote
> On 3/6/19 5:49 PM, John W. Eaton wrote:
>> On 3/6/19 5:36 PM, John W. Eaton wrote:
>>> On 3/6/19 4:41 PM, PhilipNienhuis wrote:
>>>> During "make check" I get a segfault in bsxfun:
>>>>
>>>> :
>>>> libinterp/corefcn/bitfcns.cc-tst ...............................
>>>> PASS
>>>> 60/60
>>>> libinterp/corefcn/bsxfun.cc-tst
>>>> ................................fatal:
>>>> caught signal Segmentation fault -- stopping myself...
>>>> /bin/sh: line 1: 14215 Segmentation fault (core dumped) /bin/sh
>>>> ../run-octave --no-init-file --silent --no-history -p
>>>> /home/philip/devel/octdev/dev600/test/mex
>>>> /home/philip/devel/octdev/dev600/../dev/test/fntests.m
>>>> /home/philip/devel/octdev/dev600/../dev/test
>>>> Makefile:27988: recipe for target 'check-local' failed
>>>> make[3]: *** [check-local] Error 139
>>>> :
>>>>
>>>> address@hidden dev]$ hg summary
>>>> parent: 26855:156bd21bf6bd
>>>> editor follows XDG_CONFIG_HOME for autocompletion files (bug ##55822)
>>>> branch: default
>>>> bookmarks: *@
>>>> commit: (clean)
>>>>
>>>> Mageia-6 Linux
>>>>
>>>> Anyone else seeing this?
>>>
>>> It doesn't crash for me.
>>>
>>> I'm guessing it's due to my recent changes to function handles.
>>>
>>> Can you try running the test suite with gdb?
>>>
>>> Or, can you try building with --enable-address-sanitizer-flags and see
>>> if that gives any clues about memory errors? You may also need to use
>>> --disable-java and --disable-docs to make that work.
>>
>> OK, I built with the address sanitizer and I see the following when
>> running "make check". I'll try to take a look at this problem later
>> tonight.
>
> I'm attaching a patch you can experiment with.
>
> jwe
>
> diffs.txt (4K)
> <http://octave.1599824.n4.nabble.com/attachment/4691913/0/diffs.txt>
Thanks John, that patch did the trick, "make check" finishes.
There's now one FAIL I didn't see before in ode15i. No idea if it's related,
excerpt from fntests.log follows:
>>>>> processing /home/philip/devel/octdev/dev/scripts/ode/ode15i.m
***** testif HAVE_SUNDIALS
saved_opts = warning ();
warning ("off", "all");
opt = odeset ("Jacobian", "_5yVNhWVJWJn47RKnzxPsyb_");
fail ("[t, y] = ode15i (@rob, [0, 4e6], [1; 0; 0], [-1e-4; 1e-4; 0], opt)",
"invalid value assigned to field 'Jacobian'");
warning (saved_opts);
!!!!! test failed
expected error <invalid value assigned to field 'Jacobian'>
but got <nargin: FCN must be a string or function handle>
Philip
--
Sent from: http://octave.1599824.n4.nabble.com/Octave-Maintainers-f1638794.html