Thu 19 May 2016 03:44:58 PM UTC, comment #12:
Works for me
hg id 71d8a9a0642c
|
Thu 19 May 2016 01:58:06 PM UTC, comment #11:
Yep, the problem with system.tst is a bug in octave::sys::umask. See the patch at #47961.
|
Thu 19 May 2016 01:56:29 PM UTC, comment #10:
The patch fixes the bug for me, thanks Lachlan, pushed,
http://hg.savannah.gnu.org/hgweb/octave/rev/767c81fe30ba
|
Thu 19 May 2016 01:34:02 PM UTC, comment #9:
Andrew: Tried the new patch which applied OK.
But, same hang as before in comment #6
It could be that the hang is due to one of the patches
since yesterday, so your patch may still be useful.
|
Thu 19 May 2016 12:11:07 PM UTC, comment #8:
Yes, I just noticed that JWE has added some namespace stuff.
The new patch is attached.
(file #37207)
|
Thu 19 May 2016 10:12:09 AM UTC, comment #7:
Sorry, patch fails on current system.
oct-parse.in.yy was updated last night.
|
Thu 19 May 2016 10:03:38 AM UTC, comment #6:
Andrew: After last night's updates the problem, for me,
has changed to a hang while processing:
system.tst ..................................................^C^C^Cpanic: Interrupt -- stopping myself...
I will try your patch just to see...
|
Thu 19 May 2016 09:59:50 AM UTC, comment #5:
I've found something that seems wrong in the code, which is fixed by the attached cset. Since I don't see the error messages on my system, I can't check whether or not it gets rid of the error.
Valgrind doesn't report any problems when sourcing either a function that is already loaded or one not in the path. It reports leaks when a breakpoint is hit, but it reports about the same amount of leaked memory whether or not the script is executed via "source".
(file #37204)
|
Thu 19 May 2016 08:13:15 AM UTC, comment #4:
I see essentially the same error. This is on Fedora 23
and gcc: gcc (GCC) 5.3.1 20160406 (Red Hat 5.3.1-6)
and generally the current Fedora 23 stuff.
|
Thu 19 May 2016 02:31:20 AM UTC, comment #3:
Thanks for reporting that. I'm not seeing that on my system, perhaps because I'm using a fairly old gcc.
I'll look into it.
|
Tue 17 May 2016 07:59:58 PM UTC, comment #2:
And I can confirm Doug's finding. If I backout cset http://hg.savannah.gnu.org/hgweb/octave/rev/4c080cbc4ef9 everything works as before. This was a fix for bug #33411, adding Lachlan in cc.
|
Tue 17 May 2016 07:30:26 PM UTC, comment #1:
I wasn't seeing this with 'make check', but now I can confirm this with something as simple as
Both octave-cli and octave-gui show the same "corrupted unsorted chunks" error on exit.
|
Mon 16 May 2016 10:45:07 PM UTC, original submission:
I get the following errors at the end of a make check.
I bisected and found that 4c080cbc4ef9 is giving me the errors.
When I back out 4c080cbc4ef9 then all is OK up to 09e1bb792dd9
I am on Ubuntu 15.10
|