Wed 14 May 2014 09:04:15 PM UTC, comment #10:
This should be fixed by this changeset on the development branch (http://hg.savannah.gnu.org/hgweb/octave/rev/b45d975aa38f). The "problem" is that the assert statements were written with the assumption that profile would always be called from the top-level command line. If profile() is used within an m-file to enable/disable profiling then the asserts would fire and cause the crash. My solution was simply to comment out the incorrect assert statements in the code.
|
Wed 27 Nov 2013 08:08:10 PM UTC, comment #9:
Confirmed, for what it's worth it does not happen with qt graphics toolkit.
|
Wed 27 Nov 2013 08:06:23 PM UTC, comment #8:
Oops, "graphics_toolkit fltk" is precisely what my .octaverc contains at the top. Sorry for failing to provide that crucial piece of information, as it turned out now!
(Oh, and in passing, thanks to whomever fixed the awful ~0.5-second delays when pressing the prompt history up/down keys after loading FLTK.)
|
Wed 27 Nov 2013 04:59:52 PM UTC, comment #7:
Even simpler series of commands now to reproduce the bug.
|
Wed 27 Nov 2013 05:01:24 AM UTC, comment #6:
I found a situation that fails for me. If the .octaverc file contains
then running
If there is no such line in the .octaverc or I call "run-octave -f" which does not read initialization scripts then it works.
|
Sun 13 Oct 2013 05:33:17 PM UTC, comment #5:
I am now at (hg summary)
parent: 17646:993ea21b7e03 tip
graphics.in.h: Text "edgecolor" property should default to "none".
branch: default
bookmarks: @
commit: (clean)
update: (current)
mq: (empty queue)
and the test works for me even if I don't do anything before in a session (gui or not). My configure command is:
I use gcc 4.7.2 on linux mint 14.
|
Fri 11 Oct 2013 04:48:57 PM UTC, comment #4:
Using this tip (hg id 382aaf2d533a) I still get a segfault. Moreover, I get a different result from profile.
There is no call to 'binary +' for example. So, this might have to do with compilation options. Also, I was executing the two commands as the very first thing in an Octave session, perhaps things need to be initialized to work properly.
|
Fri 11 Oct 2013 03:56:55 PM UTC, comment #3:
Can't confirm with up to date dev.
It may have been fixed at the same time as this one https://savannah.gnu.org/bugs/index.php?39524
|
Thu 08 Aug 2013 12:56:36 AM UTC, comment #2:
Confirmed on a recent tip. The profiler has ceased to function well since moving to the new parser.
|
Sun 28 Jul 2013 07:55:50 AM UTC, comment #1:
[Oops, the rest of my initial report was eaten.]
octave:2> prof_test(1)
octave:3> profexplore(profile("info"))
Top
1) : 1 calls, 12.241 total, 0.100 self
2) zeros: 1 calls, 0.005 total, 0.005 self
3) binary -: 1000 calls, 0.001 total, 0.001 self
4) binary +: 1000 calls, 0.000 total, 0.000 self
5) binary ==: 1 calls, 0.000 total, 0.000 self
profexplore> 1
../../libinterp/corefcn/profiler.cc:108: profile_data_accumulator::tree_node* profile_data_accumulator::tree_node::exit(octave_idx_type): Assertion `fcn_id == fcn' failed.
panic: Aborted -- stopping myself...
|
Sun 28 Jul 2013 07:54:36 AM UTC, original submission:
To reproduce on the default branch,
|