bugGNU Octave - Bugs: bug #55647, crash, hang, error or corrupted...


bug #55647: crash, hang, error or corrupted output when printing some demos to EPS

Submitted by:  Pantxo Diribarne <pantxo>
Submitted on:  Tue 05 Feb 2019 10:12:22 PM UTC  
Category:  Plotting with OpenGL Severity:  4 - Important
Priority:  5 - Normal Item Group:  Segfault, Bus Error, etc.
Status:  Confirmed Assigned to:  None
Originator Name:  Open/Closed:  Open
Release:  dev Operating System:  GNU/Linux

Add a New Comment (Rich Markup)

You are not logged in

Please log in, so followups can be emailed to you.


( Jump to the original submission)

Wed 11 Nov 2020 02:52:51 PM UTC, comment #11: 

Sorry. Got this wrong: We postponed defaulting to "-tight" for bug #48689. That bug does not seem to be resolved yet, too.
So the conclusion is the same: Should we revert to "-loose" again?

Markus Mützel <mmuetzel>
Project Member
Wed 11 Nov 2020 02:42:50 PM UTC, comment #10: 

I'm wondering if we should postpone defaulting to "-tight" for another version (or until this is resolved).

We did this for the same reason for Octave 5.

Markus Mützel <mmuetzel>
Project Member
Wed 11 Nov 2020 02:11:37 PM UTC, comment #9: 

While unfortunate, I don't think this particular bug should hold up the Octave release.  It is now been ~20 months since the 5.1 release.  We really need to get this out.

Rik <rik5>
Project Administrator
Wed 11 Nov 2020 01:01:38 PM UTC, comment #8: 

The original example leads to intermittent hangs (of gs.exe) and ghostscript errors with Octave 6.0.92 on Windows for me.

Do we need to take any action before Octave 6 will be released?

Markus Mützel <mmuetzel>
Project Member
Tue 26 Feb 2019 02:31:26 AM UTC, comment #7: 

Since the default was reverted back to "-loose" for the 5.1 release, this bug no longer affects Octave 5.1, right? Marking as "dev".

Mike Miller <mtmiller>
Project Administrator
Wed 06 Feb 2019 10:22:06 PM UTC, comment #6: 

@Rik: "axis auto" is OK even though it will at most avoid a crash for us, testers and developers.

Unfortunately bug #48689 is a really serious bug:

  • it affects most users: Mac, MXE, Linux with Gallium driver
  • it is easy to trigger: a single triangle drawn partly outside the axes border is enough to get a very bad vector output.
  • it is actually a bug in MESA gallium driver: colors reported in the feedback buffer to gl2ps are corrupted. I don't think anyone reported this bug to MESA but I fear it won't get much attention since it touches the deprecated GL_FEEDBACK mode.
Pantxo Diribarne <pantxo>
Project Member
Wed 06 Feb 2019 08:03:31 PM UTC, comment #5: 

As a short term fix, would it be acceptable to add "axis auto" in hold.m's BIST demo #6?

This seems like a subtle enough problem that it may not be fixed until the first bug fix release for 5.1.

Rik <rik5>
Project Administrator
Wed 06 Feb 2019 07:44:21 PM UTC, comment #4: 

@Rik: Yes the problem is present in default and may have been there for a while: before Octave 5, the default was "-loose" so we did not use epstool and probably just didn't notice the misbehavior.

@Dmitri: epstool is ran through a pipe which is fed directly with the original gl2ps-generated EPS output. Then when closing the pipe, fclose only returns when epstool has finished. So I am pretty sure there cannot be a race condition here.

Now about the second issue, the bad bbox when using "-loose", can one of you reproduce it? Actually I am only able to trigger it on my most recent system (linux Mint 19.1, whith Ghostscript 9.26 among others). On debian stretch the bbox is OK after

demo hold 6
axis auto
print -loose -depsc toto.eps

Pantxo Diribarne <pantxo>
Project Member
Wed 06 Feb 2019 05:26:54 PM UTC, comment #3: 

Yes. I am testing against stable tip.
eb1d4a619260 (stable) tip
just now hanged at ii=3.
But subsequent run manged to finish all 100 loops.


Dmitri A. Sergatskov <dasergatskov>
Wed 06 Feb 2019 04:26:36 PM UTC, comment #2: 

Is this a problem for the upcoming 5.1 release as well?  The bug is marked against the development version of Octave.

Rik <rik5>
Project Administrator
Tue 05 Feb 2019 10:54:52 PM UTC, comment #1: 

Any possibility of race condition? Like sending eps file that is still being written to epstool?


Dmitri A. Sergatskov <dasergatskov>
Tue 05 Feb 2019 10:12:22 PM UTC, original submission:  

"demo hold 6" quite reliably causes problems when printing to EPS format (see bug #55529)

for ii=1:100
  demo hold 6;
  ## axis auto
  print toto.eps -depsc ## -loose
  close all

This leads to either a crash of Octave, a hang with Ghostscript consuming 100% CPU or a simple error due to Ghostscript returning a bad status.

This example doesn't crash, hang or error out if I use "axis auto" prior to printing, so that bug #48689 doesn't come into play. Though, the output figure has a wrong bounding box. See attached "bad_bbox.eps".

This example works perfectly well when using the "-loose" print option (which used to be the default in 4.4). So it looks like epstool has problems cropping some of our gl2ps generated EPS files.

Pantxo Diribarne <pantxo>
Project Member


(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach Files:

Attached Files
file #46188:  bad_bbox.eps added by pantxo (252KiB - image/x-eps)


Depends on the following items: None found

Items that depend on this one: None found


Carbon-Copy List
  • -email is unavailable- added by mmuetzel (Posted a comment)
  • -email is unavailable- added by rik5 (Posted a comment)
  • -email is unavailable- added by dasergatskov (Posted a comment)
  • -email is unavailable- added by pantxo (Submitted the item)

    Do you think this task is very important?
    If so, you can add your encouragement to it.
    This task has 0 encouragements so far.

    Only project members can vote.




    Follow 4 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2019-02-26 mtmiller Release5.0.91 => dev
    2019-02-06 rik5 Releasedev => 5.0.91
    2019-02-06 pantxo StatusNone => Confirmed
    2019-02-05 pantxo Attached File- => Added bad_bbox.eps, #46188

    Back to the top

    Powered by Savane 3.5