eps files created by matplotlib stump evince

[Malek Ghantous]

In Ubuntu 13.04 I created an eps file in matplotlib and when trying to display the image with evince I get:

---
$ evince figure_2.eps
dictstackunderflow -4
dictstackunderflow -4

(evince:16040): EvinceDocument-CRITICAL **: ev_document_misc_pixbuf_from_surface: assertion `surface' failed

(evince:16040): GLib-GObject-CRITICAL **: g_object_unref: assertion `G_IS_OBJECT (object)' failed
dictstackunderflow -4

(evince:16040): EvinceDocument-CRITICAL **: ev_document_misc_pixbuf_from_surface: assertion `surface' failed

(evince:16040): GLib-GObject-CRITICAL **: g_object_unref: assertion `G_IS_OBJECT (object)' failed
dictstackunderflow -4
---

and so on, and the image is not displayed. gv displays the image but throws up an error diaglogue. This only happens with "usetex=True" set. It seems that for whatever reason an extra "end" is added to the eps file, since when I remove it everything seems to work fine. I'm not sure which package is actually responsible for the buggy behaviour - I've not seen this before and have been using the identical python code in the past with no issue.

[Malek Ghantous (malektronic)]

Oops, package name fail.

[WeatherGod (ben-v-root)]

Bug report has been pushed upstream: https://github.com/matplotlib/matplotlib/issues/1997

[Mike Droettboom (mdboom)]

I am unable to reproduce this using matplotlib 1.2.1, 1.2.x (maintenance branch) or master. Can you provide a short example that reproduces the bug, or an example eps file that doesn't display in evince?

[Malek Ghantous (malektronic)]

I've attached a short python file that's supposed to generate a sine
curve, and the output file I get is attached. By removing the "end"
command on line 1497 of the eps file everything works, otherwise it
throws up the errors and doesn't display. gv shows up a different
bunch of errors, though it does display the image:

----
Error: /dictstackunderflow in --end--GPL Ghostscript 9.07:
Unrecoverable error, exit code 1

Operand stack:

Execution stack:
   %interp_exit .runexec2 --nostringval-- --nostringval--
--nostringval-- 2 %stopped_push --nostringval--
--nostringval-- --nostringval-- false 1 %stopped_push 1916
1 3 %oparray_pop 1915 1 3 %oparray_pop --nostringval--
1899 1 3 %oparray_pop 1787 1 3 %oparray_pop
--nostringval-- %errorexec_pop .runexec2 --nostringval--
--nostringval-- --nostringval-- 2 %stopped_push
--nostringval--
Dictionary stack:
   --dict:1170/1684(ro)(G)-- --dict:0/20(G)-- --dict:81/200(L)--
Current allocation mode is local
Last OS error: No such file or directory
Current file position is 104278
----

[Mike Droettboom (mdboom)]

Thanks for the additional info. I am still unable to reproduce this. Do you have any non-standard rcParam settings?

[Malek Ghantous (malektronic)]

I removed everything from the .matplotlib directory and still have the
same issue. As far as I can tell I'm running a pretty stock ubuntu
13.04 install (I still haven't carried over my .profile and other
files from a previous installation), though it might have something to
do with the specific python libraries and latex packages I have
installed. I'm happy to send the output of apt-cache --install or
such if you think it will help. Or is it a ghostscript problem? I
can, on the other hand, produce ps and pdf files with otherwise the
same code with no issue.

2013/5/16 Mike Droettboom <email address hidden>:
> Thanks for the additional info. I am still unable to reproduce this.
> Do you have any non-standard rcParam settings?
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1179376
>
> Title:
> eps files created by matplotlib stump evince
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/matplotlib/+bug/1179376/+subscriptions

[Malek Ghantous (malektronic)]

Also, just to clarify regarding the comments on github, there is only
one "end" statement, but apparently no corresponding "begin" (so it's
not that they are doubled as such).

2013/5/16 Malek <email address hidden>:
> I removed everything from the .matplotlib directory and still have the
> same issue. As far as I can tell I'm running a pretty stock ubuntu
> 13.04 install (I still haven't carried over my .profile and other
> files from a previous installation), though it might have something to
> do with the specific python libraries and latex packages I have
> installed. I'm happy to send the output of apt-cache --install or
> such if you think it will help. Or is it a ghostscript problem? I
> can, on the other hand, produce ps and pdf files with otherwise the
> same code with no issue.
>
> 2013/5/16 Mike Droettboom <email address hidden>:
>> Thanks for the additional info. I am still unable to reproduce this.
>> Do you have any non-standard rcParam settings?
>>
>> --
>> You received this bug notification because you are subscribed to the bug
>> report.
>> https://bugs.launchpad.net/bugs/1179376
>>
>> Title:
>> eps files created by matplotlib stump evince
>>
>> To manage notifications about this bug go to:
>> https://bugs.launchpad.net/matplotlib/+bug/1179376/+subscriptions

[Mike Droettboom (mdboom)]

Can you try setting `rc('ps', distiller='xpdf')` and then `rc('ps', distiller='ghostscript')` and see if that makes any difference?

[Malek Ghantous (malektronic)]

rc('ps',usedistiller='xpdf') works, usedistiller='ghostscript' throws
up the same error (presumably that's the default? I don't have a
.matplotlibrc file).

I've attached the xpdf-distilled file. That workaround solves my
problem (and generates smaller and nicer eps files), but it's still
curious that ghostscript generates a faulty file. Perhaps matplotlib
should just use xpdf by default?

2013/5/17 Mike Droettboom <email address hidden>:
> Can you try setting `rc('ps', distiller='xpdf')` and then `rc('ps',
> distiller='ghostscript')` and see if that makes any difference?
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1179376
>
> Title:
> eps files created by matplotlib stump evince
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/matplotlib/+bug/1179376/+subscriptions

[Thomas Caswell (tcaswell-gmail)]

I can confirm this bug (and the work around) on 13.04, python 2.7, and current master branch of mpl.

[Mike Droettboom (mdboom)]

Yes, it is curious that ghostscript is generating a faulty file. I'm going to summarize these results back over on the matplotlib issue and see if anyone else has any ideas.

[Mike Droettboom (mdboom)]

See

https://github.com/matplotlib/matplotlib/issues/1997

for @leejjoon's resolution of this issue and

https://github.com/matplotlib/matplotlib/pull/2033

for a fix.

The patch in issue #2033 applies cleanly to 1.2.1 so Ubuntu could include the patch in its .deb (theoretically), and will make it into the next maintenance release (1.2.2) and major release (1.3.0).