Opened 12 years ago

Closed 6 years ago

#9529 closed defect (wontfix)

in, sometimes get "libpng error: Image width or height is zero in IHDR"

Reported by: rlm Owned by: jason, ncohen, rlm
Priority: major Milestone: sage-duplicate/invalid/wontfix
Component: graph theory Keywords:
Cc: chapoton, tscrim Merged in:
Authors: Reviewers: Travis Scrimshaw
Report Upstream: N/A Work issues:
Branch: Commit:
Dependencies: Stopgaps:

Status badges


To reproduce this, you can do something like the following in bash:

sage-4.5.rc1/devel/sage-main$ for i in `seq 1 20`;
> do
>     ../../sage -t sage/graphs/
> done

Change History (10)

comment:1 Changed 12 years ago by rlm

See also #5906.

comment:2 Changed 12 years ago by rlm

File "/scratch/rlmill/sage-4.5.rc1/devel/sage-main/sage/graphs/", line 267:
    sage: GP.plot()
Exception raised:
    Traceback (most recent call last):
      File "/scratch/rlmill/sage-4.5.rc1/local/bin/", line 1231, in run_one_test
        self.run_one_example(test, example, filename, compileflags)
      File "/scratch/rlmill/sage-4.5.rc1/local/bin/", line 38, in run_one_example
        OrigDocTestRunner.run_one_example(self, test, example, filename, compileflags)
      File "/scratch/rlmill/sage-4.5.rc1/local/bin/", line 1172, in run_one_example
        compileflags, 1) in test.globs
      File "<doctest __main__.example_4[13]>", line 1, in <module>
        GP.plot()###line 267:
    sage: GP.plot()
      File "/scratch/rlmill/sage-4.5.rc1/local/lib/python/site-packages/sage/misc/", line 174, in displayhook   
        print_obj(sys.stdout, obj)
      File "/scratch/rlmill/sage-4.5.rc1/local/lib/python/site-packages/sage/misc/", line 142, in print_obj
        print >>out_stream, `obj`
      File "sage_object.pyx", line 101, in sage.structure.sage_object.SageObject.__repr__ (sage/structure/sage_object.c:1370) 
      File "/scratch/rlmill/sage-4.5.rc1/local/lib/python/site-packages/sage/plot/", line 915, in _repr_
      File "/scratch/rlmill/sage-4.5.rc1/local/lib/python/site-packages/sage/plot/", line 1437, in show, **options)
      File "/scratch/rlmill/sage-4.5.rc1/local/lib/python/site-packages/sage/plot/", line 1973, in save
      File "/scratch/rlmill/sage-4.5.rc1/local/lib/python/site-packages/matplotlib/", line 1032, in savefig
        self.canvas.print_figure(*args, **kwargs)
      File "/scratch/rlmill/sage-4.5.rc1/local/lib/python/site-packages/matplotlib/", line 1455, in print_figure
      File "/scratch/rlmill/sage-4.5.rc1/local/lib/python/site-packages/matplotlib/backends/", line 366, in print_png
        filename_or_obj, self.figure.dpi)
    RuntimeError: Error building image

comment:3 Changed 12 years ago by cwitty

To reproduce this fairly quickly, type:

sage.plot.plot.DOCTEST_MODE = True
while True:
    d = DiGraph({}, loops=True, multiedges=True, sparse=True)
    GP = d.graphplot(vertex_size=100, edge_labels=True, color_by_label=True, edge_style='dashed')

at a sage: prompt.

I spent some time looking into this; here are the results I found before I gave up.

We draw our plots with bbox_inches='tight', which means:

            Bbox in inches. Only the given portion of the figure is
            saved. If 'tight', try to figure out the tight bbox of
            the figure.

There's also a parameter pad_inches:

            Amount of padding around the figure when bbox_inches is

matplotlib gets its "tight bbox" with the appropriately-named get_tightbbox() function, in src/lib/matplotlib/ Its docstring proclaims,

        It only accounts axes title, axis labels, and axis
        ticklabels. Needs improvement.

and it's true, it does need improvement.

Our graph_plot figure has "empty" axes (no title, labels, or ticklabel), but it does have an axes object with a size, which gets used for the tight bbox. I believe that size comes indirectly from get_minmax_data in sage/plot/, but I'm not sure.

I can think of a few possibilities to fix the problem, ordered from best to worst.

1) Fix matplotlib get_tightbbox() to actually be a bounding box for the drawing, not just for the "axes". This would definitely be best... it would also fix problems like graphs with loops going off the edge of the plot, etc.

2) Adjust our get_minmax_data functions to be closer to a correct bounding box. For instance, currently get_minmax_data on a Text object only looks at the center of the text, and doesn't take the size into account. There's a limit to how good a job we could do, though... we don't want to start digging through font metrics, etc.

3) Set pad_inches to make all our plots "a little bit bigger". The downside, of course, is that this simple approach would affect *all* our plots.

Like I said, I'm giving up now.

comment:4 Changed 12 years ago by jason

See #9211, where similar issues are dealt with. With the new matplotlib at #9221, we can add some extra artists whose bounding boxes should be included in the tight bbox calculations (so, for example, we should be able to add the text boxes and then the tight bbox calculations would include those bounds as well).

comment:5 Changed 12 years ago by rlm

  • Component changed from algebra to graph theory
  • Owner changed from AlexGhitza to jason, ncohen, rlm

comment:6 Changed 6 years ago by jmantysalo

  • Cc chapoton added
  • Milestone set to sage-duplicate/invalid/wontfix
  • Status changed from new to needs_review

I can not reproduce this, and this is six years old. I suppose this one can be closed as the code has been changed many times. Frédéric, do you agree?

comment:7 Changed 6 years ago by jmantysalo

  • Cc tscrim added

Travis, do you agree that we can close this?

comment:8 Changed 6 years ago by jmantysalo

Just pinging this wontfix-confirmation.

comment:9 Changed 6 years ago by tscrim

  • Reviewers set to Travis Scrimshaw
  • Status changed from needs_review to positive_review

I can't reproduce it either.

comment:10 Changed 6 years ago by embray

  • Resolution set to wontfix
  • Status changed from positive_review to closed

Determined to be invalid/duplicate/wontfix (closing as "wontfix" as a catch-all resolution).

Note: See TracTickets for help on using tickets.