IOError: cannot identify image file

[gagern]

With Sage 6.6 beta 3 and using Sage Notebook I get a backtrace when I want to see a sphere:

sage: sphere()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "_sage_input_3.py", line 10, in <module>
    exec compile(u'open("___code___.py","w").write("# -*- coding: utf-8 -*-\\n" + _support_.preparse_worksheet_cell(base64.b64decode("c3BoZXJlKCk="),globals())+"\\n"); execfile(os.path.abspath("___code___.py"))
  File "", line 1, in <module>
    
  File "/tmp/tmpWqCNEl/___code___.py", line 2, in <module>
    exec compile(u'sphere()
  File "", line 1, in <module>
    
  File "sage/local/lib/python2.7/site-packages/sage/repl/rich_output/display_manager.py", line 716, in displayhook
    return self._backend.displayhook(plain_text, rich_output)
  File "sage/local/lib/python2.7/site-packages/sage/repl/rich_output/backend_base.py", line 432, in displayhook
    return self.display_immediately(plain_text, rich_output)
  File "sage/local/lib/python2.7/site-packages/sage/repl/rich_output/backend_sagenb.py", line 361, in display_immediately
    rich_output.embed()
  File "sage/local/lib/python2.7/site-packages/sage/repl/rich_output/backend_sagenb.py", line 258, in embed
    self.save_preview()
  File "sage/local/lib/python2.7/site-packages/sage/repl/rich_output/backend_sagenb.py", line 237, in save_preview
    self.preview_png.save_as(self.preview_filename())
  File "sage/misc/cachefunc.pyx", line 2209, in sage.misc.cachefunc.CachedMethodCallerNoArgs.__call__ (build/cythonized/sage/misc/cachefunc.c:13699)
  File "sage/local/lib/python2.7/site-packages/sage/repl/rich_output/backend_sagenb.py", line 189, in preview_filename
    directory, filename = os.path.split(self._base_filename())
  File "sage/misc/cachefunc.pyx", line 2209, in sage.misc.cachefunc.CachedMethodCallerNoArgs.__call__ (build/cythonized/sage/misc/cachefunc.c:13699)
  File "sage/local/lib/python2.7/site-packages/sage/repl/rich_output/backend_sagenb.py", line 147, in _base_filename
    filename = self.sagenb_launch_script_filename()
  File "sage/misc/cachefunc.pyx", line 2209, in sage.misc.cachefunc.CachedMethodCallerNoArgs.__call__ (build/cythonized/sage/misc/cachefunc.c:13699)
  File "sage/local/lib/python2.7/site-packages/sage/repl/rich_output/backend_sagenb.py", line 122, in sagenb_launch_script_filename
    width, height = PIL.Image.open(StringIO(self.preview_png.get())).size
  File "build/bdist.linux-x86_64/egg/PIL/Image.py", line 2006, in open
IOError: cannot identify image file

Using strace I can see access to some preview.png and looking at that file I notice that it appears to be a regular PNG, concatenated with a Jmol script. So apparently two outputs ended up in the same data stream. This is likely a consequence of #17234.

Upstream: Workaround found; Bug reported upstream.

CC: @vbraun

Component: user interface

Keywords: strange

Author: Martin von Gagern

Branch/Commit: 3818998

Reviewer: Volker Braun

Issue created by migration from https://trac.sagemath.org/ticket/17940

[vbraun]

comment:1

Works for me. Can you post your strace output? Can PIL open a PNG at all on your system?

[gagern]

Replying to @gagern:

Using strace I can see access to some preview.png and looking at that file I notice that it appears to be a regular PNG, concatenated with a Jmol script. So apparently two outputs ended up in the same data stream. This is likely a consequence of #17234.

It seems I was misled: apparently JMol attaching its script to the generated PNG is normal behavior. The PNG file is apparently broken at some other point. Looking at the PNG chunks as reported by the pngchunks tool, I see that the IDAT chunk appears to be off by one byte. The chunk before that is an iTXt chunk stating the creation date. It does so in a locale-dependent manner, and in my case that means using the German abbreviation for March, which is Mär. So I'd say JMol is counting UTF-16 code units when it should be counting UTF-8 code units. This is a bug in JMol, and probably can be reproduced at least this month by using a German locale.

I'll try to see whether we can work around this issue by forcing the locale to C, like LC_ALL=C or some Java-specific analogon to this.

[gagern]

Branch: u/gagern/ticket/17940

[gagern]

Commit: 3818998

[gagern]

Author: Martin von Gagern

[gagern]

comment:4

Setting the locale to C works as expected.

---

New commits:

3818998

Trac #17940: Execute JMol in C locale.

[vbraun]

Reviewer: Volker Braun

[vbraun]

comment:5

Good detective job ;-)

Did you report it upstream?

[gagern]

comment:6

Replying to @vbraun:

Did you report it upstream?

Did so now: https://sourceforge.net/p/jmol/bugs/570/

[gagern]

Upstream: Workaround found; Bug reported upstream.

[jdemeyer]

comment:7

A bug which happens only in March :-)

This should be added to the "strange bug list" together with #10609 and #12221.

[gagern]

Changed keywords from none to strange

[gagern]

comment:8

Replying to @jdemeyer:

This should be added to the "strange bug list" together with #10609 and #12221.

Done here, shoudl I do the others as well?

[vbraun]

Changed branch from u/gagern/ticket/17940 to 3818998