Re: RE : [matplotlib-devel] crash on matplotlib.matlab import on win32 - solution and question...

[John H. <jdh...@ac...>]

>>>>> "Gregory" == Gregory Lielens <gre...@ff...> writes:


    Gregory> As you see, it crashes on BATANG.TTF, and this font file
    Gregory> is not shown in the font manager and is huge (15 Mb while
    Gregory> typical ones are aroung 100 Kb). I have to check if this
    Gregory> is a valid font file (a priori no, but could be an asian
    Gregory> font)...So now my guess is that adding a non-valid font
    Gregory> (or strange one) cause an error in one of the routines
    Gregory> that parses font files, or exceed some size somewhere,
    Gregory> causing an unclean exit..I will check what happen if I
    Gregory> remove this singular file, and check if this is really a
    Gregory> font or not...  I will give the info as soon as I have
    Gregory> it, but for now it seems this is not a real bug, at most
    Gregory> a lack of robustness when some strange files with TTF
    Gregory> extensions are present in the font directories...so not
    Gregory> sure it is worth considering, as it is more a question of
    Gregory> my windows font directory beeing corrupted :-)

    Gregory> Best regards, and thanks for your king help!

Interesting - perhaps some exception is thrown in the FT2Font
constructor because of the large file size) and this occurs before CXX
is fully set up and its  ref count is out of whack.  So when the
destructor is called because of the exception, the counts are bad.

If there is a memory or resource related problem, perhaps the
difference in the numarray and numeric builds is related to a memory
footprint of those two packages.  Really just free-styling here.

In my casual tests,  numarray does occupy a bit more memory than
Numeric does.  

If you delete this crazy 15M font file, can you run matplotlib w/o
errors?

JDH