Mail Archives: geda-user/2021/07/06/09:13:51
On Sat, 3 Jul 2021, karl AT aspodata DOT se [via geda-user AT delorie DOT com] wrote:
> Roland Lutz:
>> This looks a lot like the the issue you reported in May, so I think
>> memcheck may give a clue what's going on here.
>
> Certanly does; maybe it easier to solve with a text mode program.
>
> $ valgrind --tool=memcheck --trace-children=yes /usr/local/bin/gaf export -o test2.pdf test.sch > gaf_memcheck.log 2>&1
>
> If you are interested, the log is available in:
> http://aspodata.se/tmp/Test/gaf_pdf/
Thank you for providing the memcheck output! This revealed another use of
uninitialized variables which I missed last time; that's fixed now. But
apart from that, there's no clue on how the error could happen. If fact,
I find the *absence* of any clue rather disturbing.
> ** (export:9391): CRITICAL **: 17:18:42.168:
> o_text_get_font_size_in_points: assertion 'object != NULL' failed
I simply have no idea how this condition could be reached. There are only
two code paths that lead to this place:
- eda_renderer_draw_text (renderer, object)
- eda_renderer_get_text_user_bounds (renderer, object, ...more args...)
Both are called via functions which check for object == NULL, so except
for memory corruption (which should have been caught by memcheck), I don't
see what could have led there.
> (export:9391): libgedacairo-CRITICAL **: 17:18:42.216: file
> edarenderer.c: line 585 (eda_renderer_draw_hatch): should not be reached
A similar thing here. I only found one way for object->fill_type to have
an invalid value (when reading an invalid value from file), which I fixed
last time. So, I'm at a loss how this could happen.
Are you sure you have been using the latest Git `master'?
The only obvious thing that's different about your setup is the fact that
you are reading gedasymbols.org as a symbol library. Have you tried if
renaming your configuration files makes the errors go away?
Roland
- Raw text -