On Tue, 1 Jan 2002, Garst R. Reese wrote:
> Michael Koziarski wrote:
>
> > Bug 134:
> > Summary: Crash clicking on old-graphics in float.
> > URL: http://cvs.koziarski.com/show_bug.cgi?id=134
> >
> > The instructions are pretty simple to reproduce it.
> >
> I would not call this bug a showstopper. It requrires an old graphics in
> a minipage with a table. I have lots of old-graphics in floats.
> Garst
Actually, it also works with the new graphics inset and you don't have
to have the figure and table in a minipage; a footnote also works.
That took all of 10 seconds to verify. Other cases are likely to
exist -- although the backtraces are different for most of these
cases.
If you read deeper than the surface layer of "minipage", "footnote",
"graphics (old/new)" and "table" it should become clear that there is
something nasty lurking below.
A better arguement for letting this bug slip through might be that it
also exists in most forms within 1.1.6. But I would argue that is
even more reason why we should fix it before 1.2.0 (not necessarily
before 1.2.0pre1).
One group of crashes results in an abort with:
"pure virtual function called"
printed to the console while the other group result in a segfault.
All seem to be triggered by some access to the_locking_inset in
various classes. I haven't managed to narrow it down any better than
that.
I meant to spend most of the last week investigating this but it's
been to hot :( And I spent the last two days (almost cool at temps of
low 30s Celcius) updating patches to diffutils-2.7.6. I'll post those
later once I've completed the docs for the now working 'ChangeLog'
patch. John, you were close but I ended up with a very different
implementation to what you did.
BTW, the unbreakable prepend magic is:
@@ -0,0 +1,x @@
Allan. (ARRae)
