[Libreoffice] Class Time and its funny ideas about time

[Michael Meeks]

Hi Terrence,

On Thu, 2011-08-18 at 14:27 -0400, Terrence Enger wrote:
> I just wish I could stop thinking about ways to
> attack the problem. ... I thought of another way
> even as I was writing the previous sentence.

	:-)

> One SEGV has happened only with an earlier build.  I do have
> a backtrace.  Is it still of interest?

	Of course; LibO should never crash.

> Two SEGVs are in my latest build (master aa51fd4), but they
> do not happen again with grossly similar user actions.  I do
> have backtraces.  I even have a typescript of the entire gdb
> session.  Given my inability to make them happen again, are
> they of interest?

	Certainly; of course, if it is intermittent - it could well be memory
corruption related; if you can bear the 50-fold performance hit -
running under valgrind would be time well spent; ensure you:
export G_SLICE=always # first though. That may well give you accurate
data for the real bug; how well it works with java-ness is unclear
though.

	Very good of you to continue chasing it; and of course much of the
database stuff is -really- hard to get setup and/or test meaningfully. I
keep hoping for Wols to finish the sqlite backend and to get that
integrated, so we can do some fast/native unit tests for database bits.

	HTH,

		Michael.

-- 
 michael.meeks at novell.com  <><, Pseudo Engineer, itinerant idiot