At 4:22am -0500 Thu, 03 Mar 2011, Tor Lillqvist wrote:
for a cherry-pick candidate to libreoffice-3-3. It's a very simple&
safe fix, and fixes a crasher as reported in
But do you know how those values can end up being negative in the
first place, is that a symptom of something being horribly wrong
otherwise, and this patch now then just plasters over that and
instead of crashing it might permit silent data corruption elsewhere
then instead?
Alas, in my limited 20 minutes last night I was not able to get the full
scope in my mind, so I pose it as a question here instead: I'm a little
concerned with /why/ either of those two functions are returning a
negative number at all. (I certainly didn't give LO one in the input!)
Is there ever a time that either of those should be negative? The
rest of the code in getAllTicks() assumes that they're always >= 0.
At first glance, getTickDepth() and getMaxTickCount() read to me as
requests for non-negative values (from an English/math sense). However,
I note that the return values are sal_Int32 instead of sal_uInt32.
Going one call down in the stack trace:
sal_Int32 TickmarkHelper::getTickDepth() const
{
return m_rIncrement.SubIncrements.getLength() + 1;
}
If that's truly a length (where are the frickin' data definitions?!?!),
then, sans a wraparound, that certainly shouldn't be negative. Surely
we aren't at MAX_INT?
sal_Int32 TickmarkHelper::getMaxTickCount( sal_Int32 nDepth ) const
Again, I've run out of time, but my spidey sense is pointing to a
combination of "too small a range" = close enough to zero for the
computer to interpret as 0 or returning a negative "almost 0", and then
line 484 or 495:
484: sal_Int32 nIntervalCount = static_cast<sal_Int32>( fSub /
m_rIncrement.Distance );
I have not yet sussed out what approxSub() or isFinite() does to fSub.
Are they making nIntervalCount negative, through the fSub proxy?
495: nTickCount = nIntervalCount *
(m_rIncrement.SubIncrements[nDepth-1].IntervalCount-1);
If it's 495, I don't know how it's getting past the if above.
Thoughts?
Kevin
_______________________________________________
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice
