Re: [Libreoffice] [REVIEW] fix for a chart crasher

On Thu, 2011-03-03 at 02:22 -0700, 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?

Good point. I wondered about the same thing.

I didn't exactly look into why they (actually only one of them) ended up
negative, but from the logic of the code those values should never be
negative anyway as they represent the size of something & are used to
allocate arrays to their size.  So, even if I knew the reason for the
negative value & fixed it, I would still test for those negative values
(and they are signed integers, not unsigned).

Also, the possibility of silent data loss is unlikely, since this code
provides data for individual tick marks that are automatically
generated, and they are not stored with the document but are generated
from the min, max, and interval data (which are stored with the
document).

So, we should definitely look into why the negative value occurs in the
first place, but this change is still necessary regardless of the
outcome.

Kohei

-- 
Kohei Yoshida, LibreOffice hacker, Calc
<kyoshida@novell.com>