On 31/10/12 17:01, Tor Lillqvist wrote:
The SLED11 tinderbox has not failed yet since that change either, so I guess
you got the ugly bug. Congratulations and thanks :).
Now if only the commit message had been a bit more descriptive...
and that was after a promise made to provide better comment detail at
the opensuse conference ( after a discussion about someone else's even
briefer comments ) where I admitted also being a serial offender in that
regard
failing that, could you Noel explain what was going on, and how the
commit fixes it? (Yeah, I probably should be able to understand if
from reading the commit, but...)
I did realise the commit message was a little brief but it was too late,
anyway I tried to make up for the lack of commit description in the mail
but I guess I still didn't do a good enough job ;-), basically on export
an object id used to be created from a pointer like so
sStorageName.append('_').append(reinterpret_cast<sal_Int64>(pObj));
where sStorageName is the name of a 'folder' in the binary format
that object id was also inserted into some table in the binary format as
follows
Set_UInt32(pData,(sal_uInt32)(sal_uIntPtr)pObj) );
so basically a 64bit number was stored as a 32bit number. Because the
64bit number in question was actually a pointer it seems that mostly the
address it held was not large enough to cause trouble. The occasional
failure/core we were seeing was down to the fact on import the object id
extracted from the binary format had some value that didn't match the
'folder' name
Noel
_______________________________________________
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice
