On 2015-11-03 00:56, Michael Pyne wrote:
On Tue, November 3, 2015 00:40:58 Albert Astals Cid wrote:
Someone went a bit too far in the "port away from time_t to QDateTime"
and
changed the timestamp member of the KPixmapCacheIndexHeader struct.
According to mpyne and thiago this is a big no-no.
The reason it's a big no-no is because KPixmapCacheIndexHeader is meant
to be
held and used from mmap()'d shared memory, which generally requires the
use of
plain old data (POD) types only, to avoid inadvertent constructor or
destructor calls once the memory is freed.
QDateTime will (and is) leak applications to crash so I'd +1 the
recommendation to change back to time_t (or at least some kind of
integral
type), and the type is internal-only so there's no API concern. Just
don't
forget to bump the defined KPIXMAPCACHE_VERSION so that old caches are
flushed
;).
Also, in practice, I believe time_t is everywhere we care about - we
just have to take a little care about what API calls we use.
Alex
_______________________________________________
Kde-frameworks-devel mailing list
Kde-frameworks-devel@kde.org
https://mail.kde.org/mailman/listinfo/kde-frameworks-devel
