[Marble-devel] leak in GeoDataLineString

[Torsten Rahn]

Thanks.

I'll look at it today (Friday).

Torsten

Am Donnerstag 23 Juli 2009 23:20:29 schrieb Thibaut GRIDEL:
> Hi,
>
> I have identified a leak in
> GeoDataLineStringPrivate::toDateLineCorrected()
>
> but am unable to understand the logic so far, so am just referring this
> to you.
>
> ==21056== 321,036 (10,608 direct, 310,428 indirect) bytes in 663 blocks
> are definitely lost in loss record 383 of 410
> ==21056==    at 0x4025A2E: operator new(unsigned int)
> (vg_replace_malloc.c:224)
> ==21056==    by 0x5064656:
> Marble::GeoDataLineStringPrivate::toDateLineCorrected(Marble::GeoDataLineSt
>ring const&, QVector<Marble::GeoDataLineString*>&) (in
> /home/thibaut/sandbox/kdeedu/marble/build/local/lib/libmarblewidget.so.0.7.
>0) ==21056==    by 0x50647F5:
> Marble::GeoDataLineString::toDateLineCorrected() const (in
> /home/thibaut/sandbox/kdeedu/marble/build/local/lib/libmarblewidget.so.0.7.
>0)
>
>
> L463 unfinishedLineString is assigned a newly created dateLineCorrected
> created l428, but no other code uses it because assignment is in if
> block L463->466 and potential use of unfinishedLineString happens in
> that same test's else block L467->487
>
> If code is working then keeping dateLineCorrected L463 is the leak, and
> L478->481 can never happen.
> Otherwise I cannot tell what happens but most likely either
> unfinishedLineString or dateLineCorrected must be deleted somehow.
>
> Best Regards,
>
> Thibaut
>
> _______________________________________________
> Marble-devel mailing list
> Marble-devel at kde.org
> https://mail.kde.org/mailman/listinfo/marble-devel