[QGIS-Developer] Crashes in qgsogrfeatureiterator.cpp:52 on master

[Even Rouault]

On vendredi 10 novembre 2017 08:41:53 CET Nyall Dawson wrote:
> On 10 November 2017 at 08:24, Even Rouault <even.rouault at spatialys.com> wrote:
> > Digging further this code comes from the dev branch of qt5:
> > https://github.com/qt/qtbase/blob/dev/src/corelib/thread/qsemaphore.cpp#L4
> > 14
> > 
> > Released branches use a different logic:
> > https://github.com/qt/qtbase/blob/5.10/src/corelib/thread/qsemaphore.cpp#L
> > 217
> > 
> > Are you running bleeding edge Qt ?
> 
> No - this was just the osgeo4w dev release, so using 5.9.1.
> 
> Still, maybe it's safest if we just don't use tryAcquire for negative
> timeouts, and instead use acquire explicitly in this case...

Yes, that seems safer

And actually what you observe is logical, since 5.9 is also buggy for a dfferent reason since
QSemaphore::tryAcquire(int n, int timeout)
https://github.com/qt/qtbase/blob/5.9/src/corelib/thread/qsemaphore.cpp#L217
calls when timeout < 0
QSemaphore::tryAcquire(int n) which returns false when no resource is free
https://github.com/qt/qtbase/blob/5.9/src/corelib/thread/qsemaphore.cpp#L188

But 5.8 is ok:
https://github.com/qt/qtbase/blob/5.8/src/corelib/thread/qsemaphore.cpp#L217

The regression went per:
https://github.com/qt/qtbase/commit/eaee1209f0ead5be786e81db8aee604ccfea85b0#diff-30e4174ed07523635a22648c06274e0e

and this should occur also on any platform running qt 5.9

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20171109/4ae7d323/attachment.html>