[Libreoffice] subsequenttest hang ...

Bjoern Michaelsen bjoern.michaelsen at canonical.com
Tue Sep 27 02:18:55 PDT 2011


On Tue, 27 Sep 2011 01:53:35 -0500
Norbert Thiebaud <nthiebaud at gmail.com>
wrote:

> here the tail of the make check log.
> LOG> Log started 26.08.2011 - 11:34:33
> checking: [forms.OEditModel::com::sun::star::beans::XMultiPropertySet]
> is iface: [com.sun.star.beans.XMultiPropertySet] testcode:
> [ifc.beans._XMultiPropertySet]
> LOG> Execute: getPropertySetInfo()
> Method getPropertySetInfo() finished with state OK
> LOG> getPropertySetInfo(): PASSED.OK
> 
> LOG> Execute: setPropertyValues()
> LOG> starting required method: getPropertyValues()
> LOG> Execute: getPropertyValues()
> LOG> starting required method: getPropertySetInfo()
> Method getPropertyValues() finished with state OK
> LOG> getPropertyValues(): PASSED.OK
> 
> LOG> Changing all properties
> Method setPropertyValues() finished with state OK
> LOG> setPropertyValues(): PASSED.OK
> 
> LOG> Execute: addPropertiesChangeListener()
> LOG> starting required method: getPropertyValues()
> make[1]: ***
> [/fast/core/workdir/unxlngx6.pro/JunitTest/forms_unoapi/done] Error 1

That look suspiciously familiar. The subsequenttests in forms are still
quite a bit unstable. It tried to find the rootcause, but found that
the issue is triggered only by running a certain sequence of test one
after another, so it is not easy to isolate. But all these
problematic calls had to do with ProperySet accessors -- so likely
some PropertySet base class in forms is rotten still -- even though I
though this was fixed with:

  https://bugs.freedesktop.org/show_bug.cgi?id=35728

but obviously it isnt.

> >> so clearly running make check is not ready for prime time. I'm
> >> turning that off on my box for now.

IMHO it would still be worth a try to keep running the
subsequenttests after disabling either forms:OEditModel or more
pragmatically all unoapi tests in forms. The rest should be good.
Maybe worth a discussion on the TSC call.

That being said: That still is a serious issue regardless -- LO should
not crash in a propertyset accessor.

Best,

Bjoern

-- 
https://launchpad.net/~bjoern-michaelsen




More information about the LibreOffice mailing list