Re: sc: test failure depending on order of execution

2023-03-08 Thread Xisco Fauli

Hello Miklos,

On 8/3/23 9:09, Miklos Vajna wrote:

Hi Xisco,

On Fri, Mar 03, 2023 at 12:43:38PM +0100, Xisco Fauli 
 wrote:

Unfortunately I don't feel brave enough to change this. Is there anyone
familiar with this part of the code willing to give it a try? otherwise, I
can at least report it in Bugzilla so it doesn't get lost

Please see if https://gerrit.libreoffice.org/c/core/+/148457 and
https://gerrit.libreoffice.org/c/core/+/148460 is enough to get rid of
that workaround in the test code.


Yes, it's enough. Thanks a lot for helping with this.

I'm removing the workaround in 
https://gerrit.libreoffice.org/c/core/+/148461


Regards

--
Xisco Faulí
LibreOffice QA Team
IRC: x1sc0



Re: sc: test failure depending on order of execution

2023-03-08 Thread Miklos Vajna
Hi Xisco,

On Fri, Mar 03, 2023 at 12:43:38PM +0100, Xisco Fauli 
 wrote:
> Unfortunately I don't feel brave enough to change this. Is there anyone
> familiar with this part of the code willing to give it a try? otherwise, I
> can at least report it in Bugzilla so it doesn't get lost

Please see if https://gerrit.libreoffice.org/c/core/+/148457 and
https://gerrit.libreoffice.org/c/core/+/148460 is enough to get rid of
that workaround in the test code.

Thanks,

Miklos


Re: sc: test failure depending on order of execution

2023-03-03 Thread Xisco Fauli

Hello Miklos,

On 3/3/23 8:56, Miklos Vajna wrote:

Hmm, it feels like this is an improvement, but the root cause is why we
have a static counter for charts in the first place. Do you know where
is that counter?


There are two, the one for sc is in 
sc/source/filter/excel/xeescher.cxx:1607, added with 
b9d1610aa212c4283a8af73c6050a0e60993c484 "Merged 
xlsx-shared-oox-chart-export-part1.diff from ooo-build" and the other is 
in oox/source/export/shapes.cxx:2523 added with 
9e02fc64430ccdb311a08dd966cce61884a8b34c "Merged 
xlsx-shared-oox-chart-export-part1.diff from ooo-build"


Unfortunately I don't feel brave enough to change this. Is there anyone 
familiar with this part of the code willing to give it a try? otherwise, 
I can at least report it in Bugzilla so it doesn't get lost


Regards

--
Xisco Faulí
LibreOffice QA Team
IRC: x1sc0



Re: sc: test failure depending on order of execution

2023-03-02 Thread Miklos Vajna
Hi Xisco,

On Thu, Mar 02, 2023 at 09:31:24AM +0100, Xisco Fauli 
 wrote:
> Yes, you were right. There is a static counter that increments the chart
> export file names so it failed to find 'xl/charts/chart1.xml' because it was
> exported as 'xl/charts/chart2.xml'
> 
> Markus already added a mechanism to workaround this to the tests in chart2/
> in 5f7861e6ce212f47420533aed302fa4b5510557d "fix chart export tests
> correctly".
> 
> https://gerrit.libreoffice.org/c/core/+/148065 should fix it in the general
> case.

Hmm, it feels like this is an improvement, but the root cause is why we
have a static counter for charts in the first place. Do you know where
is that counter?

At least resetting that counter similar to how it's done in oox/
DrawingML::ResetMlCounters() would be good, and then you don't need
tricks on the testing side.

Regards,

Miklos


Re: sc: test failure depending on order of execution

2023-03-02 Thread Xisco Fauli

Hi Miklos

On 27/2/23 8:18, Miklos Vajna wrote:

My first guess would be some static variable, but who knows.


Yes, you were right. There is a static counter that increments the chart 
export file names so it failed to find 'xl/charts/chart1.xml' because it 
was exported as 'xl/charts/chart2.xml'


Markus already added a mechanism to workaround this to the tests in 
chart2/ in 5f7861e6ce212f47420533aed302fa4b5510557d "fix chart export 
tests correctly".


https://gerrit.libreoffice.org/c/core/+/148065 should fix it in the 
general case.


Thanks

--
Xisco Faulí
LibreOffice QA Team
IRC: x1sc0



Re: sc: test failure depending on order of execution

2023-02-26 Thread Miklos Vajna
Hi Xisco,

On Fri, Feb 24, 2023 at 02:40:23PM +0100, Xisco Fauli 
 wrote:
> it seems there is a test in sc/qa/unit/subsequent_export_test2.cxx that
> fails depending on the order of execution.
> 
> If testTdf121260 is executed before testTdf107586, then it passes,
> otherwise, it fails with
> 
> ##Failure Location unknown## : Error
> Test name: ScExportTest2::testTdf121260
> An uncaught exception of type com.sun.star.container.NoSuchElementException
> - 
> /home/tdf/lode/jenkins/workspace/lo_gerrit/Config/linux_clang_dbgutil_64/package/source/zippackage/zipfileaccess.cxx:263:
> at 
> /home/tdf/lode/jenkins/workspace/lo_gerrit/Config/linux_clang_dbgutil_64/package/source/zippackage/zipfileaccess.cxx:263
> 
> which means the file 'xl/charts/chart1.xml' doesn't exist in the zip file
> and parseExport fails. See https://gerrit.libreoffice.org/c/core/+/147622
> 
> Anyone has any idea why the file is not created if testTdf107586 is executed
> first ?

Not really, but the way I would debug this is to see how
XmlFilterBase::openFragmentStream() at
oox/source/core/xmlfilterbase.cxx:495 gets called in the good case, and
then hopefully the backtrace would give you some idea how it's not
called in the bad case.

My first guess would be some static variable, but who knows.

Regards,

Miklos


sc: test failure depending on order of execution

2023-02-24 Thread Xisco Fauli

Hello,

it seems there is a test in sc/qa/unit/subsequent_export_test2.cxx that 
fails depending on the order of execution.


If testTdf121260 is executed before testTdf107586, then it passes, 
otherwise, it fails with


##Failure Location unknown## : Error
Test name: ScExportTest2::testTdf121260
An uncaught exception of type com.sun.star.container.NoSuchElementException
- 
/home/tdf/lode/jenkins/workspace/lo_gerrit/Config/linux_clang_dbgutil_64/package/source/zippackage/zipfileaccess.cxx:263: 
at 
/home/tdf/lode/jenkins/workspace/lo_gerrit/Config/linux_clang_dbgutil_64/package/source/zippackage/zipfileaccess.cxx:263


which means the file 'xl/charts/chart1.xml' doesn't exist in the zip 
file and parseExport fails. See 
https://gerrit.libreoffice.org/c/core/+/147622


Anyone has any idea why the file is not created if testTdf107586 is 
executed first ?


Regards

--
Xisco Faulí
LibreOffice QA Team
IRC: x1sc0