comment about connection.py in UI testing code

[Markus Mohrhard]

Hey Stephan,

to come back to your comment from IRC earlier.

https://cgit.freedesktop.org/libreoffice/core/commit/?id=74bf4ca9f382fa12481fda18b57cdce4c0d5422b
should catch the case now that the started LibreOffice instance has already
stopped before we can connect.

Additionally I found some more information about the dbgutil assert that I
mentioned.

I managed to add some debug code and improve the order that the output for
the test is generated and found the following pattern.

tearDown: calling terminate()...
create desktop    <-- added from me as debug output
terminate            <-- added from me as debug output
caught DisposedException while TearDown
ok
test_select_text (start.SimpleWriterTest) ...
warn:sal.osl.pipe:23905:6:sal/osl/unx/pipe.cxx:497: recv() failed:
Connection reset by peer
warn:binaryurp:23583:9:binaryurp/source/reader.cxx:123: caught UNO
exception 'Binary URP bridge already disposed'
warn:binaryurp:23905:6:binaryurp/source/bridge.cxx:844: undisposed bridge,
potential deadlock ahead
python3: /home/moggi/devel/libo/binaryurp/source/binaryany.cxx:107: void*
binaryurp::BinaryAny::getValue(const com::sun::star::uno::TypeDescription&)
const: Assertion `type.get()->eTypeClass == typelib_TypeClass_ANY ||
type.equals(css::uno::TypeDescription(data_.pType))' failed.


So it seems I basically hit this if I the Desktop.terminate call has thrown
a DisposedException that we get the problem during the next start.
Additionally it seems that we might leak Bridge instances but I need to
debug that a bit more. Adding a SAL_DEBUG into the Bridge constructor and
destructor shows two created bridges but only one destroyed one per test.

Regards,
Markus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice/attachments/20161011/4d2ceb43/attachment.html>