problem

The changes.xliff diff algorithm (which is just the SdlXliff diff algorithm) fails under some circumstances.

steps to reproducce

1. Import the testdata (DE-RU) stored in TINTERNAL-27 with enabled termtagger.
2. Disable termtagger
3. Change the target of Segment Nr 17 to the content as specified in TINTERNAL-27.
4. Then finish the task.

This fails with error:

[02-May-2018 17:22:05 Europe/Vienna] ZfExtended_Log on translate5.localdev: Versand der Minidump-Fehlermails deaktiviert - Subject:
exception 'Zend_Exception' with message 'there are more closing tags then opening tags for the tags with the id __mrkTag__6d747970
653d227465726d22207472616e736c617465353a7374617475733d227072656665727265645465726d22' in /home/tlauria/www/translate5-master/appl
ication/modules/editor/Models/Export/DiffTagger/Sdlxliff.php:383

Complete stacktrace see also in TINTERNAL-27.

If the segment is edited with enabled termtagger everything works well.

possible reason

The combination of text changes and terminology leads to a MRK tag structure which seems to be invalid. One problem is possibly the non unique ID generation as visible in the above Exception message. The ID of the affected tag evaluates to its attributes: 'mtype="term" translate5:status="preferredTerm"' which is indeed not unique. So perhaps the problem is solved by ensuring that the MRKs are getting real unique IDs on internal processing.

testdata

I tried to make a simple test case, but I did not succeed. The reason is in the combination of sdl change marks depending on the changed content and term mrk tags.