Details
-
Bug
-
Resolution: Unresolved
-
None
-
None
-
Critical
-
Added "t5 patch:okapi146whitespace" CLI command to detect and add back missing whitespace characters caused by Okapi 1.46 segmenting issue
-
Empty show more show less
Description
problems
Okapi 146 had a "bug" which was segmenting in a way, that whitespace characters between Sentences were put in separte MRK tags.
After import such a task into translate5, no segment was created for that segment with that whitespace only, and therefore the mrk was not added to the target tag on export.
solution
Prio 1: Implement a CLI script, walking over all skeleton files, use the internal XML parser to compare the mrks between <source-seg> and <target> and if there are missing mrks with space only as content in target, then add them.
See therfore also: /var/www/hosting-scripts/helpers/find-in-skeletonfile.php
The CLI should by default scan and summarize how many of the tasks / files are affected and out of config which okapi version was used.
An additional option --fix must be used in order to manipulate and fix the skeleton files. The overwrite of the skeleton files must be atomic to prevent conflicts with exports currently running.
Prio 2: Validate how that could happen in translate5 and trying to fix that.
Further Steps
After rolling that out we have to analyse some installations and define if we create and additonal migration script invoking the fix script on all instances.