https://bugs.kde.org/show_bug.cgi?id=373112

            Bug ID: 373112
           Summary: Scripty commits rc.cpp files to project repo if
                    Messages.sh fails
           Product: i18n
           Version: unspecified
          Platform: Other
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: NOR
         Component: general
          Assignee: aa...@kde.org
          Reporter: kosse...@kde.org
                CC: lu...@hube-lueck.de, luigi.tosc...@tiscali.it
  Target Milestone: ---

For the marble repo Scripty recently had committed the rc.cpp artifact,
2016-11-28 01:51:36 (GMT):
https://cgit.kde.org/marble.git/commit/?id=e172f35596d8709b19531be9109c96317e0d60de
And the next day Scripty then committed a duplication of the content of the
rc.cpp file, 2016-11-29 02:07:41 (GMT):
https://cgit.kde.org/marble.git/commit/?id=a91c12e41d0f88bcb8e4a51cade4699ccbf83ed5

That ended when commits were made to the Marble repo to fix all errors of
"[...] message is missing '//~ singular' comment" & "[...] message is missing
'//~ plural' comment" reported by generate-en-po and removing the rc.cpp file
again.

The Marble code which triggered this due to the missing singular/plural
comments was committed right before that first Scripty commit, 2016-11-27
09:07:39 (GMT):
https://cgit.kde.org/marble.git/commit/?id=473d6972a28b054318e986de771066817df38de2

A first analysis shows that this happened due to
* extract-messages.sh runs Marble's toplevel Messages.sh
* that Messages.sh runs extract-tr-strings
* extract-tr-strings runs generate-en-po
* generate-en-po sees missing plural/singular fields and exits with 1
* extract-tr-strings has "set -e" and also immediately exists, bash forwards
the exit code
* Messages.sh had extract-tr-strings as last command and also exits, forwarding
again the exit code of the last command
* extract-messages.sh tests the exit code and then emits a warning about the
error code instead of removing the helper file rc.cpp

Not sure if rc.cpp files being committed have never happened before, but I
would assume errors like missing plural notations might have. Perhaps the new
translation server has a newer bash which behaves differently WRT exit code
forwarding, so this is now happening? Just a thought.

In any case, would be good to prevent the rc.cpp file being ever committed to
the repo :)

Comment in the related discussion on #kde-i18n:
<tsdgeos> figure out which git command is actaully adding files s
<tsdgeos> since afaik we should only ever modify existing ones

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to