[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-i18n-doc
Subject:    Re: [RFC] Documentation, take 2
From:       Burkhard =?iso-8859-15?q?L=FCck?= <lueck () hube-lueck ! de>
Date:       2005-12-08 21:25:37
Message-ID: 200512082225.37438.lueck () hube-lueck ! de
[Download RAW message or body]

Am Mittwoch 07 Dezember 2005 20:44 schrieb Nicolas Goutte:
>
> No, I meant to know when it has failed. For checking that it fails, you do
> not need a revision number. already today, the script can tell you that PO
> files does not match with the untranslated DocBook.
>
Sometimes yes, sometimes no :(

I have attached an svn diff for an updated kmouth docbook with an thank to 
Nicolas for revision numbers and an detailed explanation how to easy break a 
language package. 
update_xml does *not* recognize this nonmatching docbook. 

Same happens with an updated kturtle docbooks.

I tried to find, why and how this check fails, could be in updated docbooks 
with tabs at the beginning of the changed lines, but I am not shure.

Burkhard Lück


["svndiff-kmouth.patch" (text/x-diff)]

Index: l10n/documentation/kdeaccessibility/kmouth/index.docbook
===================================================================
--- l10n/documentation/kdeaccessibility/kmouth/index.docbook	(revision 483694)
+++ l10n/documentation/kdeaccessibility/kmouth/index.docbook	(working copy)
@@ -56,7 +56,32 @@
   <para>
     &kmouth; is a program that enables persons that cannot speak to let their
     computers speak. It includes a history of spoken sentences from which the
-    user can select sentences to be re-spoken.
+    user can select sentences to be re-spoken. 
+    Hi Nicolas, thanks for the docbook revision in the po files.
+    This is the way to break a language compilation with minimal warning, just because
+    update_xml sometimes does not detect nonmatching docbooks and because it is 
+    insufficient to do this test with meinproc only for the whole documentation 
+    (via index.docbook), you have to test every docbook of an application and to 
+    revert every docbook (But this is not done in the current script, so a 
+    revision test for all docbooks of an application needs to be implemented):
+    An application with index.docbook and break.docbook has already succesful generated
+    language.docbooks in the repository. break.docbook is updated and the pofiles 
+    are generated.
+    svn up break.docbook and break.po and translate.
+    break.docbook is again updated, but the changed strings are not yet merged to 
+    break.po. Do an "svn up l10n/documentation l10n/language" and you have an nonmatching
+    break.docbook.
+    Try to generate the language.docbooks for this whole kdemodul.
+    break.docbook doesn't match break.po, so po2xml writes language.docbooks with 
+    english text for the nonmatching msgid's. One of these englisch sentences in break.docbook
+    includes an entity ("ie") defined in en/user.entities, but not in this language/user.entities
+    This is true for most languages. E.g. only 3 of 40 languages with user.entities in 3.5 
+    have an entity "ie". meinproc fails because of the unknown "ie" in the language, 
+    index.docbook is reverted, but *not* the just generated language.break.docbook.
+    You commit a whole bunch of also generated files for this kdemodul, but do not notice 
+    that you also commit the "broken" file language.break.docook.
+    Someone does a checkout/update from the repository for the language, compile breaks! 
+    AFAIR this happened at least twice this year in the german language.
   </para>
   <para>
     Note that &kmouth; does not include speech synthesizer. Instead it requires


[prev in list] [next in list] [prev in thread] [next in thread]