[Aptitude-devel] Problem building documentation.

[Jens Seidel]

On Sun, Sep 06, 2009 at 08:14:07AM +0200, Omar Campagne wrote:
> I'm translating the manual into Spanish, and when I tried to build the  
> documentation yesterday, I started to get all kind of mistakes I haven't  
> seen in other occasions.

Omar finally finished translating the manual. Thanks a lot!

> I've been "playing" and learning about po4a, but I know nothing about  
> docbook. However, I find strange double entries in the msg headers in  
> all the PO files. Example:
>
> Original document:
> #. type: Content of the dselect entity
> #: en/aptitude.xml:7
> msgid "<command>dselect</command>"
> msgstr ""
>
> Translated document:
> # type: Content of the dselect entity
> #. type: Content of the dselect entity
> #: en/aptitude.xml:7
> msgid "<command>dselect</command>"
> msgstr "<command>dselect</command>"
>
> msgmerge -U po pot doesn't eliminate the extra header entries.

It cannot. According to
http://www.gnu.org/software/hello/manual/gettext/PO-Files.html
"#." starts comments extracted from the source code, these are OK.
The faulty one is "# type: Content of the dselect entity" and it is
considered as a comments made by you. Gettext tools do not and should not
remove comments added by the translator.

Don't know where the "# type:" comments comes from, maybe a buggy po4a
or gettext version? Very likely is also that an older version of gettext
was used which didn't support "#." comments well.

> As an experiment, I tried invoking ":g/.*#\ type.*/d" in Vim, in order  
> to delete all lines starting with "# type", with no avail, though (it  
> seems to me) that I started getting less mistakes at build time (or not,  
> not really sure).

Such comments ("#" followed by space " ") should never result in build errors,
really never.

I suggest you remove these comments again and send a patch for inclusion.
Ensure not to remove important comments added by yourself.

> By coincidence, that very day I was trying to build the doc, getting  
> weird errors (and going nuts :) till Daniel sent that email. Can't  
> really remember, but I think some mistakes were similar to this day.

I noticed no problems except a few errors in text2html. Maybe it's time to
switch to lynx -dump (and UTF-8 instead of latin1)?

Jens