Re: building vanilla gnu.org-i18n

[Yavor Doganov]

At Sat, 08 Dec 2007 09:48:01 +0200,
Kaloian Doganov wrote:
> Yavor Doganov <address@hidden> writes:
> 
>     AFAICS, everything is working properly now, modulo the known bugs.
>     OK to install the new templates?  I would like to test the system in
>     real simulation, i.e. to copy only the master files.
> 
> Yes, let's go ahead.  When you copy the new templates, please commit
> them even if they fail to build with the rest of the system.  We would
> like to see what the errors are.

That's self explanatory.  I'll only remove the sidebar stuff, as it's
obsolete now.  

>     I plan to start working on the CVS commands -- it may be a bit
>     premature but they're pretty important.
> 
> Could you please explain what is the desired behaviour of those SCM
> (being CVS or SVN) commands?  I remember that we have discussed this
> earlier, but I have no written record for it. :-/

Newly generated files must be cvs add'ed automatically, otherwise the
system would be much less usefull.

As I see it, there will be a script invoked by a cronjob that would
roughly do:

1) cvs update of the whole repository
2) invoke make
3) cvs commit on success

I dropped the idea of adding directories automatically -- this will be
detrimental to the repository even due to a trivial bug.  The /po
subdirs are not so many, so I'll add them manually when the time
comes.

More specifically, the system should cvs add these files if they are
created by the make process:

/prep/i18n/generic.xx.html (empty file)
/path/article.pot (when web-translators managers add a new article)
/path/article.xx.html (when a translator commits article.xx.po).

>     I was thinking also about a way to generate meaningful CVS log
>     entries, describing the changes.  Would that be useful?
> 
> It would be useful, I suppose.  I don't know who will read them when the
> system is deployed and actually used, so I can not estimate the
> importance of this feature.  But it is nice to have concsice log
> messages instead of dummy ones.

It would be useful, say, to quickly check which article.pot or
article.xx.html corresponds to which version of article.html or
article.xx.po.  I'm not sure.

>     I think yes, but I might be over-estimating this.  I see that it
>     won't be extremely hard to do it via the makefile.  Or it would be
>     easier to use Scheme for this?  I don't know, please comment.
> 
> Honestly, currently I don't have and idea how implement this.  If you
> can suggest a strategy for retrieving and composing the contents of the
> log message, then we can think whether Make/Bash or Scheme fits the
> strategy.

I was thinking about this:  When a file is modified, an extra command
will append the cvs log message from the original file.  Upon
successful make && commit, this log file will be deleted.  In case of
failure, it will not be deleted so that the accumulated messages will
be used in the next run.

>     Also, it would be nice to have a `report' target that would be
>     language specific and will output the state of all files for a
>     language team, and extended `full-report' that will check the
>     activity of all teams for a certain period.
> 
> If you can define what these reports will contain, we can think about
> how to generate them automatically.

For language teams, a `msgfmt --statistics'-similar output would be
sufficient, as a start (probably omiting files that are 100%
complete).  We'll see, it's a bit early to think about this.

>     Another idea is to implement a `check' target that would test the
>     environment and all the tools we use.
> 
> This is a nice feature too.  Are there any relevant GNU standards or
> good practices about how to implement such things?

Autoconf or DejaGNU.