Re: [bug-gettext] a .gmo file is not regenerated when its .po file changed

[Benno Schulenberg]

On Fri, May 29, 2015, at 23:03, Daiki Ueno wrote:
> Benno Schulenberg <address@hidden> writes:
> > Ah, but when that rule notices that none of the string content
> > has changed, and does not _replace/overwrite the file, it should
> > nevertheless _touch the file.
> 
> That may work with git, which doesn't record file timestamps, but might
> not work with other VCS.

It works with Subversion and with Mercurial.  And I think it will work
with any VCS, because I can't imagine that there is a VCS that will mark
a file as modified and wants to check it in just because the timestamp
changed.  Can you give an example?

When checking out files, some VCSs may restore the timestamp of the
time of check-in, but that is their problem and I don't think that
gettext needs to cater for the build problems that that creates.

>   # [...] This timestamp would not be necessary
>   # if updating the $(CATALOGS) would always touch them; however, the rule for
>   # $(POFILES) has been designed to not touch files that don't need to be
>   # changed.

It has been designed so, yes.  But it was designed wrong, in my opinion;
the content of the file does not need to be changed, and should therefore
not be changed, but the timestamp should be touched/updated so that make
(and humans) can see that things are up to date.

(By the way, what are $(CATALOGS) exactly?  How are they different from
$(POFILES) or $(GMOFILES)?)

Regards,

Benno

-- 
http://www.fastmail.com - Does exactly what it says on the tin