configure spewage on clean svn tree due to missing POTFILES

[Chris Shoemaker]

On Mon, Jan 09, 2006 at 08:51:53AM +0000, Neil Williams wrote:
> On Sunday 08 January 2006 10:02 pm, Derek Atkins wrote:
> > I just tried to build from a current SVN tree, a clean checkout
> > (effectively), and I got the following spewage from configure:
> >
> > checking what extra warning flags to pass to the C compiler...  -Werror
> > -Wdeclaration-after-statement mv: cannot stat `po/POTFILES': No such file
> > or directory
> > ./configure: line 37704: po/POTFILES.tmp: No such file or directory
> 
> I've noticed this on each build but, personally, I don't find it a problem. It 
> only happens on the CVS build. It never interferes or halts the build, it's 
> more a nuisance value.
> 
> It's brought into configure through the ALL_LINGUAS macros but when checking 
> the actual configure script, it's only a cursory tidy-up within the very last 
> stages of the macro. The command that generates the warning is intended to 
> simply refresh po/POTFILES which is something we do later in the build.
> 
> In configure:
>    if test "$gt_cv_have_gettext" = "yes"; then
>      if test "x$ALL_LINGUAS" = "x"; then
>        LINGUAS=
> <snip>to the very last command</snip>
>    rm -f po/POTFILES
>    sed -e "/^#/d" -e "/^\$/d" -e "s,.*, $posrcprefix& \\\\," -e "\$s/\(.*\) 
> \\\\/\1/" \
>         < $srcdir/po/POTFILES.in > po/POTFILES
> 
> The only way to silence the message is to output a po/POTFILES.in before 
> gettext starts work - in effect resurrecting the po/POTFILES.in stump that we 
> don't want in svn.
> 
> Is it really worth it?
> 
> I suppose a macro that touches po/POTFILES.in would shut it up.
> 
> > Um, is there some way you guys hacking on the po/POTFILES stuff
> > can handle this, please?
> 
> IMHO, we don't need to. It's a noisy nuisance warning that has no deleterious 
> effects on the rest of the build or generation of the translations. We just 
> run the refresh at a slightly later stage.
> 
> I've incorporated the po/POTFILES.in generation into the latest releases of 
> QOF and piot-qof. Trouble free because po/POTFILES.in - once built - is 
> distributed in the make dist tarball so the warning never appears in the 
> release code.
> 
> It's only CVS/svn developers who will ever see this "warning" and, IMHO, a 
> simple comment in README.svn may be the best solution. i.e. just ignore it, 
> it does no harm. There are plenty of other areas and quirks that are CVS/svn 
> only - I say ignore it.

Oh, thanks for explaining this, Neil.  I totally misunderstood Derek's
initial report.  I thought that this was halting configure.  I now see
what Derek was talking about.

IMO, if we're interested in quieting configure-noise, there are bigger
fish to fry, like noise caused by _our_ sloppy code, instead of noise
coming from some stock macro.

-chris