Re: [Qemu-devel] PO makefile rules touching source

[Fam Zheng <famz@redhat.com>]

On Fri, 08/07 10:09, Dr. David Alan Gilbert wrote:
> * Stefan Weil (sw@weilnetz.de) wrote:
> > Am 06.08.2015 um 13:54 schrieb Dr. David Alan Gilbert:
> > >Hi Stefan,
> > >   I don't really understand the way PO stuff is supposed to work,
> > >but I often get the problem that my git tree gets changed during
> > >an out of tree build, causing the po subdirectory to be
> > >dirtied.
> > >
> > >I think this is due to the rule:
> > >
> > >$(PO_PATH)/%.po: $(PO_PATH)/messages.po
> > >         $(call quiet-command, msgmerge -q $@ $< > $@.bak && mv $@.bak $@, "  GEN   $@")
> > >
> > >in po/Makefile.
> > >
> > >Why does that merge into $(PO_PATH)/%.po rather than %.po - i.e.
> > >why to the source rather than to the build directory?
> > >
> > >Dave
> > >
> > >--
> > >Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
> > 
> > Hi Dave,
> > 
> > the *.po files are initially generated, then manually
> > enhanced by the translated texts.
> > 
> > Each time when there are modifications to the
> > related source file (for QEMU: ui/gtk.c), the *.po
> > files need updates, too (new or removed texts,
> > but most often changes of line numbers).
> > 
> > That's why I have sent a patch to update the
> > *.po files for QEMU 2.4.0. With this patch
> > (which is still not applied), you would not
> > get changes when doing a complete build
> > because *.po files and ui/gtk.c would match.
> > 
> > As long as there is the mismatch, you'll
> > get modified *.po files when running
> > "make install". I usually handle this case
> > by running "git diff po|patch -p1 -R"
> > which removes the *.po modifications.
> 
> But why are these files in the source tree *ever* updated by a make install;
> if you're saying the .po files need updating, that's fine, but have the
> makefile do the update into the build directory, not the source directory.
> You might want a convenience make target for updating the source directory
> po files to make your life easier, but I don't understand why that triggers
> during a normal make install.

I agree. IMO Makefile should generally write anything to the build dir and only
read the SRC_PATH.

Fam