[OE-core] [CONSOLIDATED PULL 10/20] package.bbclass: add support to split Qt translation files

[Richard Purdie]

On Mon, 2011-06-06 at 17:27 +0000, Otavio Salvador wrote:
> On Mon, Jun 6, 2011 at 17:14, Richard Purdie
> <richard.purdie at linuxfoundation.org> wrote:
> > On Sun, 2011-06-05 at 23:44 -0700, Saul Wold wrote:
> >> From: Otavio Salvador <otavio at ossystems.com.br>
> >>
> >> There're many Qt applications that provide translation files in '.qm'
> >> format however those weren't being splitted as GetText based
> >> ones.
> >>
> >> Signed-off-by: Otavio Salvador <otavio at ossystems.com.br>
> >
> > Am I right in assuming all these qt applications use one of the qt
> > classes?
> 
> I'd expect it.
> 
> > I do wonder if this shouldn't be part of one of the qt classes as part
> > of a package_do_split_locales_append() or prepend() style function from
> > a code separation point of view.
> 
> I believe there's going to have a lot of code duplication between the
> two to do that.

Not really. Something like:

python package_do_split_locales_qm() {
   import re

   dvar = bb.data.getVar('PKGD', d, True)
 
   # Check of Qt translation files
   qm_re = re.compile("(.*)\.qm$")
   qm_files = {}
   for root, dirs, files in os.walk(dvar):
       for file in files:
           qm_file = qm_re.match(file)
           if qm_file:
               locale = qm_file.group(1)
               relpath = os.path.join(root, file).replace(dvar, '', 1)
               if relpath:
                   if not qm_files.has_key(locale):
                       qm_files[locale] = []
                   qm_files[locale].append(relpath)
 
    locale_re = re.compile("^.*([a-z]{2}(_[A-Z]{2})?)$")
    for l in qm_files:
        ln = legitimize_package_name(locale_re.match(l).group(1))
        pkg = pn + '-locale-' + ln
        bb.data.setVar('FILES_' + pkg, " ".join(qm_files[l]), d)
}

PACKAGE_PREPROCESS_FUNCS += "package_do_split_locales_qm"

which doesn't really duplicate anything with package.bbclass. To make
that work we'd need to set PACKAGELOCALES and change package.bbclass to
something like:

diff --git a/meta/classes/package.bbclass b/meta/classes/package.bbclass
index 1e6a872..66796e7 100644
--- a/meta/classes/package.bbclass
+++ b/meta/classes/package.bbclass
@@ -358,12 +358,18 @@ python package_do_split_locales() {
 
 	localedir = os.path.join(dvar + datadir, 'locale')
 
-	if not os.path.isdir(localedir):
+	locales = set()
+	if os.path.isdir(localedir):
+		locales |= os.listdir(localedir)
+
+	extralocales = d.getVar("PACKAGELOCALES", True)
+	if extralocales:
+		locales |= extralocales
+
+	if not locales:
 		bb.debug(1, "No locale files in this package")
 		return
 
-	locales = os.listdir(localedir)
-
 	# This is *really* broken
 	mainpkg = packages[0]
 	# At least try and patch it up I guess...
@@ -378,7 +384,11 @@ python package_do_split_locales() {
 		ln = legitimize_package_name(l)
 		pkg = pn + '-locale-' + ln
 		packages.append(pkg)
-		bb.data.setVar('FILES_' + pkg, os.path.join(datadir, 'locale', l), d)
+		files = os.path.join(datadir, 'locale', l)
+		extrafiles = d.getVar('FILES_' + pkg, True)
+		if extrafiles:
+			files = extrafiles + " " + files
+		bb.data.setVar('FILES_' + pkg, files, d)
 		bb.data.setVar('RDEPENDS_' + pkg, '%s virtual-locale-%s' % (mainpkg, ln), d)
 		bb.data.setVar('RPROVIDES_' + pkg, '%s-locale %s-translation' % (pn, ln), d)
 		bb.data.setVar('SUMMARY_' + pkg, '%s - %s translations' % (summary, l), d)

and there are probably neater ways to do this.

> > I do appreciate the need to influence
> > the list of locales although you could do that through directory
> > creation or an extra variable but we could enhance package.bbclass to
> > support that.
> 
> I don't understand what you meant by this. Mind to elaborate it a bit more?

See above for the kind of thing I'm thinking of.

> ...
> >> +             if qm_files.has_key(l):
> >> +                     locale_re = re.compile("^.*([a-z]{2}(_[A-Z]{2})?)$")
> >> +                     ln = legitimize_package_name(locale_re.match(l).group(1))
> >> +                     files += qm_files[l]
> >
> > Regardless, this is changing ln under some weird circumstances.
> > Shouldn't the code adding this to the locales list be handling this
> > translation? I can see potential duplication between the arrays and a
> > host of other nasty bugs with this code as it stands :/. Please take
> > that regexp out of this loop at the very least.
> 
> Where do you suggest me to move the regexp to?

Well, I was thinking about the loop above this one just over qm_files
instead of all locales. Also note that re.compile can be reused and can
also live outside any looping.

Cheers,

Richard