[OE-core] [PATCH] package.bbclass: allow shell-style wildcards in PRIVATE_LIBS

[Quentin Schulz]

Hi Alex,

On Wed, Sep 04, 2019 at 12:11:29PM +0200, Alexander Kanavin wrote:
> On Wed, 4 Sep 2019 at 09:32, Quentin Schulz <
> quentin.schulz at streamunlimited.com> wrote:
> 
> > > +++ b/meta/classes/package.bbclass
> > > @@ -1646,7 +1646,8 @@ python package_do_shlibs() {
> > >                  prov = (this_soname, ldir, pkgver)
> > >                  if not prov in sonames:
> > >                      # if library is private (only used by package) then
> > do not build shlib for it
> > > -                    if not private_libs or this_soname not in
> > private_libs:
> > > +                    import fnmatch
> > > +                    if not private_libs or len([i for i in private_libs
> > if fnmatch.fnmatch(this_soname, i)]) == 0:
> >
> > We only need to know if this_soname is matching one of the patterns, we
> > don't need to check for each and every pattern.
> >
> > Something like:
> >
> > for pattern in private_libs:
> >         if fnmatch.fnmatch(this_soname, pattern)]) == 0:
> >                 sonames.add(prov)
> >                 break
> >
> > would be possible?
> >
> 
> Actually, this suggested code snippet looks altogether wrong to me. We do

Indeed.

> need to check the file against every pattern, and only if it matches none
> of them, it gets added to sonames. Note the original code says 'this_soname

So we just need to invert the logic actually.

matched = False
for pattern in private_libs:
	if fnmatch.fnmatch(this_soname, pattern)]) == 0:
		matched = True
		break

if not matched:
	sonames.add(prov)

> not in private_libs', which also means going over every item in the list.
> 

I'd hope Python devs did optimization on this one and stopped at the
first match in the list, but I'm not one of them so...

BR,
Quentin
-- 
StreamUnlimited Engineering GmbH
High Tech Campus Vienna, Gutheil-Schoder-Gasse 10, 1100 Vienna, Austria
quentin.schulz at streamunlimited.com, www.streamunlimited.com