Re: [pango] 1.36.4

[Behdad Esfahbod <behdad behdad org>]

Thanks Emmanuele.

I certainly didn't intend to break Pango.  In the future, send me an email if
I do, because that's not intended.  In this case, I was under the impression
that the API that I touched has no users left whatsoever.  I was wrong,
because I have not deprecated and removed pango-ot.h yet.

At any rate, added back the old API in HarfBuzz.  Since it didn't go into any
HarfBuzz release, we're safe there.  All needed is a new pango release.

Cheers, and thanks again,
behdad

On 14-06-24 04:04 AM, Emmanuele Bassi wrote:
> hi Behdad;
>
> On 23 June 2014 22:59, Behdad Esfahbod <behdad behdad org> wrote:
> > Ugh.  Wish I had in mind that a release is coming.  My bad.
> >
> > I changed API in harfbuzz master, which broken Pango build.  ebassi updated
> > Pango and required harfbuzz 0.9.25, which does NOT have the API change.
>
> I was trying to put the GNOME autobuilder back to working, since it
> had been failing for a whole night; given that Pango sits fairly low
> on the stack, it not building was preventing other modules from
> building and the tests from running.
>
> I bumped the HB requirement in Pango to 0.9.29, which was the version
> in Git that held the commit that changed the API, not to 0.9.25.
>
> if the harfbuzz API can change without the version not getting bumped
> in Git, then it's a problem for every other project that depends on
> harfbuzz.
>
> > Alternatively, and perhaps this is the better option: I can add the old API
> > back in...   The code involved is in pango-ot, that is fairly unused.  I can
> > remove the hb function call there completely so newer Pango will keep building
> > with a variety of harfbuzz versions.  There won't be any lost functionality in
> > practice.
> >
> > WDYT?
>
> adding back the old symbol, maybe as a wrapper that passes NULL with a
> deprecation warning, is perfectly fine by me.
>
> ciao,
>  Emmanuele.

-- 
behdad
http://behdad.org/