GTK+ broken: anyone knows why?

[Shreevatsa R]

On Tue, May 6, 2008 at 2:31 AM, Ryan Schmidt <ryandesign at macports.org> wrote:
> On May 6, 2008, at 00:24, Shreevatsa R wrote:
>
>> On Mon, May 5, 2008 at 11:36 PM, Ryan Schmidt wrote:
>>>
>>> On May 5, 2008, at 09:24, Shreevatsa R wrote:
>>>
>>>
>>>> The GTK+ port (gtk2) is broken in multiple ways:
>>>>
>>>> 1. Did anything change in the base recently? It used to be that
>>>> variants chosen are passed on to dependencies as well, but it appears
>>>> that now `port install gtk2 +quartz` and `port install gtk2 +quartz
>>>> +no_x11` do different things, although the port gtk2 has "variant
>>>> quartz requires no_x11 ..." in it. That is, `port install gtk2
>>>> +quartz` installs the X11 version of pango, etc., instead of pango
>>>> +no_x11. Same with cairo.
>>>>
>>>
>>>  I don't know if something changed in base with regard to this. If this
>>> worked before, can you do some regression testing to see what revision of
>>> base broke it?
>>
>> Oh, this is strange: I tested with 1.6.0, and it doesn't seem to work
>> with that either: "port install gtk2 +quartz" doesn't seem to pull in
>> +no_x11 for dependencies. I (and others) were quite sure this used to
>> work just a month ago or so, but if it doesn't work with 1.6.0... what
>> can it mean? 1.6.0 hasn't been changed in a long time, right? (Or are
>> the files at
>> http://svn.macports.org/repository/macports/downloads/MacPorts-1.6.0/
>> generated from SVN, thus "MacPorts 1.6.0" doesn't refer to any fixed
>> thing? If this is the case, I will do more detailed regression testing
>> to pin down the date, and thus the revision which changed things.)
>
> 1.6.0 today is the same as 1.6.0 when it was created in r32094 on
> 2007-12-16. The disk images and the installer packages on them are created
> manually, not by an automated process. Well, they're created using the port
> called "macports" but this is kicked off by a person, and the result is
> tested, and then the disk images are manually committed into the repository.
> There is a document somewhere on our website describing what all gets done
> when a release is to be made.
>
> If MacPorts base didn't change, maybe the portfile did. Check the log of the
> portfile in the repository.

I guess it was never working, and my memory's just playing tricks on
me. I guess I was always doing "+no_x11" manually after all. Even the
FAQ says "Is it possible to have a dependency on a specific variant of
another port? [...] No."
So I guess this part never worked at all. And no one pointed this out
when I was submitting useless patches at
http://trac.macports.org/ticket/14305 :P

>>>> 2. After installing (`port install gtk2 +quartz +no_x11`) and
>>>> installing any GTK+ application on top of it, all windows display only
>>>> boxes; fonts don't work. This might be related to the recent changes
>>>> in cairo, or pango, or something.
>>>>
>>>
>>>  I believe I've had some other reports of this but I don't know how to
>>> fix
>>> it. I started a dialog with the developers of pango but I seem to have
>>> forgotten to follow up. I'm still not sure if this is a pango problem or
>>> a
>>> cairo problem.
>>>
>>
>> Do you have references to that dialogue, to maybe know the current
>> state of things?
>> This is nothing terribly urgent of course :)
>
> This is the thread I was thinking of:
>
> http://mail.gnome.org/archives/gtk-i18n-list/2008-April/msg00004.html
>

Thanks. I am trying to install from older Portfiles in an attempt to
track down the bug, but it appears that the previous pango portfile
had a "patch-ltmain.sh.diff" which it is unable to fetch now. Maybe
the contents of that patchfile were important?