Python PortGroup resets depends_lib ?

[Michael Dickens]

I changed that line in reaction to py*-pyside[-tools] failing to build
on buildbots & then tracking the issue down to the Python 1.0 PortGroup
setting the depends, not appending to it. It's a simple fix that should
work well for many ports.

In case you were wondering, the depends_lib statements on lines 110 and
187 are for when {$subport eq $name || $subport eq ""} -> when this is
the py-foo version, not a numbered version. So, they have no impact on
actual ports; they just depend on the default pyXY-foo port & don't do
anything else. - MLD

On Thu, Oct 8, 2015, at 07:55 AM, Aljaž 'g5pw' Srebrnič wrote:
> On 8 ottobre 2015 at 13:14:31, René J.V. (rjvbertin at gmail.com) wrote:
>>
>>
>> On Thursday October 08 2015 11:46:49 Aljaž Srebrnič wrote:
>>
>>
>>> Hm, that’s interesting. Indeed on lines 110 and 187 of python-
>>> 1.0.tcl depends_lib is used directly, but on lines 100 and 191 depends_lib-
>>> append is used. In the qt4 PortGroup only depends_lib-append is used
>>> though. I wonder what happens if you call the python PortGroup
>>> first, and the qt4 one below it?
>>
>>
>> I haven't tried that in this case, but when I wrote my own port for
>> py-kde4 I learnt that you either use the Python portgroup as the last
>> included portgroup (and put up with its side-effects), or you don't
>> use it - depending on whether the project uses `setup.py` for
>> building. If you include the Qt4 portgroup after the Python portgroup
>> it's quite possible that you have to reset things that are more
>> complicated to reset than depends_lib.
>>
>> I haven't yet looked at python-1.0.tcl in detail; do you think
>> there's a reason to use depends_lib directly rather than appending to
>> it, on the lines you cite? There are no comments to explain the
>> difference (line 187 really should).
>>
>> BTW, line 100 was changed from `depends_lib` to `depends_lib-append`
>> yesterday, I guess I should check whether that change was in reaction
>> to this thread and if it addresses the issue I ran in to. The commit
>> message isn't very explicit about what's being appended to rather
>> than overwritten ...
>>
> I don’t know the reason depends_lib is user, but in my (limited)
> knowledge, -append should equally work. Michael, what do you think?


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.macosforge.org/pipermail/macports-dev/attachments/20151008/a14f3737/attachment.html>