Re: wip/qtcreator package

Hi Matthias

Thanks for your reply.

Matthias Drochner wrote:
> [email protected] said:
>> The assumption is that pkgsrc  always sets the qmake mkspec to
>> "netbsd-g++", regardless of the host OS.
> 
> I might have missed something, but I haven't seen that.

 From x11/qt4-tools/buildlink3.mk:

...
# the way the spec files are currently instealled via pkgsrc, they
# are all identical anyway, so just pick one and point to it.
# Without specifying QMAKESPEC, qmake can't figure it out on some
# platforms (solaris for example)
CONFIGURE_ENV+=         QMAKESPEC=${QTDIR}/mkspecs/netbsd-g++
MAKE_ENV+=              QMAKESPEC=${QTDIR}/mkspecs/netbsd-g++
...

Also qmake on DragonFly defaults to mkspec "freebsd-g++" but pkgsrc sets 
it to "netbsd-g++" when i build wip/qtcreator here.

> Unless told otherwise, qmake uses the "default" mkspec
> which is set up at qmake build time to match the
> current configuration. See the "pre-configure" target in
> Makefile.common.

Actually i don't see where Makefile.common is pulled in through the 
buildlink3.mk chain - i thought Makefile.common is used for qt4 only, 
not for depending packages?

> 
>> The dl library is part of the libc on BSD, but has to be specified on
>> other systems like linux. QMake and pkgsrc should set "-ldl" directly
>> for those systems.
> 
> On systems where it is needed, -ldl gets put into the SYS_LIBS
> qmake variable, so this should be pulled in.
> (I don't see Linux there, perhaps there is some other magics,
> or it needs to be added.)
> 

Ok, so this means that we could also abandon this portability hack in 
the qtcreator build system completely and make patch-ae patch it away. 
Looks like we are fine with that anyway, it's more about what solution 
to advertise upstreams. Seems to me that the portability hack they put 
in was just a fix for some bad qmake configurations in the first place.



>> Is the '${FIND} -exec' used for the install routine in
>> wip/qtcreator/Makefile portable?
> 
> It doesn't look bad, but I think some "pax -rw" is simpler.
> 

The advantage of using "${FIND} -exec ${INSTALL_DATA}" is that owner and 
mode are set accordingly, thus i opt to stick with this method unless 
there is reason against it.



>> Optional Debugging Helper
> 
> As sad as it is, but NetBSD is out of luck here anyway because
> debugging of multithreaded programs stopped working after
> some pthread changes, before the 5.0 release.

AFAIK DragonFly has a similar problem, but i don't think that's a reason 
not to make the Debugging Helper library build. In fact it's just a 
wrapper for gdb output to display Qt4 objects nicely.


If these issues are ok i will try to fix the Debugging Helper library 
and ask for a review on the mailing list later this week.

Best regards

Florian Walpen

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with 
Crystal Reports now.  http://p.sf.net/sfu/bobj-july