Re: glib 1.1.12, gtk1.1.12 for debian/ppc
On Fri, Jan 15, 1999 at 12:24:58PM +0200, Konstantinos Margaritis wrote:
> On Fri, 15 Jan 1999, Sven LUTHER wrote:
> > i managed to download glib/gtk+ 1.1.12, and compiled it here on my sparc
> > station, but on my linux box at home, i tried to launch dbuild on it, and was
> > gratified with an internal compiler error, i had no time to go further into the
> > problem, but i upgraded all current gcc, binutils and other such stuff, and
> > launched it again, the problem persisted. I would not call an internal gccc
> > error a trivial matter. (but like said , i had no time to look into it.)
> Well, from a certain point of view you are right, an internal gcc error
> should not be called trivial, but then it is a well known one, and we
> could do nothing except wait for a new egcs version. Unless of course we
> overcome it. As it was, the glib and gtk 1.1.12 that I built, were failing
> to compile correctly. The problem was on two files, testglib and
> testgthread (if I remember the names correctly). They would not compile
> with optimizations on. So, I compiled the two files with no optimizations
> whatsoever and the process continued with no problem. Note that these
> files are only there to test proper compiling and linking of the library,
> so optimization is totally unnecessary on them. It's not glib's fault that
> egcs fails to do full optimization on these two files. Note also that gtk
> 1.1.12 builds with no problems. It's glib that was the culprit, but it was
> a requirement for gtk that glib 1.1.12 was installed.
i think in this case the correct thing to do is to fill a bug against this package, did you do it ? you state there that the package don't compile, because of so and so, that they should apply the appended patch. and also for such important package, you are not developper and thus cannot upload the package, you send a copy the problem to the debian-ppc mailing list.
if you don't tell us the problems, there is no possibility that it will be fixed ...
maybe we should advertize this someplace on our web page ? in a faq or something ?
> > It would be nice if we ppc-developper could have a web page or another system
> > or even a central coordinator of some kind, who would assign packages to
> > people, even working ones, and each of them would receive a mail when a new
> > version of this packages have succesfully build, are ignored by dbuild or
> > failed to build for some reason, maybee with the tail of the resulting message
> > fail or something appended. This port-developper could also receive copies of
> > the bug message for the corresponding packages, and other such stuff.
> > I guess this would demand some work from us, but it would be nice to have it,
> > or at least some web page where each ppc-developper of the most important stuff
> > is listed, and also where one could volunteer (by emailing to a coordinator for
> > example) to take care of a particular important package.
> I second that, as it is, debian development is too i386-based, and other
> ports are made to look like they are always in catch-up mode. That may be
> true till now, but it doesn't have to be so. They are ports, that is
> different paths from the main course, not behind the main course.
> > ok i talked a lot, i will try to post here the output of the glib1.1.12 dbuild
> > process, or someone could tell me what to do with a gcc internal error, and i
> > will upload the packages asap.
> see above.
> > notice that the gtk/glib packages are at 1.1.5 or something such because the
> > latest two build that i tried had problem. i was able to build 1.1.9 for ppc,
> > but i had to make changes in the .orig.tar.gz, and thus didn't upload it. This
> > was just before i left for holyday. (and the n the 1.1.5 packages are less than
> > a month old, and there should be 1.2 soon)
> still that's no reason that we should wait for 1.2 and be stuck with 1.1.5
> when the cause is something like an optimization flag of egcs.
sure, but as all debian developper we have a real life beside it, and i have been busy. ...
with the info you give to me now, i wil be able to fix it, but best would be to fill a complete bug report to the package maintainer, as said above...
i will try to upload a package this weekend ...