Re: RFS: iulib (5th attempt)
Hi Charles,
2009/7/19 Charles Plessy <plessy@debian.org>:
> our tools do not invoke the build target twice in a row, but real people
> sometimes have a good reason to do. For instance, I tried to build iulib it did
> not work because I was missing dependancies, and when I installed them and
> called the build target again, it triggered the bug discussed above. A good
> makefile should be resilient to this.
How did you do this? Simply running
$ ./debian/rules build
a second time, produces:
dh --with quilt build
and dh knows it isn't necessary and doesn't build.
> Talking about dependancies, since there are only three versions of the libtool
> package in circulation, you can probably depend on versions higher than Lenny.
> Alternatively, you can support both: it looks easy. It would be especially
> useful if you think that there are chances your package would be backported.
> If the scons migration for the 0.4 version implies to drop libtool, you do not
> need to put too much energy into this…
The latter. I want to get a clean package into the NEW queue, and
while it sits there, I'll work on v0.4.
> A much more problematic issue is the binary dependancies of libiulib0. You have
> devised a long list by hand, but it is not accurate for Sid, which is the
> target of your package. There is at least one error: libavcodec51 does not
> exist there. Do you have evidence that ${shlibs:Depends} is not enough?
No. I've removed everything but ${shlibs:Depends}, ${misc:Depends}
> Probably. Maybe you can submit a whishlist bug, or if you have time, try to
> write such a check.
Done[1,2].
> PS: do you know how to write a symbols control file? (me not).
dpkg-gensymbols -O -plibiulib0
gives me output... But how to know I am using it correctly...
I've uploaded the updated package to mentors.
Regards
Jeff
[1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=537768
[2] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=537769
Reply to: