Re: RFS: libmesh
Ondrej Certik wrote:
* debian/changelog contains Closes: #nnnn, because although I filled
the ITP bug, I still didn't receive the email with the bug number
* although mentors.debian.net says no lintian errors, their scripts
don't work, because there are in fact the following two errors:
wrong-bug-number-in-closes
sharedobject-in-library-directory-not-actually-a-shlib usr/lib/libmesh.so
The first is caused by my first (*), the second is caused by wrong
shlibs files. Could you please help me how to write the debian/rules
correctly? The upstream Makefile is quite broken, it doesn't have the
"install" rule, so I am copying the library by hand in debian/rules.
And also I don't know how to correctly handle the shlibs.
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=426734 seems to have
appeared now.
* the build-depends are fine (builds in pbuilder), but the depends for
libmesh0.6.0 don't include the petsc library - but I believe that
fixing the previous point will just fix this one automatically
* the debian/rules contains some unnecessary things, but I just left
it there until we fix the above problems
* also please check that I removed all the non-free things correctly
This doesn't look terribly portable (what's it going to do on hurd??) to me:
ls lib/i486-pc-linux-gnu_opt/*
cp lib/i486-pc-linux-gnu_opt/libmesh.so $(meshbin)
($shell gcc -dumpmachine) might be what you're looking for?
Also section 1 of [1] is relevant here too I think, in that (your rules
at least) don't look very versioned to me.
It's also generally better to use install or even dh_install instead of cp
# dh_makeshlibs
This line needs to be uncommented. This will go someway towards fixing
the lintian problem. People also like complaining about all the
commented lines left in from the template you used. E.g. there's no
point having #dh_installpam hanging around in your debian/rules
I noticed the following: "/usr/bin/make -j2" I don't think parallel
building is considered the done thing in debian/rules, at least that
seemed to be how I remmeber the discussion in [0] going. People don't
like the smaller buildds getting hit like that.
I haven't looked over the dfsg-ness yet. I think it's considered good
practice to document somewhere what had to go and why to make it dfsg.
Either way it'd be quite handy for me. Is it algorithm that's non-free?
Or documentation?
I'd quite like to see libmesh in Debian, as my own work seems to be
heading more and more towards numerical solutions of PDEs and libmesh
looks like it's worth trying out.
I'll take another look over the package again later for you. Feel free
to ask more questions if you like :-)
Alan
[0] http://lists.debian.org/debian-policy/2001/03/msg00368.html
[1] http://www.debian.org/doc/debian-policy/ch-sharedlibs.html
Reply to: