Re: Do the HPPA "binary-only" NMUs violate the GPL?
> > clif (0.93-1.0.1) unstable; urgency=low
> > * va_arg fixes for gcc 2.96 and later. See #103683
> This is certainly in poor taste. If the HPPA people can't wait for the
> maintainer to turn around a new package, they should do a source NMU - it's
> scarcely much more work than a binary-only upload. Like any other, this rule
> is made to be broken, but one shouldn't make a habit of it.
As the one doing a fair percentage of these uploads, here is the logic I
was given by some porters (who have done other architectures) when they
told me to do binary NMU's with a patch in the BTS.
1) GPL is met, since the source is available. All information needed to
get the source is provided in the ftp archive and the .changes file.
2) there are long build-dependency chains which take forever to get built
if one package takes time to get there. In order for the build daemon
to find the package, it must be in the archive.
3) The next upload from the maintainer should in theory include the fix,
and if it doesn't, then the ported-to architecture remains out-of-date,
but no one else is affected. It is also possible that the fix breaks
another architecture, but that the breakage would not be noticed by
the porter, who is unfamiliar with the package and is just dealing
with porting issues.
In cases where the change starts getting complex, I am working with the
maintainer to incorporate the fix, possibly doing a source NMU.