Bug#27906: PROPOSED] Binary-only NMU's

[Paul Slootman <paul@wau.mis.ah.nl>]

On Mon 19 Oct 1998, Ian Jackson wrote:
> Paul Slootman writes ("Bug#27906: PROPOSED] Binary-only NMU's"):
> ...
> > If you're saying that each and every binary version should be accompanied
> > with corresponding source only when a release is made, then the whole
> > problem could be circumvented by making the bug report with the diffs
> > severity: important; that would prevent the package being released until
> > the bug is closed (and hence the patches being included in a new upload),
> > right?
> 
> This is a nice idea for a simple fix, but I think it has two problems:
> 
> 1. A `Severity: important' bug doesn't prevent us _distributing_ the
> binaries, just from _releasing_ with that version in our current
> frozen tree.  Thus, we'd still be distributing binaries with no
> corresponding sources, violating our social contract and eg the GPL.
> 
> 2. That puts porters' patches, which are usually not important except
> to the port(s), on the critical path for the release.

I think that [2] follows from [1], because the package should _not_
be released if there is a NMU diff in the BTS, because otherwise
we'll be releasing a binary package without corresponding source in
that same release...

The critical phrase is "except to the port(s)". As the port's binary
package is built _with_ those patches, we can't release the package
in that port if the patches aren't included in the source release.
I do hope you're not contemplating different source release fo the
different architectures?   ... no, you _can't_ :-)


Paul Slootman