[Pkg-octave-devel] Epoch in version number of octave2.1

[Dirk Eddelbuettel]

On 26 November 2005 at 23:34, Rafael Laboissiere wrote:
| With the release of octave2.9_2.9.4-1, a terrible mistake was made: the
| virtual package octave_2.9.4-1 was included in the upload.  This means
| that octave2.9 takes precedence over octave2.1.  Obviously, this is not
| what we want,  By the way, I already changed the package to produce the
| octave virtual package only for octave2.1.
| 
| I tried to get the ftp admins to remove the octave_2.9.4-1_<arch>.deb
| files from the packages pool, by filing the bug report:
| 
|     htpp://bugs.debian.org/339445
| 
| I have been told that the ftp admins will take no action and that it is our
| responsibility to fix the problem.

That's understandable given their resources, and the amount of errors us 1000
maintainers manage to produce :)

| There are (at least) two possible solutions:
| 
| First, by introducing an epoch in the version number of octave2.1. The
| next release will have number 1:2.1.72-7 and will supplement the 2.9.4-1
| one.  

I really dislike epochs. Very inelegant. OTOH they do fix the problem.

| Second, by restoring the generation of the octave virtual package by
| octave2.9 and making it depend on octave2.1.

Clever. But doesn't it require octave2.9 to be installed to get octave2.1 ?
Or simply 2.9 to be in the archive to produce a package "octave" that itself
depends on 2.1, so in a way 2.9 pointing to 2.1 and giving users 2.1 even if
it comes from a 2.9 source. That sounds nice.
 
| None of these options is ideal.  Which one do you prefer?

I had one quick idea while reading your email: "introduce another layer of
indirection", ie upload a new package octave-virtual that produces a package
octave. You can keep its version numbers distinct from Octave. I believe
emacsen-common does just that. 

But then your second suggestion already does that. Maybe that's the way to
go?

Dirk

-- 
Statistics: The (futile) attempt to offer certainty about uncertainty.
         -- Roger Koenker, 'Dictionary of Received Ideas of Statistics'