[SciPy-dev] version numbers.

[Pearu Peterson]

Hi,

On Tue, 19 Feb 2002, eric wrote:

> So, here are my proposed modification to Pearu's versions:
> 
>        <major>.<minor>.<micro>-<release_level>-<cvs-minor>.<cvs-micro>.<serial>
> 
> <major>, <minor>, and <micro> are all human specified and won't change
> that often (where should we store these anyway?).

You can hardcode it into version_template for now until we get a
better idea.

> One other question -- do we need all three of these --
> <cvs-minor>.<cvs-micro>.<serial> -- or could they combined into one
> CVS based number?  If the 3 separate numbers are valuable in their own
> rights, I'm happy to keep them around.  But if we're just looking for
> uniqueness, then we could lump them all together.

I think you can leave out <cvs-micro> that is the approx. total number of
files. <cvs-minor> should be adequate to reflect the changes in
adding or removing files, nobody will really care how many there are
files anyway. So, use <cvs-minor>.<serial> but I think you should keep
them appart as in theory one could have situation:
  2.34  -> 234
  23.4  -> 234

It is expected that the <serial> part will change most rapitly and
therefore will have large values. In order to reduce the length of the
version string, we could represent <serial> as a hexadecimal in it.

I have studied a possibility to let CVS server to calculate these
cvs-numbers. Indeed, it is possible but the corresponding script needs
some adjustements. So, I hope that the frustation with updating version
numbers in CVS should be temporary.

Regards,
	Pearu