Re: ITK debian packages
Hello Andreas,
Thank you for your input; I think I agree with you.
But to be clear: it was actually Brad King (upstream author) that I
was soliciting an opinion from, because in a previous message of this
thread he said:
If you want to be sure, I suggest you package GCC-XML with the
version number from the date the ITK 2.4 branch was made, which is
2005-11-29. Therefore the GCC-XML version number should be
something like 0.7.20051129. Note that odd minor versions of
GCC-XML are development versions so the patch level is just the
date.
That sounds like he is suggesting to modify the patch level directly.
I would prefer to use what I believe is a more common Debian
convention, where the version used is "${UPSTREAM}+cvs${DATE}. With
this convention, gccxml is version 0.7.0+cvs20060226 and cableswig is
version 0.1.0+cvs20060226.
On Sun, Feb 26, 2006 at 11:34:57AM +0100, Andreas Tille wrote:
> On Sat, 25 Feb 2006, Steve M. Robbins wrote:
>
> >The resulting binary reports itself as "2.7.0" (gccxml --copyright)
> >presumably from the file gxConfigure.h which contains:
> >
> >#define GCCXML_VERSION "0.7"
> >#define GCCXML_VERSION_FULL "0.7.0"
> >#define GCCXML_VERSION_MAJOR 0
> >#define GCCXML_VERSION_MINOR 7
> >#define GCCXML_VERSION_PATCH 0
> >
> >Is your suggestion that I edit gxConfigure.h and change the patch
> >value to be the date (e.g. 20060121)?
>
> As I wrote in my previous mail: Just leave the source as is and
> add the CVS date to the Debian package version.
Oh! Maybe I have the convention mixed up. Are you saying it should
be 0.7.0-20060226 for gccxml?
On Sun, Feb 26, 2006 at 11:33:09AM +0100, Andreas Tille wrote:
> On Sun, 26 Feb 2006, Steve M. Robbins wrote:
> >So Brad, what do you suggest for versioning? The Debian practice
> >would be to label a CVS snapshot from 2006-02-26 as 0.1.0+cvs20060226.
> >Is that reasonable or do you folks prefer what you suggested for
> >gccxml, namely replace the patch number with the date, i.e.
> >0.1.20060226?
>
> The Debian versioning is just a convention that made sense for
> a lot of packages. While I do not think that patching the source
> is necessary to reflect this marking of a CVS snapshot I would
> stick to this naming scheme. But this is more or less a suggestion
> that you *might* follow if you think it is reasonable.
Regards,
-Steve
Reply to: