VendorRelease and VendorString policy in recent servers

[Kalle Vahlman]

Hi!

We are currently facing a bit of a dilemma with the VendorRelase
numbers in our work[1] and I was wondering if there was a defined
policy for this.

AFAIK the VendorRelease is a number tied to a particular VendorString.
So far everything has been fine and dandy as the numbers have steadily
increased. But apparently since the 1.3 release or so the
VendorRelease number starts over. There is mixed results in a quick
survey over how different distros report this, for example Debian Etch
and Ubuntu Feisty report 7.x but Debian Sid apparently tells 1.4 and
Gentoo reports small numbers already starting from 1.3.

These differences make checks like this break:

  http://gitweb.freedesktop.org/?p=cairo;a=blob;f=src/cairo-xlib-surface.c#l2011

This means that currently cairo will think that some servers on some
distros (or when the server is "hand-build") have the bug in question
even if they don't, thus hitting fallbacks which potentially include
roundtrips to the server.

So the questions become:

 1. Is there a defined policy on how the numbering proceeds or did
this "just happen"?
 2. Why the VendorString wasn't changed even though the numbering
changed radically?

I guess we could just change it to 7.3 for our builds, or just limit
the cairo check between known X.Org versions, but I'd rather see a
better solution for all this...


[1] http://movial.fi/en/company/newsroom/News_and_Events/?NewsId=72

-- 
Kalle Vahlman, zuh at iki.fi
Powered by http://movial.fi
Interesting stuff at http://syslog.movial.fi