Re: [PROPOSAL] Debian Release Plan
On Saturday 02 August 2003 12:15, Nathanael Nerode wrote:
> Perhaps the time has come to reconsider the requirement that, to be
> releaseable, all packages must be release-ready on all 11
> previously-released architectures, and in sync on all 11 architectures.
> That's a lot to keep in sync, especially when upstream doesn't care
> about some of the architectures. :-( Of course, there are already
> options individual maintainers can use to deal with such issues, such as
> declaring their packages to be non-m68k or non-s390 (for instance).
> Perhaps this should be used more aggressively. :-/
I realize that this position is unpopular, but I would have to agree. Perhaps
m68k, etc. should not be supported until Sarge-r1 (i.e. the latest release
for m68k would be Woody-r-whatever until Sarge-r1 is released with new
support). Although this could make the security team's life even more
complicated :-/
I had a vague idea which I believe would be easy to implement. I might just
try it later today, depending on how complicated it turns out to be. The
testing scripts block entry when a package is not buildable/has bugs/etc. on
any architecture for which it claims support... it should be simple to make
modified versions which would tell us what an i386-only (or whatever
architecture combination) release would be like at any time. It becomes more
complicated when dealing with RC bugs than it is with the buildds, because
they don't have architecture tags (some of them have [subject prefixes] but I
wouldn't expect that be true of all of them). Anyway, if someone else wants
to try this, go for it :-) I may try, later tonight or this weekend, and
I'll post if I get any interesting statistics. Or damn lies. Or benchmarks.
Have a nice day!
-thomas
Reply to: