| 1 |
On Fri, 20 May 2005 15:51:51 -0500 Brian Jackson <iggy@g.o> |
| 2 |
wrote: |
| 3 |
| Wouldn't it be better from a QA perspective to go back to the (really) |
| 4 |
| old policy of dropping anything you can't test on. I know that puts |
| 5 |
| more work on you guys, but this is only going to get worse as we get |
| 6 |
| more devs. Wouldn't it be better to nip this in the bud now. Maybe |
| 7 |
| broaden the arch teams by giving some devs access to remote boxes. |
| 8 |
|
| 9 |
Not really. Dropping to ~arch when bumping works well. Sure, ~arch does |
| 10 |
occasionally end up broken, but it's better than us lagging behind |
| 11 |
massively. There're too many packages and not enough people these |
| 12 |
days... |
| 13 |
|
| 14 |
The assumption is, if foo-1.2 works on, say, sparc, then foo-1.3 |
| 15 |
probably will too to the extent that we're happy for it to go to ~sparc. |
| 16 |
On the other hand, we're *not* confident enough in upstreams' abilities |
| 17 |
to always put out perfect releases that we're prepared to move things to |
| 18 |
stable without explicit testing. |
| 19 |
|
| 20 |
See, we *really* don't want arch to get broken. We'd rather ~arch didn't |
| 21 |
break either, of course, but taking the occasional hit there is |
| 22 |
acceptable if it lets us keep everything up to date. |
| 23 |
|
| 24 |
| Get every dev access to all the supported arches (some of this could |
| 25 |
| probably be done with emulators of some sort, qemu or somesuch). Make |
| 26 |
| them test on every arch before they change any keywords. |
| 27 |
|
| 28 |
Not gonna happen. Emulators don't cut it and won't find all the problems |
| 29 |
(but they will find a load of other bogus non-issues). Plus, from |
| 30 |
experience I'd say that at least half our devs wouldn't have a clue |
| 31 |
where to start when doing arch testing... |
| 32 |
|
| 33 |
Then there's the issue of most alt-archs having far higher QA standards |
| 34 |
than x86 anyway, and us not wanting to sink to what x86 considers |
| 35 |
acceptable for marking stable. |
| 36 |
|
| 37 |
From experience -- the current policy as it is now *works*, so long as |
| 38 |
everyone follows it. |
| 39 |
|
| 40 |
-- |
| 41 |
Ciaran McCreesh : Gentoo Developer (Vim, Shell tools, Fluxbox, Cron) |
| 42 |
Mail : ciaranm at gentoo.org |
| 43 |
Web : http://dev.gentoo.org/~ciaranm |
Archives