We have had some discussions about this on IRC but it is time for an official decision.
The number of devs/TUs using i686 is ever decreasing. I am now moved to an x86_64 machine so my i686 signoffs will disappear and there are very few others who do signoff for i686.
So I would propose: 1) x86_64 signoff policy stays the same 2) i686 only requires one signoff 3) the one i686 signoff can be waived if there is a testsuite that passes
Comments?
Allan
Am 22.04.2013 12:31, schrieb Allan McRae:
So I would propose:
- x86_64 signoff policy stays the same
- i686 only requires one signoff
- the one i686 signoff can be waived if there is a testsuite that passes
Thanks for bringing this up to make it finally an official policy. This might need some changes to archweb though.
Greetings,
Pierre
On Mon, Apr 22, 2013 at 5:45 AM, Pierre Schmitz pierre@archlinux.de wrote:
Am 22.04.2013 12:31, schrieb Allan McRae:
So I would propose:
- x86_64 signoff policy stays the same
- i686 only requires one signoff
- the one i686 signoff can be waived if there is a testsuite that passes
Thanks for bringing this up to make it finally an official policy. This might need some changes to archweb though.
The default "# of signoffs to require" would be easy enough to configure based on architecture. I'll spike this right now.
Allan, for point (3), I wonder if some more makepkg meta info would come in handy here? We now record checkdeps and makedeps, but also recording
function = build function = package function = check
or something like that might be useful for this and other tools. A even more sophisticated approach might record the total time each of these steps took...food for thought.
-Dan
On 22/04/13 23:31, Dan McGee wrote:
On Mon, Apr 22, 2013 at 5:45 AM, Pierre Schmitz pierre@archlinux.de wrote:
Am 22.04.2013 12:31, schrieb Allan McRae:
So I would propose:
- x86_64 signoff policy stays the same
- i686 only requires one signoff
- the one i686 signoff can be waived if there is a testsuite that passes
Thanks for bringing this up to make it finally an official policy. This might need some changes to archweb though.
The default "# of signoffs to require" would be easy enough to configure based on architecture. I'll spike this right now.
Allan, for point (3), I wonder if some more makepkg meta info would come in handy here? We now record checkdeps and makedeps, but also recording
function = build function = package function = checkor something like that might be useful for this and other tools. A even more sophisticated approach might record the total time each of these steps took...food for thought.
We could... or we could store the build logs that are always generated when using makechrootpkg? I think they are always useful to refer back to.
I don't think time spent on these is a useful metric.
Allan
On Mon, Apr 22, 2013 at 5:45 AM, Pierre Schmitz pierre@archlinux.de wrote:
Am 22.04.2013 12:31, schrieb Allan McRae:
So I would propose:
- x86_64 signoff policy stays the same
- i686 only requires one signoff
- the one i686 signoff can be waived if there is a testsuite that passes
Thanks for bringing this up to make it finally an official policy. This might need some changes to archweb though.
Changes for (2) rolled out to archweb today; all signoffs for future i686 packges will be set to only require 1 by default. (3) will take a bit more legwork to make it easy for archweb to determine if a check function exists, etc.
-Dan
arch-dev-public@lists.archlinux.org