On 2015-02-27 8:46 AM, Ben Hearsum wrote:
On 2015-02-26 05:57 PM, Mike Hoye wrote:
Instead, I propose that we stand up a variety of old, slow, minimum-spec
"approved" development boxes, build on them daily. If forward progress
means we need to abandon a category of developers' machines, I ask that
we try to do that with as much transparency and advance notice as we'd
grant a category of users in the same position.
Monitoring things that only build on a nightly basis has proven to be
terribly difficult. Nightly only jobs (that don't have equivalent CI
jobs) are hidden on treeherder/tbpl, and issues tend to go unnoticed for
weeks. Bustage to jobs like this isn't usually cause for backouts either.
I've had a couple of conversations with engineers and managers about
this in various fora. Consensus seems to be that this is kind of a crap
situation to have fallen into inadvertently and my heart's in the right
place but the cost-benefit fixing this is not clear.
And that's without considering backing out the relevant changes, which
we'd need a fantastic reason I don't have to justify. So barring some
significant command-line-option insight in the next few days I'm going
to update the relevant documentation, wontfix 1137346, and put some cron
jobs in some VMs in the hope that this class of problem doesn't sneak up
on us again.
- mhoye
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform
