Closed
Bug 507952
Opened 15 years ago
Closed 1 month ago
Explore providing more continuity for release {tinder-|moz}configs
Categories
(Camino Graveyard :: General, defect)
Tracking
(Not tracked)
RESOLVED
INCOMPLETE
People
(Reporter: alqahira, Assigned: alqahira)
References
()
Details
Attachments
(4 files)
Right now we fork a tinder-config for releases for the first milestone on a branch, and then we just just rsync the release config over from the previous milestone/release for each new release and run a sed on it. This works well enough, except 1) it's one more thing that has to be done manually on the box itself 2) there's not currently any (automated) way of preserving continuity for configs, particularly in case of acts of $DEITY 3) we periodically make changes to {tinder-|moz}configs over the course of normal development, and then we have to remember to port them to the release configs before spinning (there's a somewhat lengthy note stressing this point in Branch Tag Build Release) 4) probably something else I'm not remembering at the moment I haven't given this a ton of thought/investigation yet, but I'm thinking we could do something like this: 1) Create a new tinderbox-config, release/, per branch, with the initial version forked from the nightly builder (for 1.6.x and 2.0.x, from the existing "release" configs, merged and cleaned up to match the nightly ones) 2) Make a new tree on the nightly/release box, $branchname-release, whose tinderbox-configs/ folder pulls from the new release/ tinderbox-config above 3) Instead of adding a new tree for each milestone/release in multi-config.pl, multi-config.pl would be set up for $branchname-release with --config-cvsup-dir, and spinning each release would involve just uncommenting the tree; it would cvs up the config and start a clobber release build (for absolute freshness, we could still rsync --delete the source and object hierarchies from the $branchname-release tree to simulate the current rsync of only tinderbox and config files) 4) Branching and tagging would involve bumping the numbers and tag (and/or assorted other config settings) in the release/ {tinder-|moz}configs in addition to client.mk and camino/config/version.txt 5) Any time we modify a {tinder-|moz}config during development, we'd also modify release/, just like it was cb-miniosx01/ or any other (real) box I haven't really checked to see if this will all work as I imagine it, and reading back over it after I've written it all down now, I wonder whether it's actually creating more work than it's saving (and preserving). I guess option B would be to write a tool that we keep under version control and check out on the tinderbox that copies the parent branch mozconfig, copies the parent branch tinder-config.pl and prompts you for values that it replaces in the copy, and does some sort of sanity-check. Dunno. It's been bothering me while Sam's been hanging out with hackers all week, so I'm putting pen to paper as it were, and now it can bother everyone else, too ;)
Assignee | ||
Comment 1•15 years ago
|
||
Just so we'll have some record/archive, here's the release tinder-config from 2.0b3.
Assignee | ||
Comment 2•15 years ago
|
||
Just so we'll have some record/archive, here's the release tinder-config from 1.6.8.
Assignee | ||
Comment 3•14 years ago
|
||
I thought about this again today randomly while in the shower, and I basically came up with comment 0 all over again. Instead of part 2 and 3, we'd keep the traditional $version trees for the releases and just make sure their --config-cvsup-dir values were set to the release tag. In part 4, the changes would be done on the mainline (like relnotes and installer/Makefile.in) so that the mainline always had the most current values; it would just be tagged for the release. This is maybe a little less work (at least while releases are in CVS; it gets a little more annoying once we're in Hg, since we'll have to juggle two VCSes and repos). In the meantime, I've been periodically pulling the tinder-configs off the box, so I have local copies.
Assignee | ||
Comment 4•13 years ago
|
||
…And 2.1a1.
Assignee | ||
Comment 5•12 years ago
|
||
I need to pull the one for 2.1.2, which has a larger Gecko timeout value, and I updated it tonight for the new symbolpush server, so that doesn't fail on release :P
Assignee | ||
Comment 6•12 years ago
|
||
You need to log in
before you can comment on or make changes to this bug.
Description
•