Closed Bug 1224628 Opened 9 years ago Closed 9 years ago

Investigate http://hg.mozilla.org/integration/gaia-2_5

Categories

(Developer Services :: Mercurial: hg.mozilla.org, defect)

Product:
Developer Services
Component:
Mercurial: hg.mozilla.org
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(Not tracked)

Status:
VERIFIED FIXED

People

(Reporter: Pike, Assigned: hwine)

References

Details

Attachments

(1 file)

disable v2.5 processing in bumper
4.73 KB, patch
KWierso
: review+
Details | Diff | Splinter Review
It seems that something's funky with http://hg.mozilla.org/integration/gaia-2_5, the SHAs don't seem to match up between http://hg.mozilla.org/integration/gaia-2_5 and http://hg.mozilla.org/integration/gaia-central. Example: http://hg.mozilla.org/integration/gaia-central/rev/f9329944fe8c is a changeset from 2014, but http://hg.mozilla.org/integration/gaia-2_5/rev/f9329944fe8c doesn't find a revision. We're populating the en-US repos for l10n by using hg convert (with tweaks), so the difference in hashes means that we can't use the existing system to treat 2.5 as a branch of gaia-central.
PS: Nothing on my side is currently using the repo. If I'm really the only consumer of this repo, I'm in no way bound to its current state, and don't mind it being wonky, reset, talked to, whatever is needed.
Notes: this *might* be related to the "multiple roots detected" hook issues encountered during original branching. The other user of this repo is the bumper bot -- if hashes are reset, some one time manual changes may be needed.
Per IRC conversation, somehow Git commit 98187f837d7c514205c70ad41a0c13de713fec50 was converted differently between gaia-central and gaia-2_5. This caused SHA-1 divergence between the two repos.
Just spent a few mins, no idea if helpful. https://bitbucket.org/durin42/hg-git/commits/8e17f8a16269 is a commit where a few revision IDs changed in tests/test-git-submodules.t, dating to 2014-12-30
Per #vcs plan from Friday, starded copy of gaia-central: 11:29 < hwine> | How does this sound as a path forward: recopy the conversion repo from gaia central to new-2.5; hg strip & git "strip" everything past 2.5 branch point; restart 2.5 conversion 11:29 < gps>| i approve
Assignee: nobody → hwine
Status: NEW → ASSIGNED
update: conversion ongoing (In reply to Axel Hecht [:Pike] from comment #5) > https://bitbucket.org/durin42/hg-git/commits/8e17f8a16269 is a commit where > a few revision IDs changed in tests/test-git-submodules.t, dating to > 2014-12-30 Yes, gps noticed this as well. We should be pinned to a version prior to that, so it wouldn't affect things. In any case, the current approach should ensure no divergence as it doesn't replace the original changeset.
https://hg.mozilla.org/integration/gaia-central/graph now has arcs for both master and 2.5
See Also: → 1225465
Status: - The integration/gaia-central repository (bug 1225465) is now correct and in production. - Some issues remain with integration/gaia-2_5 (this bug), but that branch is broken for other reasons atm. - Plan is to disable gaia v2.5 support in bumper, to avoid mismatch with b2g44-v2_5 gecko and resulting multi-minute time penalty per bumper run. - once repo is fixed, bumper will be re-enabled.
per plan in comment 9, disable gaia v2.5 branch bumper processing. No file => no processing. QED
Attachment #8689230 - Flags: review?(nthomas)
Comment on attachment 8689230 [details] [diff] [review] disable v2.5 processing in bumper Hwine asked me to rubberstamp this. Happy to oblige.
Attachment #8689230 - Flags: review+
Attachment #8689230 - Flags: review?(nthomas)
Depends on: 1226463
bug 1226463 fix landed & deployed. Manual run shows good results, so re-enabled b2g_bumper per comment 13 by hg backout of attachment 8689230 [details] [diff] [review]
b2g_bumper triggered builds for v2.5 are running again. First build can be tracked at https://treeherder.mozilla.org/#/jobs?repo=mozilla-b2g44_v2_5&revision=589a9b96c5bc ni: Pike - can you report on the status of the l10n dashboard please? Do the issues appear to be resolved? ni: nhirata - can you update on the status of the v2.5 build gaia/gecko matching please?
Flags: needinfo?(nhirata.bugzilla)
Flags: needinfo?(l10n)
I download the newer Aries 2.5 Dogfood build to do verification. The Gaia commit number within Aries build matches the newer commit on both sides (Github and HG). Flame 2.5 build is under confirmation. # Build info Build ID 20151120074230 Gaia Revision b7f130c36588e308784cd1b7c8a28585c9819076 Gaia Date 2015-11-20 05:09:18 Gecko Revision 589a9b96c5bc632bcc336d0543442983cc1f61dd Gecko Version 44.0a2 Device Name aries Firmware(Release) 4.4.2 Firmware(Incremental) eng.worker.20151120.064943 Firmware Date Fri Nov 20 06:49:51 UTC 2015 Bootloader s1 # Gaia commit on Github https://github.com/mozilla-b2g/gaia/commit/b7f130c36588e308784cd1b7c8a28585c9819076 # Gaia commit on HG http://git.mozilla.org/?p=releases/gaia.git;a=commit;h=b7f130c36588e308784cd1b7c8a28585c9819076
Update: There are no known issues at this time. Just leaving open until positive verification for: - fxos, v2.5 - gaia l10n
Leaving my needinfo open for a few more, still grokking the universe here. One thing I found is that https://hg.mozilla.org/integration/gaia-central/rev/19cd7f783ded seems to be a 2.5 changeset in reality. It's the only one of its kind I found, and it's parent is fine, AFAICT. It's one of three heads, not sure what it is about changeset: 19523:bf01d5808391 bookmark: master parent: 19521:7cb565d70183 parent: 19522:c8788891b795 user: Andrew Sutherland <asutherland@asutherland.org> date: Tue Jul 09 14:27:02 2013 -0700 summary: Merge pull request #10872 from asutherland/bug_852369
Update the verification status. Checking the latest Flame 2.5 user build and tracing commit history, I also don't find any problem and no exceptional case was reported from testers. # Build info Device: Flame-KK v2.5 Gaia-Rev 5839f17dedc757947c9531dc0d66c3c49119f5ea Gecko-Rev 3a45ca93dd447e046baa2c7590f60ae008e438f8 Build-ID 20151123163948 Version 44.0a2 Bootloader L1TC000118D0 # Gaia commit on Github https://github.com/mozilla-b2g/gaia/commit/5839f17dedc757947c9531dc0d66c3c49119f5ea # Gaia commit on HG http://git.mozilla.org/?p=releases/gaia.git;a=commit;h=5839f17dedc757947c9531dc0d66c3c49119f5ea
Blocks: 1227473
OK, l10n is looking good now, thanks for your help. Both central and 2.5 are now updated. I haven't yet restarted the cron job, because of the stray head(s) below. The stray head below isn't a major concern to the l10n side, but it might be good to have a sanity check on that? (In reply to Axel Hecht [:Pike] from comment #18) > Leaving my needinfo open for a few more, still grokking the universe here. > > One thing I found is that > https://hg.mozilla.org/integration/gaia-central/rev/19cd7f783ded seems to be > a 2.5 changeset in reality. It's the only one of its kind I found, and it's > parent is fine, AFAICT. > > It's one of three heads, not sure what it is about > > changeset: 19523:bf01d5808391 > bookmark: master > parent: 19521:7cb565d70183 > parent: 19522:c8788891b795 > user: Andrew Sutherland <asutherland@asutherland.org> > date: Tue Jul 09 14:27:02 2013 -0700 > summary: Merge pull request #10872 from asutherland/bug_852369
Flags: needinfo?(l10n) → needinfo?(hwine)
(In reply to Axel Hecht [:Pike] from comment #20) > OK, l10n is looking good now, thanks for your help. Yay! > Both central and 2.5 are now updated. I haven't yet restarted the cron job, > because of the stray head(s) below. > > The stray head below isn't a major concern to the l10n side, but it might be > good to have a sanity check on that? Multiple heads, especially on gaia-central, are "okay for now" if they don't break your system. (There are also multiple roots.) > It's one of three heads, not sure what it is about > > changeset: 19523:bf01d5808391 > bookmark: master > date: Tue Jul 09 14:27:02 2013 -0700 Thanks for pointing out that the bookmark "master" was stuck. It appears to have been stuck since that commit. I just "unstuck" that bookmark, and it now correctly shows as being aligned with 'tip' on the 'default' branch. If you do not see that in your repository after a pull, you may need to delete the stuck bookmark locally first: hg bookmark -d master The next pull will bring restore the bookmark from the server. If I understood your message correctly, you are clear to restart your cron jobs.
Flags: needinfo?(hwine)
agreed, cron job is running again.
On today's builds 2.5 branch shows : * gecko: http://hg.mozilla.org/releases/mozilla-b2g44_v2_5/rev/89b4f90d2865e62b9d898655cf082902cf2572a0 * gaia : e05621cbfd92b3fb4e5aef86621c57cd68fb0414 ( last 2.5 gaia ) The 2.5 builds that are being built currently on taskclusters are as follows : aries-dogfood (debug) :ok nexus-4-kk-ota (debug): ok nexus-5-l-ota (debug) : ok flame-kk-ota (debug) : oddity that it's not the same pull as the above. Having said that it's still 2.5 http://hg.mozilla.org/releases/mozilla-b2g44_v2_5/rev/33a575e711faf3344aa2e31ca2ea066b4cd8aafa gaia revision : 07462becf08f0c26ebd64daf89646e7403a336c5 ( second to last checkin ) on pvtbuild server : we have no recent 2.5 builds.
Flags: needinfo?(nhirata.bugzilla)
emulator ics and flame-kk builds were added to buildbot in bug 1229418.
Closing this bug, as the repositories are back in sync, confirmed in comment 23 If there are build issues (what's building what where), separate bugs should be opened with the correct team (TC/BB - transition periods are no fun).
Status: ASSIGNED → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
from comment 23. Verified with the builds that are being built currently. There's other issues with 2.5 which seem like build issues, not repo issues.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: