No GV Nightly since Oct 04: beetmover jobs aren't part of the graph anymore
Categories
(Firefox Build System :: Task Configuration, defect, P1)
Tracking
(firefox-esr60 unaffected, firefox-esr68 unaffected, firefox67 unaffected, firefox68 unaffected, firefox69 unaffected, firefox70 unaffected, firefox71 fixed)
Tracking | Status | |
---|---|---|
firefox-esr60 | --- | unaffected |
firefox-esr68 | --- | unaffected |
firefox67 | --- | unaffected |
firefox68 | --- | unaffected |
firefox69 | --- | unaffected |
firefox70 | --- | unaffected |
firefox71 | --- | fixed |
People
(Reporter: jlorenzo, Assigned: nalexander)
References
(Regression)
Details
(Keywords: regression)
Attachments
(1 file, 1 obsolete file)
The latest ship_geckoview
jobs[1] don't have any beetmover jobs anymore, causing maven.mozilla.org to not serve the latest geckoview nightlies.
This is caused by lines like [3] (which were landed as part of bug 1580356): the beetmover jobs depend on some signed bits, otherwise they get filtered out.
[1] For instance: https://tools.taskcluster.net/groups/MzzN5YhUSlaP9lYvgeaGcQ
[2] https://maven.mozilla.org/?prefix=maven2/org/mozilla/geckoview/geckoview-nightly-x86_64/
[3] https://hg.mozilla.org/mozilla-central/rev/1ffe2e3d6e19a351575ac986c43155ed8e8d90bc#l3416.57
Assignee | ||
Comment 1•6 years ago
|
||
jlorenzo: so I agree this is "caused" by the lines that you link, but what is to be done? The BM jobs very carefully arrange to depend on both build
and build-signing
jobs, and at this time there's no sensible notion of "signing" mobile/android
jobs. (We'd be signing an AAR. It would make more sense to sign the old target.maven.zip
, but that's gone. Is there any advantage to signing an AAR? Is the signature even part of the BM job?). It seems to me that BM should move the results of the Ngv job rather than muck around with quasi-signed artifacts, but I don't have strong opinions here. Thoughts?
Reporter | ||
Comment 2•6 years ago
|
||
I'm sorry for the lack of context of my first comment. Starting bug 1548545, we've been gpg-signing maven artifacts, including AARs[1]. See this beetmover task definition[2], as an example:
{
"locale": "en-US",
"paths": {
"public/build/maven/org/mozilla/geckoview/geckoview-nightly-armeabi-v7a/71.0.20191002033852/geckoview-nightly-armeabi-v7a-71.0.20191002033852-javadoc.jar.asc": {
"checksums_path": "geckoview-nightly-armeabi-v7a-71.0.20191002033852-javadoc.jar.asc",
"destinations": [
"maven2/org/mozilla/geckoview/geckoview-nightly-armeabi-v7a/71.0.20191002033852/geckoview-nightly-armeabi-v7a-71.0.20191002033852-javadoc.jar.asc"
]
},
"public/build/maven/org/mozilla/geckoview/geckoview-nightly-armeabi-v7a/71.0.20191002033852/geckoview-nightly-armeabi-v7a-71.0.20191002033852-sources.jar.asc": {
"checksums_path": "geckoview-nightly-armeabi-v7a-71.0.20191002033852-sources.jar.asc",
"destinations": [
"maven2/org/mozilla/geckoview/geckoview-nightly-armeabi-v7a/71.0.20191002033852/geckoview-nightly-armeabi-v7a-71.0.20191002033852-sources.jar.asc"
]
},
"public/build/maven/org/mozilla/geckoview/geckoview-nightly-armeabi-v7a/71.0.20191002033852/geckoview-nightly-armeabi-v7a-71.0.20191002033852.aar.asc": {
"checksums_path": "geckoview-nightly-armeabi-v7a-71.0.20191002033852.aar.asc",
"destinations": [
"maven2/org/mozilla/geckoview/geckoview-nightly-armeabi-v7a/71.0.20191002033852/geckoview-nightly-armeabi-v7a-71.0.20191002033852.aar.asc"
]
},
"public/build/maven/org/mozilla/geckoview/geckoview-nightly-armeabi-v7a/71.0.20191002033852/geckoview-nightly-armeabi-v7a-71.0.20191002033852.pom.asc": {
"checksums_path": "geckoview-nightly-armeabi-v7a-71.0.20191002033852.pom.asc",
"destinations": [
"maven2/org/mozilla/geckoview/geckoview-nightly-armeabi-v7a/71.0.20191002033852/geckoview-nightly-armeabi-v7a-71.0.20191002033852.pom.asc"
]
}
},
"taskId": "XzYkpxoSQP-8vu1xwBm-Jw"
}
This is the reason why the beetmover jobs depend on build-signing
. To me, the fix is to remove the disable-build-signing: true
lines from:
android-api-16-nightly/opt
.android-aarch64-nightly/opt
android-x86_64-nightly/opt
android-x86-nightly/opt
Please let me know if you need more information 🙂
[1] For instance: https://maven.mozilla.org/?prefix=maven2/org/mozilla/geckoview/geckoview-nightly-x86_64/71.0.20191004094656/, see the .asc files.
[2] https://tools.taskcluster.net/groups/Yi9DecphRRWvoJY15d_Pkg/tasks/drbDj483Q5W07Y2J894q6A/details
Assignee | ||
Comment 3•6 years ago
|
||
Try build with the simple backout is percolating at https://treeherder.mozilla.org/#/jobs?repo=try&revision=2c734fe28520b7b366f9fde446831b0515b8e774. My guess is that we want to remove the beetmover
kind entirely, unless it has application across trees. There are no Fennec beets remaining!
Assignee | ||
Comment 4•6 years ago
|
||
Bug 1580356, which removed Fennec from mozilla-central, was too
aggressive. That ticket reasoned that since there's no longer an APK
to sign, that build-signing should be disabled. However,
build-signing is more general: since Bug 1548545, GeckoView artifacts
have been GPG signed. The beetmover tasks expect those signed
artifacts, and without them, nothing gets published to
maven.mozilla.org.
This simply reverts the parts of Bug 1580356 that removed the
build-signing tasks from the taskgraph.
Assignee | ||
Comment 5•6 years ago
|
||
jlorenzo: the patch you pushed to try (https://hg.mozilla.org/try/rev/b4e2404d6d381e2254c18e1cf09093113079d9d6) makes sense to me. I'd go further and remove fennec_*
and the beetmover
kind -- if removing the APK handling is okay (across repositories) then removing this must be okay too --but that could also be follow-up. Can you commandeer the revision so we can get this landed?
Reporter | ||
Comment 6•6 years ago
|
||
Updated•6 years ago
|
Comment 10•6 years ago
|
||
bugherder |
Assignee | ||
Comment 11•6 years ago
|
||
And we're back, folks: I see a BM-GV
job and I see that https://maven.mozilla.org/?prefix=maven2/org/mozilla/geckoview/geckoview-nightly/71.0.20191011093010/ is looking happy and healthy.
Great work, Johan and Mihai!
Updated•6 years ago
|
Updated•4 years ago
|
Description
•