Closed Bug 685166 Opened 8 years ago Closed 8 years ago

Port the work to enable mozconfigs to live in the source tree

Categories

(Thunderbird :: Build Config, defect)

defect
Not set

Tracking

(thunderbird12 fixed, thunderbird13 fixed, thunderbird-esr1012+ fixed)

RESOLVED FIXED
Thunderbird 14.0
Tracking Status
thunderbird12 --- fixed
thunderbird13 --- fixed
thunderbird-esr10 12+ fixed

People

(Reporter: standard8, Assigned: jhopkins)

References

Details

(Keywords: meta, Whiteboard: [ETA ?])

Attachments

(2 files, 5 obsolete files)

Firefox have been working on putting their mozconfigs in the source tree. This will make it simpler at merge time, as any differences from central to aurora (for example) will just be automatically merged across, and we won't need to do a separate analysis in the build-config repository for what needs to be merged across mozconfigs.
Here is the Firefox bug:

Bug 558180 - Move mozconfigs into source tree
Blocks: 698843
If we could start by copying the mozconfigs into comm-central, that would help me with bug 698843: I need to tweak the mozconfigs to ALWAYS_RUN_CLIENT_PY (currently only done on tryserver).
Attachment #572893 - Flags: review?(mbanner)
Comment on attachment 572893 [details] [diff] [review]
copy of buildbot-configs/thunderbird/* mozconfigs

Actually, now I look at what Firefox is doing, I believe this wants to be:

- You land only the mozconfigs for a particular version on that branch.
- The mozconfigs are renamed and structured thus:
-- <platform>/<type>
--- where type is one of debug, release, nightly, l10n-mozconfig etc.
Attachment #572893 - Flags: review?(mbanner) → review-
Depends on: 703232
<Standard8> jhopkins: the point is that if central has a change in mozconfig, it'll automatically get merged across to aurora etc at the right times
<jhopkins> Standard8: ah, i had thought debug flags might be an issue
<Standard8> jhopkins: shouldn't be, all our builds should have separate configs
Attached patch comm-aurora mozconfigs (obsolete) — Splinter Review
Assignee: nobody → jhopkins
Attachment #572893 - Attachment is obsolete: true
Status: NEW → ASSIGNED
Attachment #600173 - Flags: review?(mbanner)
Attached patch comm-beta mozconfigs (obsolete) — Splinter Review
Attachment #600174 - Flags: review?(mbanner)
Attached patch comm-central mozconfigs (obsolete) — Splinter Review
Attachment #600175 - Flags: review?(mbanner)
Attached patch comm-release mozconfigs (obsolete) — Splinter Review
Attachment #600177 - Flags: review?(mbanner)
Comment on attachment 600175 [details] [diff] [review]
comm-central mozconfigs




>+mk_add_options CLIENT_PY_ARGS="--mozilla-repo=http://hg.mozilla.org/mozilla-central --skip-inspector --skip-venkman --skip-chatzilla --hg-options='--verbose --time' --hgtool=../tools/buildfarm/utils/hgtool.py --tinderbox-print"

Can we drop the mozilla-repo argument here and rely on the client.py defaults? This would make it much easier for each merge every 6 weeks, as we only need to update client.py.

Also, please ensure the order of arguments is consistent across the files.

>diff --git a/mail/config/mozconfigs/linux64/debug b/mail/config/mozconfigs/linux64/debug

>+mk_add_options ALWAYS_RUN_CLIENT_PY=1
>\ No newline at end of file

Please fix these (there's several of them).


>diff --git a/mail/config/mozconfigs/macosx-universal/unittest b/mail/config/mozconfigs/macosx-universal/unittest
>diff --git a/mail/config/mozconfigs/win32/unittest b/mail/config/mozconfigs/win32/unittest

I don't believe we need these files, if we do then I'd expect to have Linux versions as well.

I'd also expect release mozconfigs, but we can add those later if you want.
Attachment #600175 - Flags: review?(mbanner) → review-
Comment on attachment 600173 [details] [diff] [review]
comm-aurora mozconfigs

I think the same comments apply here, cancelling reviews until there's updated patches.
Attachment #600173 - Flags: review?(mbanner)
Attachment #600174 - Flags: review?(mbanner)
Attachment #600177 - Flags: review?(mbanner)
Mark: I'll get your review on the comm-central mozconfigs and do the others if/when these are approved.
Attachment #600175 - Attachment is obsolete: true
Attachment #604463 - Flags: review?(mbanner)
Comment on attachment 604463 [details] [diff] [review]
comm-central mozconfigs

Review of attachment 604463 [details] [diff] [review]:
-----------------------------------------------------------------

This is fine, with the minor nits fixed. We will need release mozconfigs to go on central as well, but I know you're still getting that part of automation running, so I'm happy for this to land with the nits fixed and the release mozconfigs part as follow-up.

::: mail/config/mozconfigs/linux64/debug
@@ +27,5 @@
> +CONCURRENCY=$(( `grep processor /proc/cpuinfo | wc -l` + 2 ))
> +mk_add_options MOZ_MAKE_FLAGS="-j$CONCURRENCY"
> +
> +# Run client.py
> +mk_add_options CLIENT_PY_ARGS="--hg-options='--verbose --time' --hgtool=../tools/buildfarm/utils/hgtool.py --skip-chatzilla --skip-chatzilla --skip-inspector --skip-venkman --tinderbox-print"

nit: You've got --skip-chatzilla in here twice.

::: mail/config/mozconfigs/linux64/nightly
@@ +28,5 @@
> +CONCURRENCY=$(( `grep processor /proc/cpuinfo | wc -l` + 2 ))
> +mk_add_options MOZ_MAKE_FLAGS="-j$CONCURRENCY"
> +
> +# Run client.py
> +mk_add_options CLIENT_PY_ARGS="--hg-options='--verbose --time' --hgtool=../tools/buildfarm/utils/hgtool.py --skip-chatzilla --skip-chatzilla --skip-comm --skip-inspector --skip-venkman --tinderbox-print"

nit: --skip-chatzilla twice

::: mail/config/mozconfigs/win32/debug
@@ +15,5 @@
> +export MOZ_DEBUG_SYMBOLS=1
> +
> +
> +# Run client.py
> +mk_add_options CLIENT_PY_ARGS="--hg-options='--verbose --time' --hgtool=../tools/buildfarm/utils/hgtool.py --skip-chatzilla --skip-chatzilla --skip-inspector --skip-venkman --tinderbox-print"

nit: --skip-chatzilla is here twice

::: mail/config/mozconfigs/win32/nightly
@@ +14,5 @@
> +# Needed to enable breakpad in application.ini
> +export MOZILLA_OFFICIAL=1
> +
> +# Run client.py
> +mk_add_options CLIENT_PY_ARGS="--hg-options='--verbose --time' --hgtool=../tools/buildfarm/utils/hgtool.py --skip-chatzilla --skip-chatzilla --skip-comm --skip-inspector --skip-venkman --tinderbox-print"

nit: --skip-chatzilla twice
Attachment #604463 - Flags: review?(mbanner) → review+
Committed comm-central mozconfigs in http://hg.mozilla.org/comm-central/rev/fe140018319f
Comment on attachment 604463 [details] [diff] [review]
comm-central mozconfigs

[Triage Comment]
a=me for landing this on comm-beta.
Attachment #604463 - Flags: approval-comm-beta+
Attachment #600177 - Attachment is obsolete: true
Attachment #600174 - Attachment is obsolete: true
Attachment #600173 - Attachment is obsolete: true
TODO: We still need release mozconfigs.
Release mozconfigs with s/release/${MOZ_UPDATE_CHANNEL}/ in the --enable-update-channel line (which I believe FF automation sets correctly), and also the necessary lines for client.py calls.
Attachment #615895 - Flags: review?(jhopkins)
Comment on attachment 615895 [details] [diff] [review]
release mozconfigs

Review of attachment 615895 [details] [diff] [review]:
-----------------------------------------------------------------

The only discrepancy I noticed were these lines missing from the linux32 mozconfig (compared to the current mozconfig in buildbot-configs):

< # XXX disable pedantic until we get MOZ_IPC or we resolve the build issues
< ac_add_options --disable-pedantic

Looks good - thanks, Mark!
Attachment #615895 - Flags: review?(jhopkins) → review+
(In reply to John Hopkins (:jhopkins) from comment #19)
> The only discrepancy I noticed were these lines missing from the linux32
> mozconfig (compared to the current mozconfig in buildbot-configs):
> 
> < # XXX disable pedantic until we get MOZ_IPC or we resolve the build issues
> < ac_add_options --disable-pedantic

Yes, that's expected, see the review request on bug 737028!
Component: Release Engineering → Build Config
Product: Mozilla Messaging → Thunderbird
QA Contact: release → build-config
Target Milestone: --- → Thunderbird 14.0
Version: other → Trunk
Checked in:

http://hg.mozilla.org/comm-central/rev/56ea15c0e0a2

As far as I know we don't need mozconfigs for l10n, although admittedly I'm confused by the ones Firefox has in-tree - the l10n base seem to all reference l10n-central regardless of what tree they are on.

If we do need those, then we can either reopen or file a new bug.
Status: ASSIGNED → RESOLVED
Closed: 8 years ago
Resolution: --- → FIXED
Comment on attachment 615895 [details] [diff] [review]
release mozconfigs

[Triage Comment]
Going to take these forward so that we've got them on aurora & beta hence ready for the switch.
Attachment #615895 - Flags: approval-comm-beta+
Attachment #615895 - Flags: approval-comm-aurora+
Comment on attachment 604463 [details] [diff] [review]
comm-central mozconfigs

[Triage Comment]
Taking this onto esr as well, as we'll need it there, and currently its NPOTB.
Attachment #604463 - Flags: approval-comm-esr10+
Attachment #615895 - Flags: approval-comm-esr10+
You need to log in before you can comment on or make changes to this bug.