Windows log reads: cp: cannot create regular file `c:/builds/moz2_slave/tb-c-cen-w64-00000000000000000/build/objdir-tb/_tests/testing/mochitest/browser/browser/base/content/test/appUpdate/updater.exe': No such file or directory Linux log reads: cp: cannot create regular file `/builds/slave/tb-c-cen-l64-00000000000000000/build/objdir-tb/_tests/testing/mochitest/browser/browser/base/content/test/appUpdate/updater': No such file or directory Mac reads: rsync: mkdir "/builds/slave/tb-c-cen-m64-00000000000000000/build/objdir-tb/_tests/testing/mochitest/browser/browser/base/content/test/appUpdate" failed: No such file or directory (2)
This comes from bug 893505: https://hg.mozilla.org/mozilla-central/rev/898a2071de2d which is a massive change.
Summary: All builds busted on all platforms as of 2017-04-09: Missing appUpdate/updater → Port bug 893505 to C-C - All builds busted on all platforms as of 2017-04-09: Missing appUpdate/updater
It seems we need to work here: https://hg.mozilla.org/mozilla-central/rev/898a2071de2d#l50.15, 50.51 and 50.57. Could a ifdef work here?
>> Could a ifdef work here? Just looking at it I would say yes.
But how exactly? I've never worked in tests.
Hmm standard makefile syntax. I would say either checking that MOZ_APP_NAME is not in thunderbird, instantbird or seamonkey or check if the directory for the rsync/copy exists. There should be no fallout in other branches because this just copies the updater for tests to a specific target location. The browser tests are never run for c-c apps.
Maybe like this. I'm not even sure what the language is there. It's not Python since that has if: conditional stuff and no endif. I imitated what was there, including the tabs.
Here a try including the M-C changeset: https://treeherder.mozilla.org/#/jobs?repo=try-comm-central&revision=28960a37e0c6ec35d9ad844f72241b871f53d0ef (In reply to Frank-Rainer Grahl from comment #5) > Hmm standard makefile syntax. > I would say either checking that MOZ_APP_NAME is not in thunderbird, > instantbird or seamonkey or check if the directory for the rsync/copy exists. Hmm, didn't see this comment, so can you please suggest something. ifneq (thunderbird,$(MOZ_APP_NAME)) How is the "not in"?
Trying the second version: https://treeherder.mozilla.org/#/jobs?repo=try-comm-central&revision=e7b074054fe16bb2e06c7daf409c294383942a2f Feel free to cancel those runs if you have a better idea.
All good things come in threes :) MOZ_INSTANTBIRD is not defined right now in "im" but that could be done later. Essentially checks if one of the 3 vars is defined (check for non space) and does only set the dir if not.
Do IB and SM do tests? IB I'm sure they don't because they rely on TB's tests. So no need for at least IB.
>> Do IB and SM do tests? SM does. Mostly broken because no one has time to fix them but they are run. My local build failed. Need to check the syntax.
Don't ask :) Works for me locally. Need to check Firefox.
Attachment #8856229 - Attachment is obsolete: true
(In reply to Frank-Rainer Grahl from comment #12) > Works for me locally. Need to check Firefox. Great, since my two try pushes failed, here goes the FRG version: https://treeherder.mozilla.org/#/jobs?repo=try-comm-central&revision=a1d37b5d035aa13a471b17f6acedc6397c76a44c
Also started a FX try: https://hg.mozilla.org/try/rev/e295d6fd2e7d6c3361a12d04389b08c9e3e26684
> Also started a FX try: My local one is still compiling. Full build with tests is slowwww...
Local builds are well with Firefox and SeaMonkey so TB shouldn't be any different.
Attachment #8856224 - Attachment is obsolete: true
Attachment #8856225 - Attachment is obsolete: true
Comment on attachment 8856236 [details] [diff] [review] 1354866-build-bustage-mochitest-update-smtb.patch Sorry about the review SPAM. One review will do, so I'm putting authors and reviewers from bug 893505. Thunderbird is completely busted and we need to fix this quickly.
(In reply to Jorg K (GMT+2) from comment #13) > Great, since my two try pushes failed, here goes the FRG version: > https://treeherder.mozilla.org/#/jobs?repo=try-comm- > central&revision=a1d37b5d035aa13a471b17f6acedc6397c76a44c That came out green.
Comment on attachment 8856236 [details] [diff] [review] 1354866-build-bustage-mochitest-update-smtb.patch Can't this just use MOZ_BUILD_APP to check whether browser/ is actually shipped? I'm assuming that fennec doesn't run any of this. More generally, I'm confused that the test dir here is in browser/ and this is in a toolkit/ makefile.
> Can't this just use MOZ_BUILD_APP to check whether browser/ is actually shipped? Would need to be tested for 'browser' and you must be sure that its really only needed there. An "ifdef MOZ_PHOENIX" would probably be more simple then. This seems to be the corresponding define for Firefox. > More generally, I'm confused that the test dir here is in browser/ and this is in a toolkit/ makefile. I agree.
So how can we solve this? We can't hold Thunderbird development hostage with four lines in a makefile somewhere. M-C is merging about 300 changesets a day and we must *absolutely* be able to build to stand the chance to keep up with those changes. So if the original change wasn't correct due to a mix of browser/ and toolkit/, let's back it out and land it properly. We're accepting responsibility for TB going bust due to our technical debt when we're not keeping up with, for example, replacing deprecated interfaces, but this one here is nothing we can or should be fixing.
Comment on attachment 8856236 [details] [diff] [review] 1354866-build-bustage-mochitest-update-smtb.patch This should be fine to get Thunderbird building again. If it needs changes it can be fixed later. Adding r=chmanchester since he is the build peer that reviewed bug 893505
Comment on attachment 8856236 [details] [diff] [review] 1354866-build-bustage-mochitest-update-smtb.patch Review of attachment 8856236 [details] [diff] [review]: ----------------------------------------------------------------- r=me with one nit addressed. ::: toolkit/mozapps/update/updater/updater-xpcshell/Makefile.in @@ +7,5 @@ > > XPCSHELLTESTDIR = $(topobjdir)/_tests/xpcshell/toolkit/mozapps/update/tests > MOCHITESTCHROMEDIR = $(topobjdir)/_tests/testing/mochitest/chrome/toolkit/mozapps/update/tests > + > +ifeq ($(MOZ_SUITE)$(MOZ_THUNDERBIRD),) The prevailing style in our Makefiles seems to be to put the empty string first, so starting this line with `ifeq (,$(...` reads a little easier to me.
Attachment #8856236 - Flags: review?(cmanchester) → review+
Attachment #8856669 - Attachment description: 1354866-build-bustage-mochitest-update-smtb.patch - not fixed → 1354866-build-bustage-mochitest-update-smtb.patch - nit fixed
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/mozilla-inbound/rev/1ba02ff5671b Skip updating mochitest/browser for TB and SM. r=rstrong,chmanchester
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → FIXED
Target Milestone: --- → Thunderbird 55.0
You need to log in before you can comment on or make changes to this bug.