974166 - Allow the possibility of spidermonkey builds on more platforms

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Steve Fink [:sfink] [:s:]]

I am looking at adding in Windows shell builds to keep them alive, and probably to satisfy bug 751700 as well.

Comment 1

[Steve Fink [:sfink] [:s:]]

Attachment: Allow the possibility of spidermonkey builds on more platforms

The first step is to just define the builds. I *think* this patch produces exactly zero new builds. It just creates several new spidermonkey builders. It does allow for a not-on-try-by-default optimized ARM simulator build, but I think that should be fine.

Then again, I could also be confusing myself. I guess I should run this on my master and see what builds it queues up.

Comment 2

[Christian Holler (:decoder)]

We should add a --disable-threadsafe build too, because that really breaks all the time and is very important for fuzzing.

Comment 3

[[e-mail] (:bhearsum)]

Comment on attachment 8377886 [details] [diff] [review]
Allow the possibility of spidermonkey builds on more platforms

Review of attachment 8377886 [details] [diff] [review]:
-----------------------------------------------------------------

> The first step is to just define the builds. I *think* this patch produces
> exactly zero new builds. It just creates several new spidermonkey builders.
> It does allow for a not-on-try-by-default optimized ARM simulator build, but
> I think that should be fine.

Looks that way to me after doing comparisons with builder_list.py (https://hg.mozilla.org/build/braindump/file/default/buildbot-related/builder_list.py):
(buildbot)➜  mozilla  diff -Naur ~/tmp/{before,after}.txt
(buildbot)➜  mozilla  
 

It's probably not now, but at some point there's going to have to be a larger conversation about how much we're willing to spend on spidermonkey builds. We've got a more strict budget than in the past, so we won't be able to add things quite as freely as before.

::: mozilla/config.py
@@ +1423,5 @@
> +            'linux-debug': ['arm-sim', 'warnaserrdebug'],
> +            'win32': ['generational', 'warnaserr'],
> +            'win32-debug': ['generational', 'warnaserrdebug'],
> +            'win64': ['generational', 'warnaserr'],
> +            'win64-debug': ['generational', 'warnaserrdebug'],

I'm not sure about adding 64-bit windows stuff. The Firefox 64-bit windows build is still basically unsupported, and completely hidden.

Comment 4

[Steve Fink [:sfink] [:s:]]

Comment on attachment 8377886 [details] [diff] [review]
Allow the possibility of spidermonkey builds on more platforms

Cancelling review because I'm seeing some extra builds getting scheduled. Several are builds I will end up wanting, but I didn't think they'd be there yet.

Comment 5

[Steve Fink [:sfink] [:s:]]

Attachment: Allow more flexible control of project try_by_default setting

Comment 6

[Steve Fink [:sfink] [:s:]]

Attachment: Allow the possibility of spidermonkey builds on more platforms

I required more granularity in try_by_default in order to prevent this from creating more try builds by default. This triggers the same set of jobs as without this patch, though it does add some new builders (selectable via trychooser.)

Comment 7

[[e-mail] (:bhearsum)]

Comment on attachment 8379822 [details] [diff] [review]
Allow more flexible control of project try_by_default setting

Review of attachment 8379822 [details] [diff] [review]:
-----------------------------------------------------------------

::: mozilla/config.py
@@ +1405,5 @@
>      'spidermonkey_try': {
>          'enable_try': True,
> +        'try_by_default': {
> +            'rootanalysis': True, # all platforms for which it is defined
> +            'generational': set(['linux64-debug']),

Are you sure this does something? I was expected to see a misc.py patch here...I don't see anything looking at try_by_default for anything other than True/False.

Comment 8

[[e-mail] (:bhearsum)]

Comment on attachment 8379823 [details] [diff] [review]
Allow the possibility of spidermonkey builds on more platforms

Review of attachment 8379823 [details] [diff] [review]:
-----------------------------------------------------------------

::: mozilla/config.py
@@ +1417,5 @@
> +            'linux64-debug':  ['rootanalysis', 'generational', 'exactrooting', 'warnaserrdebug'],
> +            'win32': ['generational', 'warnaserr'],
> +            'win32-debug': ['generational', 'warnaserrdebug'],
> +            'win64': ['generational', 'warnaserr'],
> +            'win64-debug': ['generational', 'warnaserrdebug'],

Did you see my earlier question about win64?:
(In reply to Ben Hearsum [:bhearsum] from comment #3)
> ::: mozilla/config.py
> @@ +1423,5 @@
> > +            'linux-debug': ['arm-sim', 'warnaserrdebug'],
> > +            'win32': ['generational', 'warnaserr'],
> > +            'win32-debug': ['generational', 'warnaserrdebug'],
> > +            'win64': ['generational', 'warnaserr'],
> > +            'win64-debug': ['generational', 'warnaserrdebug'],
> 
> I'm not sure about adding 64-bit windows stuff. The Firefox 64-bit windows
> build is still basically unsupported, and completely hidden.

Comment 9

[Steve Fink [:sfink] [:s:]]

Attachment: Make spidermonkey try_by_default into a dict

Doh! Sorry about the bbc piece. Here it is.

Comment 10

[Steve Fink [:sfink] [:s:]]

Attachment: Allow the possibility of spidermonkey builds on more platforms

Oh, and I forgot about win64 too.

I'd like to have win64 builds available (as in possible, but off unless you ask for them.) But when looking through the spidermonkey.sh script, it turns out I don't have a way to make them 64-bit yet. So right now it's best to leave them out.

Comment 11

[Steve Fink [:sfink] [:s:]]

http://hg.mozilla.org/build/buildbot-configs/rev/107e4e3a07d9
http://hg.mozilla.org/build/buildbot-configs/rev/b89a04d4d878

Comment 12

[John Hopkins (:jhopkins)]

in production.

Comment 13

[Ed Morley [:emorley]]

13:35:59 - Ms2ger: Tomcat|sheriffduty, edmorley: the failing windows SM runs on https://tbpl.mozilla.org/?tree=Try&rev=34b1cf48d26e don't seem to run on inbound? Ok to hide?
14:02:33 - edmorley: Ms2ger: yup
14:03:30 - edmorley: Ms2ger: history on them says they are new as of 24th
14:04:41 - edmorley: Ms2ger: bug 974166

Comment 14

[Steve Fink [:sfink] [:s:]]

Yes. I will be fixing them in bug 975011. Except I'll probably disable them entirely at the same time. Or rather, I will replace them with a single warnings-as-warnings shell build of some flavor.

If I were a good boy, I'd disable them right now.