911248 - Clean up TBPL's job types and some other UI fixups

Status: RESOLVED FIXED

(Tree Management Graveyard :: TBPL, defect)

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Attachment: Clean up the OSX job names

OSX 10.5 is obsolete and the OSX 10.6 job names are a bit simpler now. This also moves macosx64 to the 10.7 line so that desktop B2G builds show up with the other builds.

Comment 2

[Ryan VanderMeulen [:RyanVM]]

Attachment: Remove the distinction between Fedora and Ubuntu Linux jobs

Tell me this doesn't look better.

We aren't duplicating jobs between Fedora/Ubuntu at this point, so let's treat Linux jobs as Linux regardless of which slaves they run on.

Comment 3

[Ryan VanderMeulen [:RyanVM]]

Attachment: Remove obsolete Talos job types

Comment 4

[Ryan VanderMeulen [:RyanVM]]

Attachment: Remove obsolete B2G beta channel builds

I confirmed with fabrice that these builds are long-obsolete. We don't even run these builds on the (ever-running) v1.0.1 tree.

Comment 5

[Ryan VanderMeulen [:RyanVM]]

Attachment: Split the Marionette unit and webapi tests into seperate symbols

This has annoyed me for awhile. Has jgriffin's blessing.

Comment 6

[Ryan VanderMeulen [:RyanVM]]

Attachment: Give Android l10n nightlies a unique symbol

I went with NaL to fit the convention we're using for the B2G builds, but I'm open to other suggestions.

Comment 7

[Ryan VanderMeulen [:RyanVM]]

Would be nice if the l10n nightlies were displayed like mochitests, i.e. NaL(1 2 3 4 5). Would save a lot of space.

Comment 8

[Ryan VanderMeulen [:RyanVM]]

Attachment: Give Android l10n nightlies a unique symbol and group them

This groups them. Looks like UserInterface.js special cases mochitest? Guess we'd have to extend that.

Comment 9

[Ed Morley [:emorley]]

Comment on attachment 797926 [details] [diff] [review]
Remove obsolete Talos job types

Review of attachment 797926 [details] [diff] [review]:
-----------------------------------------------------------------

r=me presuming flicking around both trunk and older trees (eg esr17, release, b2g18) doesn't show any generic 'T' jobs :-)

Comment 10

[Ed Morley [:emorley]]

(In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #8)
> This groups them. Looks like UserInterface.js special cases mochitest? Guess
> we'd have to extend that.

That's done here:
https://hg.mozilla.org/webtools/tbpl/file/16cf91ab5d57/js/Data.js#l725

Guess we could make the regexp more generic (ie "anything-[0-9]+"), as long as it didn't catch things like "trobocheck2" etc.

Comment 11

[Ed Morley [:emorley]]

Comment on attachment 797953 [details] [diff] [review]
Give Android l10n nightlies a unique symbol and group them

Review of attachment 797953 [details] [diff] [review]:
-----------------------------------------------------------------

I think I'd prefer "Nl" or "NL" rather than "NaL", since the only reason the B2G builds are "BgL" is since they are b2g gecko builds and not normal full product builds. 

Also, for Asan, We have "An" (ie type then nightly), which is yet another convention... hmm.

I'm wondering if we should make this more generic too, eg:

L10n(N1 N2 N3...)

Since (a) We may end up with these on other platforms too, (b) Android is already stated on the row name, so seems redundant to state it again.

Comment 12

[Ryan VanderMeulen [:RyanVM]]

It also occurs to me that the Windows regexes in Data__getMachine may need some cleanup. I didn't look closely at those earlier.

Comment 13

[Ryan VanderMeulen [:RyanVM]]

Well, the ASAN-specific names are going away soon, so we don't need to worry about that one :)

I'm fine with L10n(N1 ...). IIRC, we do eventually want L10n jobs showing on tbpl, so that does seem better in the long-run.

Comment 14

[Ryan VanderMeulen [:RyanVM]]

(In reply to Ed Morley [:edmorley UTC+1] from comment #9)
> r=me presuming flicking around both trunk and older trees (eg esr17,
> release, b2g18) doesn't show any generic 'T' jobs :-)

You presume correctly :). I removed a good one first to make sure I knew what I was looking for, then proceeded to remove them one at a time, reloading and checking after each.

Comment 15

[Ryan VanderMeulen [:RyanVM]]

(In reply to Ed Morley [:edmorley UTC+1] from comment #11)
> Also, for Asan, We have "An" (ie type then nightly), which is yet another
> convention... hmm.

While (most of) the ASAN special casing will be going away soon, we also have Static Checking builds that have an Sn. I can live with it for now, though I'll admit that I'm already contemplating ideas for grouping builds together.

(In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #12)
> It also occurs to me that the Windows regexes in Data__getMachine may need
> some cleanup. I didn't look closely at those earlier.

Nope, stuck with the older rev3 slaves on the older branches still. Oh well.

Comment 16

[Ryan VanderMeulen [:RyanVM]]

Attachment: Give L10n repacks their own job type

This implements what we discussed. I re-used the "N" symbol since it's group inside L10n(). I debated using "L" instead, so it would be L10n(L1 L2 L3 ...), but I think this works for now. If we start adding other L10n job types, we may want to reconsider then.

Comment 17

[Ryan VanderMeulen [:RyanVM]]

Attachment: Rename "Win" to "Windows" in the OS names list

There are already other jobs with longer names. AFAICT, this doesn't cause any text cutoff problems and is more consistent with how we display other OSes.

Comment 18

[Ryan VanderMeulen [:RyanVM]]

Attachment: Capitalize the job flavors

This was annoying me while adding the ASAN stuff because of how out of place it looked. Additionally, having them all lower-case is inconsistent with the OS names are displayed in the job list.

Comment 19

[Ryan VanderMeulen [:RyanVM]]

Attachment: De-parenthesize VM OS names

Shortens them up a bit and personally, I don't think it hurts anything from a "what job is that?" standpoint. This also makes Win2012 jobs not get cut off when looking at the Date branch.

Comment 20

[Ryan VanderMeulen [:RyanVM]]

Attachment: Move b2g18 v1.0.1 under v1.1hd

b2g18 and b2g18_v110hd logically belong together. I'll save the "1.0.1 shouldn't even be on TBPL anymore" argument for another bug.

Comment 21

[Ed Morley [:emorley]]

Comment on attachment 798114 [details] [diff] [review]
Give L10n repacks their own job type

Review of attachment 798114 [details] [diff] [review]:
-----------------------------------------------------------------

r=me with these changes :-)

::: js/Config.js
@@ +267,5 @@
>                            "Talos other",
>                            "Talos dromaeojs",
> +                          "Talos Performance"],
> +    "L10n Repack": ["L10n Nightly",
> +                    "L10n Repack"]

This just needs to be: ["L10n Nightly"]
(ie no "L10n Repack" inside the group)

@@ +323,5 @@
>      "B2G Gecko Nightly": "Ng",
>      "B2G Gecko Localizer Nightly": "NgL",
>      "B2G Device Image Nightly": "Ni",
>      "B2G Marionette-Enabled Device Image Nightly": "NiM",
> +    "L10n Repack": "L10n",

Perhaps pop this with the other line, just above it.

::: js/Data.js
@@ +552,5 @@
>          /static analysis/i.test(name) ? "Static Checking Build" :
>          /valgrind/i.test(name) ? "Valgrind Build" :
>          /dxr/i.test(name) ? "DXR Index Build" :
>          /build$/i.test(name) ? "Build" :
> +        /l10n/i.test(name) ? "L10n Repack" :

This line isn't needed :-)

Comment 22

[Ryan VanderMeulen [:RyanVM]]

I was thinking that the "L10n Repack" lines would serve as a generic "l10n" type for any eventual additions along the same line as a 'T' is now.

Comment 23

[Ed Morley [:emorley]]

Comment on attachment 798210 [details] [diff] [review]
Rename "Win" to "Windows" in the OS names list

Review of attachment 798210 [details] [diff] [review]:
-----------------------------------------------------------------

Is going to take me a while to get used to (think the short 'Win' lines currently help my eye calculate where in the platform list I'm glancing), but makes sense, so lets try it.

Comment 24

[Ed Morley [:emorley]]

Comment on attachment 798212 [details] [diff] [review]
Capitalize the job flavors

Review of attachment 798212 [details] [diff] [review]:
-----------------------------------------------------------------

Again looks weird after so long lowercase, but makes sense :-)

Comment 25

[Ed Morley [:emorley]]

One more possible consistency change... we have:
Linux32
Linux64
Windows 7 x64 Opt

I think we should:
Linux32 -> Linux (or Linux x86; but think it would be easier to visually scan without the x86)
Linux64 -> Linux x64

So that the spacing is consistent.

Comment 26

[Ryan VanderMeulen [:RyanVM]]

Funny, I was thinking about doing that too. Why not. Also, ping on comment 22 (in case it got lost in the r+ shuffle).

Comment 27

[Ed Morley [:emorley]]

Re comment 22, shall we wait until any other l10n jobs crop up? Since fairly easy to add (and deploy) changes, don't think we gain much (and if we start pre-emptively adding many new potential job types, we'll start bloating the regexp :-)).

Comment 28

[Ryan VanderMeulen [:RyanVM]]

Works for me. Pushed with nits addressed.

https://hg.mozilla.org/webtools/tbpl/rev/d32bad9bd5ca
https://hg.mozilla.org/webtools/tbpl/rev/1034568e267a
https://hg.mozilla.org/webtools/tbpl/rev/eb089117c783
https://hg.mozilla.org/webtools/tbpl/rev/3bf9a12b0d90
https://hg.mozilla.org/webtools/tbpl/rev/37a327715035
https://hg.mozilla.org/webtools/tbpl/rev/7b4591864764
https://hg.mozilla.org/webtools/tbpl/rev/bba4f0851ee2
https://hg.mozilla.org/webtools/tbpl/rev/b50ede895cc5
https://hg.mozilla.org/webtools/tbpl/rev/ac8f27167c2a
https://hg.mozilla.org/webtools/tbpl/rev/2f6f28e799e2

Comment 29

[Ed Morley [:emorley]]

In production :-)