582697 - Scrape num_ctors for linux builds

Status: RESOLVED FIXED

(Tree Management Graveyard :: TBPL, defect)

Description

[Chris AtLee [:catlee]]

We've just started reporting the number of global initializers for linux builds to Tinderbox.  Could TBPL grab this value and include it in the build information?

Comment 1

[Phil Ringnalda (:philor)]

Sigh. So much bustage everywhere you look.

You TinderboxPrint num_ctors: 338, and then three buildsteps later, you TinderboxPrint <a title="num_ctors" href='http://graphs.mozilla.org/graph.html#tests=[{"test":81,"branch":23,"machine":748}]'>num_ctors:338</a>. Please do that just once, not twice.

http://hg.mozilla.org/users/mstange_themasta.com/tinderboxpushlog/file/b646a0721da5/js/MachineResult.js#l60 is pretty much completely busted - it tries to stop us from double-showing the revision that we show off to the left, but the revision is TinderboxPrinted as "revision:", not "rev:", except on try.

It tries to stop us from showing the slave name, because we used to show it in the annotated summary. Now we don't show it there, and we should show it from the scrape, but we can't, because Try does TinderboxPrint: s: try-linux-slave13 and then two steps later does TinderboxPrint: s: 9528112e87253c5a8dc437209d067b03daafcbe6. Please stop doing that so we can stop hiding "s:".

It also is what hides num_ctors:, since that unanchored /s\:/ is "drop every scrape line with an s: anywhere in it on the ground."

And it's anybody's guess what /try\-/ is trying to hide, since I don't see anything (currently) on try like that.

Comment 2

[Phil Ringnalda (:philor)]

Oh, the revision is whichever wherever, rev:f2a2adaaacba0f0b1d470430c4fb8f6fb6125fbf or revision: f2a2adaaacba or rev:f2a2adaaacba within a single tree and within a single OS. Convenient.

Comment 3

[Phil Ringnalda (:philor)]

Attachment: don't drop s:

This gives us num_ctors:, and gives me my s: slavename for builds (I'll worry about tests and talos later, since I filed a bug for that), and hides the many revisions that are put in the scrape as revision: rather than rev:. I looked again, and still can't find any "try-.*:" - I'm betting at one point it was "try-revision:" or something like that.

The only downside is that if we deploy faster than releng reconfigs for bug 686831 (not very likely), on try we'll show

s: slavename
s: longcsetidnumber

until they do. Small chance, small cost.

Comment 4

[Phil Ringnalda (:philor)]

http://hg.mozilla.org/users/mstange_themasta.com/tinderboxpushlog/rev/e6b69bb53a1b looking fine on https://tbpl-dev.allizom.org/