798798 - Turn off codesighs everywhere

Status: RESOLVED DUPLICATE of bug 803736

(Release Engineering :: General, defect)

Description

[Phil Ringnalda (:philor)]

* The number is meaningless, and was replaced by a meaningful one in bug 659950 a year and a half ago (though it's still not deployed)

* On Mac, it adds 5 minutes to the build time; on mock, it added 30 minutes, or a 20 minute timeout and pointless red, except

* On non-try, it's been 50% broken since last January, by fetching the old log and putting it in one filename, then telling codesighs that it's in a different filename

* On non-PGO Linux64, it reports the total of code+data as being zero bytes, so it's rotting, but because it's meaningless even when it's non-zero, there is absolutely no chance of anyone fixing it

We're all used to having Granny sit in her rocker in the corner, but she's been dead for several years now, and the corpse is starting to stink - it's time to bury her.

Comment 1

[Phil Ringnalda (:philor)]

Attachment: fix v.1

I'm sort of surprised to find it here, but despite the many wonderful varieties of places where we turn it off, this seems to be the single place where we turn it on.

If you're looking for the signature on the death warrant and the call to the governor's office which is answered with "no, we're not commuting the sentence," they are in bug 798243 comment 1; if you're looking for the poor sucker who is going to have to remove those disabling bits below, and my disabling on try bits, and the disabling of b2g a different way from the bit below and the disabling of android and every other use of "codesighs" anywhere, that'd be the assignee of bug 668219.

Comment 2

[[e-mail] (:bhearsum)]

Comment on attachment 668882 [details] [diff] [review]
fix v.1

Review of attachment 668882 [details] [diff] [review]:
-----------------------------------------------------------------

I'm pretty sure this patch on its own will cause syntax problems due to missing variables. Also, we should just rip out support completely rather than disabling. That involves misc.py, process/factory.py, steps/test.py, steps/mock.py (wtf?), and process/release.py. Are up for doing that?

Comment 3

[Phil Ringnalda (:philor)]

Oh, did I misunderstand how .get works? I thought that line was getting the value for enable_codesighs (which we only ever set to False), and if the value didn't exist, setting it to True, and that my patch would instead set it to False.

Comment 4

[[e-mail] (:bhearsum)]

(In reply to Phil Ringnalda (:philor) from comment #3)
> Oh, did I misunderstand how .get works? I thought that line was getting the
> value for enable_codesighs (which we only ever set to False), and if the
> value didn't exist, setting it to True, and that my patch would instead set
> it to False.

Looks like I just read it wrong. In any case, I'd like this to be ripped out completely.

Comment 5

[Phil Ringnalda (:philor)]

I was hoping bug 668219 would rip it out, since that's even in its summary, but perhaps since that took a full year just to decide on the name in bug 668216, it can't be trusted to do so.