It seems like a rather specific requirement. Also, I don't know how well
adding requirements to tier 1 jobs has worked in practice for forcing
change. There are requirements around the allowable amount of
intermittent orange, too, aren't there? And I can't say I've noticed
people jumping up to fix intermittents as a result, though the threat of
de-listing has provided _some_ level of "encouragement". I guess the
sheriffs would know.
You're talking C++ stacks, I assume. What about the various layers of
harnesses? What about NS_ASSERTIONs (or whatever those noisy things
are)? Are stacks always the most critical bit of information, or would
it be better to require full output for failing tests? Or logged input
for tests where that makes sense?
I'd be ok with requiring stacks, whatever exactly that means, but I'd
rather have a good place to go to see the current state of all the
different jobs in terms of what debugging features they provide, how to
run them, and a list of bugs for everything that's missing.
(Perhaps I'm a little biased, in that I maintain a set of builds that do
not have stacks currently. Then again, I just went to a bunch of work to
add them, which is waiting on review, so I agree they are important and
useful.)
On 07/10/2015 10:06 AM, Andrew McCreight wrote:
Are we going to have tests for this? Does working include being properly
symbolicated?
But yes, it is extremely annoying to have failures without stacks. (See bug
1165469, which was only figured out once somebody reproduced the test
locally, because there was no stack.)
On Fri, Jul 10, 2015 at 9:58 AM, Kyle Huey <m...@kylehuey.com> wrote:
Any reason not to require this?
- Kyle
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform