1163573 - Process Type section in Signature Summary is empty

Status: RESOLVED WORKSFORME

(Socorro :: General, task)

Description

[Robert Kaiser]

It looks like something has broken the Process Type section in the Signature Summary, it's always empty now.

See for example this one, where it should say 100% plugin:
https://crash-stats.mozilla.com/report/list?signature=hang%20%7C%20CleanupPerAppKey%20%7C%20MsgWaitForMultipleObjectsEx%20%7C%20MsgWaitForMultipleObjects%20%7C%20mozilla%3A%3Aipc%3A%3AMessageChannel%3A%3AWaitForInterruptNotify%28%29%20%7C%20mozilla%3A%3Aipc%3A%3AMessageChannel%3A%3ACall%28IPC%3A%3AMessage%2A%2C%20IPC%3A%3AMessage%2A%29

Comment 1

[David Teller [:Yoric] - still alive but not very active]

This is now indirectly a blocker for e10s.

Comment 2

[David Teller [:Yoric] - still alive but not very active]

STR:
* Fetch https://crash-stats.mozilla.com/api/SuperSearch/?async_shutdown_timeout=%21__null__&_results_number=10&_columns=process_type
* Expect a column "process type"

(Same query entered through the WebApp, has an empty column for process type).

Comment 3

[[DEACTIVATED] Adrian Gaudebert]

KaiRo, when the "process_type" is null, that means it's a "browser" crash, right? If that's the case, then SuperSearch is behaving correctly, it just doesn't have the right data. I suppose the fix for that would be to have yet another processor rule to transform empty process_type to "browser".

Comment 4

[David Teller [:Yoric] - still alive but not very active]

So how can we find out whether a crash is in a parent process/child process/single process?

Comment 5

[[DEACTIVATED] Adrian Gaudebert]

What would you normally use? Is that what `process_type` is for? If so, the possible values of that field that I am aware of are "browser", "plugin" and "content". Not sure what corresponds to what in your list of processes. 

Anyway, if I am not mistaking (and I'm quite sure I am not), an empty value in Super Search means that it is actually a "browser" process_type. So if you're not seeing anything in the process_type column, then that means everything is a "browser" process type.

Comment 6

[David Teller [:Yoric] - still alive but not very active]

(In reply to Adrian Gaudebert [:adrian] from comment #5)
> What would you normally use? Is that what `process_type` is for? If so, the
> possible values of that field that I am aware of are "browser", "plugin" and
> "content". Not sure what corresponds to what in your list of processes.

Well, "content" == "child process", so that's one good :)
In single process mode, "browser" should be "single process".
In multi-process mode, "browser" should be "parent process".

So if there is an information that can tell me whether the user is running in multi-process mode, I'll have everything I need.

> Anyway, if I am not mistaking (and I'm quite sure I am not), an empty value
> in Super Search means that it is actually a "browser" process_type. So if
> you're not seeing anything in the process_type column, then that means
> everything is a "browser" process type.

Ok, good to know.

Comment 7

[Robert Kaiser]

(In reply to Adrian Gaudebert [:adrian] from comment #3)
> KaiRo, when the "process_type" is null, that means it's a "browser" crash,
> right?

Yes, but this bug is about the old signature summary, not about Super Search, right?

Yoric, the process type for the browser process should always be empty (in the raw data) or something like "browser" or "main" (in beautified data). The way to determine if e10s is on is so look for DOMIPCEnabled=1 in raw data (or "dom_ipc_enabled" "exists" in terms of SuperSearch).

Socorro should not overload a field that has clear meaning in the raw data. It's the type of process that has crashed, which is (in current times) either the main process (currently called "browser" but in the light of FxOS, "main" might be a better name), a content process, or a plugin process.

Comment 8

[[DEACTIVATED] Adrian Gaudebert]

Yeah, sorry KaiRo, we hijacked this bug. :)

So Yoric I think your problem is solved by what KaiRo said, right? The `dom_ipc_enabled` filter should be what you are looking for. 

If we come back to the problem of this bug as originally filed, I do not know what's wrong with the summary in report/list/, but the new version in signature report seems to work fine, see for example: https://crash-stats.mozilla.com/signature/?product=Firefox&signature=F1398665248_____________________________

As such, I am willing to not try to fix this problem and just let it die with the old signature summary.

Comment 9

[David Teller [:Yoric] - still alive but not very active]

Sorry about the hijacking, I misunderstood the bug. As far as I am concerned, all is now well :)

Comment 10

[Will Kahn-Greene [:willkg] ET needinfo? me]

I can't find an instance of the problem described in comment #0. Maybe in the last 5 years, this was fixed somehow.

Marking as WORKSFORME. If someone has steps to reproduce, we can re-open.