Bug 96342
| Summary: | "FAIL: Timed out waiting for notifyDone to be called" should be TIMEOUT failures, not TEXT | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Dirk Pranke <dpranke> |
| Component: | Tools / Tests | Assignee: | Nobody <webkit-unassigned> |
| Status: | NEW | ||
| Severity: | Normal | CC: | adamk, jberlin, simon.fraser, slewis, thorton |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
Dirk Pranke
It seems to me that if a test fails with the above message, we should be treating this as a TIMEOUT failure (and possibly killing WTR to reset the state as much as possible) rather than a TEXT failure; I can understand the existing implementation since ORWT didn't have a concept of a TIMEOUT failure, but perhaps we should change things in NRWT to check for this message and treat it as a TIMEOUT?
Any objections?
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Stephanie Lewis
Logically this makes sense.
My only concern is that a timeout currently usually means the process hung, whereas these bugs usually mean the test didn't complete for some other reason. Thus a current timeout is much more serious bug than these failures are.
Tim Horton
I think this makes sense, though I'm not sure it needs to restart the child...
Dirk Pranke
(In reply to comment #1)
> Logically this makes sense.
>
> My only concern is that a timeout currently usually means the process hung, whereas these bugs usually mean the test didn't complete for some other reason. Thus a current timeout is much more serious bug than these failures are.
Historically, in chromium (when we ran with TestShell instead of DRT) the timeout was enforced by TestShell itself, and things basically ran until the watchdog fired, much like with this error.
The only reason we assume the process is hung now is because we can't distinguish between a hung process and the case where we know DRT timed itself out. We could consider this message to be a timeout, and treat a hung process more like a crash (and only restart in the latter case).