Bug 63845
| Summary: | NRWT appears to run slower on the bots (when run with one child process) | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | WebKit Review Bot <webkit.review.bot> |
| Component: | New Bugs | Assignee: | Adam Barth <abarth> |
| Status: | RESOLVED FIXED | ||
| Severity: | Normal | CC: | abarth, aroben |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Bug Depends on: | |||
| Bug Blocks: | 34984 | ||
WebKit Review Bot
NRWT appears to run slower on the bots (when run with one child process)
Requested by abarth on #webkit.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Adam Barth
For example:
http://build.webkit.org/builders/SnowLeopard%20Intel%20Release%20%28Tests%29/builds/31015
http://build.webkit.org/builders/SnowLeopard%20Intel%20Release%20%28Tests%29/builds/31017
Adam Barth
As sizable fraction of this turns out to be related to checking for ReportCrash all the time. With that code disabled, we get down to 20 mins. ORWT looks to have been about at 18ish minutes.
Feel free to re-open this bug if you'd to get those last two minutes back. There's going to be a big win from turning on parallelism, so we're probably better off optimizing in that configuration.
Adam Roben (:aroben)
But we can't just leave the ReportCrash code disabled. That would reintroduce bug 55907!
Adam Roben (:aroben)
Adam explained to me that he fixed the slowness in a way that made things work just fine for the single process case, but not for the multi-process case. The latter is now covered by bug 63949.