Bug 84196 – run-webkit-tests picked up an old crash log

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Simon Fraser (smfr)]

http://build.webkit.org/results/Lion%20Debug%20(WebKit2%20Tests)/r114380%20(6092)/fast/loader/form-state-restore-with-locked-back-forward-list-crash-log.txt is a crash log from:

Date/Time:       2012-04-02 03:32:12.508 -0700

yet the tests ran on 4/17. I guess the bot has accumulated a lot of crash logs, and the scripts picked up an old one because the PIDs have wrapped around.

The scripts should probably delete old crash logs from the machines at some point.

[Dirk Pranke]

Good point. It should be easy enough to change the code to notice when run-webkit-tests starts and not pick up any crash logs older than that start time.

[Simon Fraser (smfr)]

Another one:
http://build.webkit.org/results/Lion%20Debug%20(WebKit2%20Tests)/r114521%20(6121)/fast/dom/htmlcollection-detectability-crash-log.txt

Date/Time:       2012-04-11 00:07:16.193 -0700

[Dirk Pranke]

Created attachment 137778 [details]
Patch

[Dirk Pranke]

I think this patch should fix things, although I haven't tested this against a real-world instance.

[Ryosuke Niwa]

Comment on attachment 137778 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=137778&action=review

> Tools/Scripts/webkitpy/common/system/crashlogs.py:38
> +    def find_newest_log(self, process_name, pid=None, include_errors=False, newer_than=None):

I'm not sure if newer_than makes sense as the variable name. I think start_time would be semantically clearer.

[Dirk Pranke]

(In reply to comment #5)
> (From update of attachment 137778 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=137778&action=review
> 
> > Tools/Scripts/webkitpy/common/system/crashlogs.py:38
> > +    def find_newest_log(self, process_name, pid=None, include_errors=False, newer_than=None):
> 
> I'm not sure if newer_than makes sense as the variable name. I think start_time would be semantically clearer.

Interesting, newer_than seems clear to me (show me crash logs newer than X), but start_time not at all (logs don't have a start time) ... why do you think the opposite? Are you thinking that you're looking for processes started after X? The process doesn't have to have started after X, though, just crashed after X.

Maybe "created_after" would be better, or "crashed_after", or just "after"?

Note that there is a minor correctness issue in that you actually want ctime > X, not mtime > X, but there is no filesystem.ctime() routine. I could add one, but it seemed unnecessary since ctime will presumably equal mtime nearly all of the time ...

[Ryosuke Niwa]

Comment on attachment 137778 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=137778&action=review

>>> Tools/Scripts/webkitpy/common/system/crashlogs.py:38
>>> +    def find_newest_log(self, process_name, pid=None, include_errors=False, newer_than=None):
>> 
>> I'm not sure if newer_than makes sense as the variable name. I think start_time would be semantically clearer.
> 
> Interesting, newer_than seems clear to me (show me crash logs newer than X), but start_time not at all (logs don't have a start time) ... why do you think the opposite? Are you thinking that you're looking for processes started after X? The process doesn't have to have started after X, though, just crashed after X.
> 
> Maybe "created_after" would be better, or "crashed_after", or just "after"?
> 
> Note that there is a minor correctness issue in that you actually want ctime > X, not mtime > X, but there is no filesystem.ctime() routine. I could add one, but it seemed unnecessary since ctime will presumably equal mtime nearly all of the time ...

Ah, I see what you're saying. Okay.

[Dirk Pranke]

Committed r114577: <http://trac.webkit.org/changeset/114577>

[Dirk Pranke]

*** Bug 85267 has been marked as a duplicate of this bug. ***