Gecko Profiler shouldn't include data-urls in profile strings (for perf reasons)
Categories
(Core :: JavaScript Engine, task, P1)
Tracking
()
People
(Reporter: tcampbell, Assigned: tcampbell)
References
(Blocks 1 open bug)
Details
Attachments
(1 file)
When profiling Bug 1576999, we spend an enormous amount of time in the profiler itself cloning strings for data-url filenames.
We can either cap filename lengths here or replace all filenames that start with 'data:' with a placeholder.
| Assignee | ||
Comment 1•7 years ago
|
||
When profiling scripts with large numbers of dynamic Functions and
complex URLs the profiler overhead can severely skew results. This puts
a cap on the filename length we send to profiler to reduce the impact of
this when using data urls.
Comment 3•7 years ago
|
||
| bugherder | ||
Comment 4•7 years ago
|
||
Backed out changeset 5e767d6cba5a (Bug 1577161) for causing Bug 1579038.
Backout link: https://hg.mozilla.org/integration/autoland/rev/dca9b6dd135f3230cb7f5516003df450efcb06bf
Comment 5•7 years ago
|
||
Backout merged: https://hg.mozilla.org/mozilla-central/rev/dca9b6dd135f
| Assignee | ||
Comment 6•7 years ago
|
||
The issue is that char_traits::find(..) requires length to cover valid memory rather than stopping at first match. On Windows, the SIMD implementation will read beyond null-pointer and touch an unmapped page.
Comment 7•7 years ago
|
||
Maybe you can use js_strchr_limit?
Updated•7 years ago
|
Comment 8•6 years ago
|
||
There's a r+ patch which didn't land and no activity in this bug for 2 weeks.
:tcampbell, could you have a look please?
For more information, please visit auto_nag documentation.
Comment 10•6 years ago
|
||
| bugherder | ||
| Assignee | ||
Updated•6 years ago
|
Comment 11•6 years ago
|
||
| bugherder landing | ||
Updated•6 years ago
|
Description
•