Revert column number dimension to "code units", not "code points", for FF68 Beta/Release
Categories
(Core :: JavaScript Engine, task, P1)
Tracking
()
Tracking | Status | |
---|---|---|
firefox-esr60 | --- | unaffected |
firefox66 | --- | unaffected |
firefox67 | --- | unaffected |
firefox68 | + | fixed |
People
(Reporter: tcampbell, Assigned: Waldo)
References
(Regression)
Details
(Keywords: regression)
Attachments
(1 file)
Due to a few lingering perf issues, we should disable the UTF-8 column fix before FF68 merges.
One option might be to revert the following two patches:
https://hg.mozilla.org/mozilla-central/rev/8eaa269b362f
https://hg.mozilla.org/mozilla-central/rev/1664505ad313
This is the ifdef that switches to the slow mode, and reverts a conversion to using UTF-8 so that is isn't plagued by colno errors.
Jeff may have some other ideas on what the best approach to achieve this is.
Updated•5 years ago
|
Assignee | ||
Updated•5 years ago
|
Assignee | ||
Updated•5 years ago
|
Assignee | ||
Comment 1•5 years ago
|
||
Assignee | ||
Comment 2•5 years ago
|
||
https://treeherder.mozilla.org/#/jobs?repo=try&revision=353a9a34b632b36e6bbfcba7ce1d1a2377b36e09
This patch also reverts the function of JS::Evaluate
to inflate as it did before, to fully undo the effect of the last landing in bug 1504947, and it adds a new JS::EvaluateDontInflate
that does the UTF-8-aware tokenizing JS::Evaluate
that processes as UTF-8 currently does. There are no callers of the new function; I figure they can be added just like for all the other compilation APIs when needed.
Builds from the try-rev above should be suitable for testing to verify that the various perf regresions are gone.
Pushed by jwalden@mit.edu: https://hg.mozilla.org/integration/autoland/rev/2ae93b4399fb Revert column-number counts to counts of code units for release-or-beta, pending perf considerations in column-number counting being fixed. r=tcampbell
Comment 4•5 years ago
|
||
bugherder |
Updated•2 years ago
|
Updated•2 years ago
|
Description
•