WebKit Bugzilla
New
Browse
Log In
×
Sign in with GitHub
or
Remember my login
Create Account
·
Forgot Password
Forgotten password account recovery
RESOLVED FIXED
196208
[GTK][WPE] Disable process warming
https://bugs.webkit.org/show_bug.cgi?id=196208
Summary
[GTK][WPE] Disable process warming
Patrick Griffis
Reported
2019-03-25 11:20:29 PDT
[GTK][WPE] Disable process warming
Attachments
Patch
(1.69 KB, patch)
2019-03-25 11:23 PDT
,
Patrick Griffis
no flags
Details
Formatted Diff
Diff
Patch
(1.92 KB, patch)
2019-03-26 02:38 PDT
,
Philippe Normand
no flags
Details
Formatted Diff
Diff
Show Obsolete
(1)
View All
Add attachment
proposed patch, testcase, etc.
Patrick Griffis
Comment 1
2019-03-25 11:23:11 PDT
Created
attachment 365877
[details]
Patch
Michael Catanzaro
Comment 2
2019-03-25 11:40:46 PDT
Comment on
attachment 365877
[details]
Patch Process warming should already be disabled because PSON is disabled. Is there a bug in the logic for that? This is OK as an emergency measure, but the comment needs a FIXME, because this is very bad. Process warming is not optional for us any more than bubblewrap sandboxing is; we have to make them both work together. Without process warming, PSON will be a severe performance penalty. And we have to make PSON work for Spectre mitigation. I don't think prewarming is fundamentally incompatible with the bubblewrap sandbox. Prewarmed processes should just use the same WebsiteDataStore as their corresponding web view. If the application creates a view with a different WebsiteDataStore, then of course the prewarming will be defeated and a new process will need to be spun up, but that's an unusual case.
Michael Catanzaro
Comment 3
2019-03-25 11:47:05 PDT
(In reply to Michael Catanzaro from
comment #2
)
> Prewarmed processes should just use the same WebsiteDataStore as > their corresponding web view. If the application creates a view with a > different WebsiteDataStore, then of course the prewarming will be defeated > and a new process will need to be spun up, but that's an unusual case.
And if this logic has to be platform-specific because it's not needed by Apple, then so be it. We've accepted the constraint that all data directories must be known before launching the process, and it's not practical to revisit that, so the inescapable conclusion is WebsiteDataStore has to be ready when the process is launched.
Patrick Griffis
Comment 4
2019-03-25 11:51:18 PDT
(In reply to Michael Catanzaro from
comment #2
)
> Comment on
attachment 365877
[details]
> Patch > > Process warming should already be disabled because PSON is disabled. Is > there a bug in the logic for that?
I could not reproduce any issue without forcing PSON on, so I think its fine. Agreed about this being temporary though.
Chris Dumez
Comment 5
2019-03-25 12:17:01 PDT
Comment on
attachment 365877
[details]
Patch View in context:
https://bugs.webkit.org/attachment.cgi?id=365877&action=review
r=me
> Source/WebKit/UIProcess/glib/WebProcessPoolGLib.cpp:90 > + // Process warming is incompatible with the fact our WebProcessProxy::platformGetLaunchOptions()
You may want a FIXME comment as the performance of process-swap-on-navigation is not going to be great without process-prewarming.
Philippe Normand
Comment 6
2019-03-26 02:38:34 PDT
Created
attachment 365952
[details]
Patch
WebKit Commit Bot
Comment 7
2019-03-26 03:16:52 PDT
Comment on
attachment 365952
[details]
Patch Clearing flags on attachment: 365952 Committed
r243490
: <
https://trac.webkit.org/changeset/243490
>
WebKit Commit Bot
Comment 8
2019-03-26 03:16:54 PDT
All reviewed patches have been landed. Closing bug.
Note
You need to
log in
before you can comment on or make changes to this bug.
Top of Page
Format For Printing
XML
Clone This Bug