75559 - Style re-resolution due to content state changes needs to be optimized

Status: RESOLVED WONTFIX

(Core :: CSS Parsing and Computation, defect, P3)

Description

[Marc Attinasi]

When some piece of content has a state change we currently resolve all style for
that element and all of its children. This is terribly inefficient, and causes
incredible performance problems when rules like '*:active' are present because
any state change ends up re applying ALL rules on potentially all elements (if
the BODY or root element is the one who's state is changing). A smarter approach
would be to simply re-apply the style rules that have dynamic pseudos in them,
as these are the only ones that can change when a content state change happens.
Even better would be to restrict it to rules that have a specific dynamic
pseudo, like the :active when a content state change for 'active' happens, or
:hover when a content state for hover happens.

The hack put in for bug 58821 can then be removed, allowing a wider application
of the dynamic pseudos without suffering from abysmal performance.

Comment 1

[James Green]

CC self, perf keyword and nominating for 0.9.1, 1.0, since we wouldn't want to
forget about this for too long :)

Comment 2

[Hixie (not reading bugmail)]

cc'ing hyatt since :active, :hover and :focus (and later :disabled, :enabled,
:checked, :selected, and what not) are things that I believe interest him. ;-)

Comment 3

[Marc Attinasi]

Adding the URL from the old bug - Ian, please don't remove that page!

Also, setting milestone to Moz 1.0 for now - far too many 0.9.1 bugs on my lap
already, and this one does not crash or hang or corrupt data ;)

Comment 4

[Marc Attinasi]

I wonder if this was fixed with the RuleTree landing...

Comment 5

[Kevin McCluskey (gone)]

Reassigning to Pierre

Comment 6

[Markus Hübner]

Wouldn't this be a perfect 1.0 candidate?

Comment 7

[Kevin McCluskey (gone)]

Bulk moving Mozilla1.01 bugs to future-P1

Comment 8

[Boris Zbarsky [:bzbarsky]]

*** Bug 132598 has been marked as a duplicate of this bug. ***

Comment 9

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

My patch in bug 5693 has some optimizations -- it makes HasStateDependentStyle
state-specific.

Bug 96984 is about removing the hack.

Comment 10

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

See also bug 137395, which is a bug that we should try to fix when we consider a
better way to optimize this.

Comment 11

[Madhur Bhatia]

cc'ing myself

Comment 12

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Assigning pierre's remaining Style System-related bugs to myself.

Comment 13

[Boris Zbarsky [:bzbarsky]]

So now that we have HasStateDependentStyle, is this still an issue?

Comment 14

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

No.