87381 - BRS_ISINLINEINCRREFLOW schedules an extra dirty reflow

Status: VERIFIED FIXED

(Core :: Layout, defect, P4)

Description

[Chris Waterson]

...and I don't understand why. I've removed the code and done some smoketesting,
and as far as I can tell it's completely vestigial. (This code dates back to
some of buster's ``typing optimizations'' that I found to be rather
impenetrable.) Anyway, I'll attach a patch that removes the code. If people
could try the patch, or explain to me why we need the code, I'd appreciate it!

Comment 1

[Chris Waterson]

Attachment: remove ReflowDirtyChild() call when BRS_ISINLINEINCRREFLOW

Comment 2

[Chris Waterson]

Okay, I see why buster put this here, and I also see why I think it's vestigial.
Reflows are only ever targeted at block frames (see the ReflowDirtyChild() method).

So, how can we _ever_ end up with a reflow that's targeted at an inline? When
the inline contains a text field, which contains a block frame for its magic
<div> anonymous content.

I'm dubious that this code is worth keeping in the tree. It obfuscates the fact
that inline frames cannot be targeted for reflow. (It seems like a half-hearted
attempt to begin generalizing the reflow system so that inline frames _could_ be
individually targeted for reflow: the only situation in which we'd currently
have a BRS_ISINLINEINCRREFLOW is when targeting reflow at a text widget, which
IsIncrementalDamangeConstrained() detects, and bails us out of.)

I move to strike it from the record. Thoughts?

Comment 3

[Chris Waterson]

> So, how can we _ever_ end up with a reflow that's targeted at an inline?
> When the inline contains a text field, which contains a block frame for its
> magic <div> anonymous content.

I've just realized that this isn't the only case: <button> creates an 
area-frame, which is a block frame, and can therefor be the target of an 
incremental reflow.

Comment 4

[Chris Waterson]

Attachment: test case; <button> with kooky button:hover style

Comment 5

[Chris Waterson]

So, the above test case seems to work fine without the extra dirty reflow. It
has a button with an outrageous hover pseudo rule that dramatically changes the
size of the button.

Comment 6

[Marc Attinasi]

Looks good to me: [s]r=attinasi

Comment 7

[rbs]

r=rbs

Comment 8

[Chris Waterson]

Checked in.

Comment 9

[Chris Petersen]

Marking verified per last comments.