ggarand

.add KHTML/KJS changes for 4.4.1

Tue, 02 Mar 2010 16:05:22 +0000

Commit by ggarand :: r1097995 /trunk/www/sites/www/announcements/changelogs/ (2 files):

.add KHTML/KJS changes for 4.4.1

checking at last those baselines. Pheew.

Tue, 02 Mar 2010 14:18:41 +0000

Commit by ggarand :: r1097953 regression/trunk/tests/ (18 files in 4 dirs):

checking at last those baselines. Pheew.

test + baseline

Tue, 02 Mar 2010 14:11:44 +0000

Commit by ggarand :: r1097951 regression/trunk/tests/khtmltests/regression/ (4 files in 2 dirs):

test + baseline

checking most [RP] changes. .Mostly due to kling's rounding of CSS font sizes...

Tue, 02 Mar 2010 13:49:41 +0000

Commit by ggarand :: r1097943 regression/ (102 files in 30 dirs):

checking most [RP] changes.
.Mostly due to kling's rounding of CSS font sizes patch
.Some changes due to flow-around-floats'float clearing behaviour change. Verified.

initialize variable, fix comparison warning

Tue, 02 Mar 2010 13:04:51 +0000

Commit by ggarand :: r1097931 khtml/trunk/KDE/kdelibs/khtml/xml/ (dom_docimpl.cpp xml_tokenizer.cpp):

initialize variable, fix comparison warning

got a better idea here: instead of letting the textarea be always...

Tue, 02 Mar 2010 09:28:22 +0000

Commit by ggarand :: r1097847 khtml/trunk/KDE/kdelibs/khtml/ (css/html4.css rendering/render_form.cpp):

got a better idea here:

instead of letting the textarea be always ScrollBarAsNeeded,
we may mirror the overflow-x/y property and default to overflow-y:scroll

This matches MSIE7 so should be perfectly compatible with sites that want
to turn the scrollbars to auto for cosmetic reasons (with overflow:auto),
while retaining as default the very nice "this is a form widget"
accessibility hint that the vertical scrollbar naturally provides.

CCBUG: 221830

better spec conformance for loading of script generated sub-documents: No longer...

Tue, 02 Mar 2010 09:28:14 +0000

Commit by ggarand :: r1097846 khtml/trunk/KDE/kdelibs/khtml/ (4 files in 2 dirs):

better spec conformance for loading of script generated sub-documents:

No longer use an had hoc auto-close mode for frames opened by
document.write, rely on explicit close() instead.

-matches MSIE/Firefox : unclosed iframe keep owner document in loading
 state for as long as explicit close() hasn't been called.
.matches HTML5 draft for load event emission,
  cf. events/iframe_onload2.html

commit missing bits and reenable delayed execution of scripts. this "stage hunk...

Tue, 02 Mar 2010 09:28:07 +0000

Commit by ggarand :: r1097844 khtml/trunk/KDE/kdelibs/khtml/ (3 files in 2 dirs):

commit missing bits and reenable delayed execution of scripts.

this "stage hunk for commit" right click feature of git-gui is
super addictive, but it really can turn your tree into a mess :-}

fix my previous commit and rearrange some bits. Meanwhile, I need to do more...

Sun, 28 Feb 2010 09:28:48 +0000

Commit by ggarand :: r1097036 khtml/trunk/KDE/kdelibs/khtml/html/ (htmltokenizer.cpp htmltokenizer.h):

fix my previous commit and rearrange some bits.

Meanwhile, I need to do more tests and analysis, so disable the lock
until I have been able to do that.

top/bottom extra are already accounted in offsetHeight BUG: 200342

Sun, 28 Feb 2010 01:00:34 +0000

Commit by ggarand :: r1096958 khtml/trunk/KDE/kdelibs/khtml/rendering/render_table.cpp:

top/bottom extra are already accounted in offsetHeight

BUG: 200342

change the way we handle scripts to match other UA's behaviour. It is now...

Sat, 27 Feb 2010 03:32:34 +0000

Commit by ggarand :: r1096622 khtml/trunk/KDE/kdelibs/khtml/ (4 files in 2 dirs):

change the way we handle scripts to match other UA's behaviour.

It is now defacto standard to have scripts wait for any loading stylesheet
before they are considered for execution (thus providing more reliable
CSS computed values for render objects).

This is a port of two patches, the first implementing the above by Antti
Koivisto <anti a apple dt com>, the second being a time watcher for
yielding execution of possibly accumulated scripts, patch by Yael Aharon
<yael.aharon a nokia dt com> with some putative improvements -
probably overkill though, as I couldn't even craft a
testcase that would trigger it, and the original BR doesn't seem to
address a specific issue. Anyway, it's ported now ;-/

I did not address the first patch's limitations so this does not apply to
inline scripts yet (FIXME)

improvements of resources scheduling and overall speed progress of KIO over the...

Sat, 27 Feb 2010 03:32:20 +0000

Commit by ggarand :: r1096621 khtml/trunk/KDE/kdelibs/khtml/ (khtml_part.cpp khtmlview.cpp):

improvements of resources scheduling and overall speed
progress of KIO over the past year(s) allow to greatly optimize the paint
suppression and initial layout delay algorithm.

TTFP is halved on most pages, without any flashing!

port KHTML to Andrea's excellent scheduler. This, together with the next patch,...

Sat, 27 Feb 2010 03:32:08 +0000

Commit by ggarand :: r1096620 khtml/trunk/KDE/kdelibs/khtml/misc/ (loader.cpp loader.h):

port KHTML to Andrea's excellent scheduler.

This, together with the next patch, greatly improves Konqueror's
objective speed (Page Loading Time) as well as the subjective speed
(Time To First Paint).
Our PLT was already very nice, but this just blows things off.
Will have to explain that in greater details, because it's just
beautiful :-)

automatically merged revision 1088982: don't allow any synchronous update of...

Wed, 24 Feb 2010 02:18:01 +0000

Commit by ggarand :: r1095343 khtml/branches/KDE/4.4/kdelibs/khtml/rendering/ (render_layer.cpp render_layer.h):

automatically merged revision 1088982:
don't allow any synchronous update of scrollbars during layout

automatically merged revision 1088979: great, looks like these two Qt...

Wed, 24 Feb 2010 02:15:00 +0000

Commit by ggarand :: r1095342 khtml/branches/KDE/4.4/kdelibs/khtml/rendering/render_replaced.cpp:

automatically merged revision 1088979:
great, looks like these two Qt regressions have finally been fixed,
so we can enable back the optimized code and close all the printing bugs
that closing/restoring the QPainter triggered

I'll do that in two commits so that problems can be pin-pointed :

1) stop ending the QPainter before painting widgets to buffers,
 it's not needed anymore and Qt has poor support for that.

 Fixes printing on pages with form widgets.

BUG: 197402
BUG: 214352

CSS3 Backgrounds & Borders: implement 'contain' and 'cover' values for...

Wed, 24 Feb 2010 00:55:31 +0000

Commit by ggarand :: r1095329 khtml/trunk/KDE/kdelibs/khtml/ (8 files in 2 dirs):

CSS3 Backgrounds & Borders:

implement 'contain' and 'cover' values for background-size

CSS3 Backgrounds & Borders compliance fix. use the correct property values for...

Wed, 24 Feb 2010 00:55:11 +0000

Commit by ggarand :: r1095328 khtml/trunk/KDE/kdelibs/khtml/css/ (5 files):

CSS3 Backgrounds & Borders compliance fix.

use the correct property values for bg-clip/bg-origin.
i.e. suffix them with '-box' as mandated by CR specification.

automatically merged revision 1094436: let's grab more stack room if the...

Mon, 22 Feb 2010 21:09:39 +0000

Commit by ggarand :: r1094496 kjs/branches/KDE/4.4/kdelibs/kjs/ (regexp.cpp regexp.h):

automatically merged revision 1094436:
let's grab more stack room if the operating system agrees.
this will let really ill-designed regexps match +/-32Kb in the worst case
(recursion at each char).

Regular expressions unable to perform their function in such conditions
deserve to die a preposterous death.

automatically merged revision 1094434: InlineBox::deleteLine shouldn't try to...

Mon, 22 Feb 2010 21:09:02 +0000

Commit by ggarand :: r1094495 khtml/branches/KDE/4.4/kdelibs/khtml/rendering/render_line.cpp:

automatically merged revision 1094434:
InlineBox::deleteLine shouldn't try to remove() the line box before
disposing of it.

This causes crashes in release mode because remove() will call to a
possibly invalid parent (this parent-is-invalid circumstance isn't fully
clear to me, though), the link to which is only nulled out in
Debug mode.

Let's play this low risk : align Release on Debug for branch backport...

BUG: 193717

automatically merged revision 1094432: avoid doing any unneeded work in...

Mon, 22 Feb 2010 21:08:01 +0000

Commit by ggarand :: r1094493 khtml/branches/KDE/4.4/kdelibs/khtml/rendering/render_layer.cpp:

automatically merged revision 1094432:
avoid doing any unneeded work in scrollToOffset when the offset ends up