Bug 106107 – REGRESSION(SUBPIXEL_LAYOUT): el.offsetWidth < el.clientWidth for elements of a certain size

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Brad Vogel]

Repro:
1) Run the attached file in Chrome 21 or later, or Webkit Nightly.

Actual Result:
You will see the alert "BUG: offsetWidth is smaller than clientWidth".

Expected Result:
You will see the alert "NO BUG: unable to repro"

Tested in:
Firefox 17.0.1: PASS
Chrome 20: PASS
Safari 6.0.2 (7536.26.17): PASS
Chrome 21: FAIL
Chrome Canary 26.0.1374.1: FAIL
Webkit Nightly Version 6.0.2 (7536.26.17, 537+): FAIL

Notes:
Per the W3C spec, offsetWidth should never be smaller than clientWidth.

If you change the attached repro case to use "margin: 1.2em" instead of "margin: 1.1em", it will not repro the bug.

Impact:
This breaks scripts that use the following code to check if an element is scrollable:
    var isScrollable = el.offsetWidth < el.scrollWidth;
because if the bug is reproducing, offsetWidth will be smaller than clientWidth, but scrollWidth will equal clientWidth.

[Brad Vogel]

Created attachment 181330 [details]
Original repro case

[Alexey Proskuryakov]

Regressed in <http://trac.webkit.org/changeset/133351>.

[Emil A Eklund]

Looking into this now.

[Emil A Eklund]

Created attachment 181577 [details]
Patch

[Levi Weintraub]

Comment on attachment 181577 [details]
Patch

LGTM.

[Brad Vogel]

LGTM!

[Levi Weintraub]

You may want to merge this into m25 given it's super low risk and high return.

[WebKit Review Bot]

Comment on attachment 181577 [details]
Patch

Clearing flags on attachment: 181577

Committed r139013: <http://trac.webkit.org/changeset/139013>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Ryosuke Niwa]

This patch caused a build failure on Windows:
http://build.webkit.org/builders/Apple%20Win%20Debug%20%28Build%29/builds/60388/steps/compile-webkit/logs/stdio

7>LINK : warning LNK4075: ignoring '/INCREMENTAL' due to '/LTCG' specification
7>WebCore.lib(RenderRegionSet.obj) : error LNK2005: "public: virtual int __thiscall WebCore::RenderBox::pixelSnappedOffsetWidth(void)const " (?pixelSnappedOffsetWidth@RenderBox@WebCore@@UBEHXZ) already defined in WebCore.lib(RenderingAllInOne.obj)
7>WebCore.lib(RenderRegionSet.obj) : error LNK2005: "public: virtual int __thiscall WebCore::RenderBox::pixelSnappedOffsetHeight(void)const " (?pixelSnappedOffsetHeight@RenderBox@WebCore@@UBEHXZ) already defined in WebCore.lib(RenderingAllInOne.obj)
7>WebCore.lib(MathMLInlineContainerElement.obj) : error LNK2005: "public: virtual int __thiscall WebCore::RenderBox::pixelSnappedOffsetWidth(void)const " (?pixelSnappedOffsetWidth@RenderBox@WebCore@@UBEHXZ) already defined in WebCore.lib(RenderingAllInOne.obj)
7>WebCore.lib(MathMLInlineContainerElement.obj) : error LNK2005: "public: virtual int __thiscall WebCore::RenderBox::pixelSnappedOffsetHeight(void)const " (?pixelSnappedOffsetHeight@RenderBox@WebCore@@UBEHXZ) already defined in WebCore.lib(RenderingAllInOne.obj)
7>   Creating library C:\cygwin\home\buildbot\slave\win-debug\build\WebKitBuild\Debug\lib\WebKit.lib and object C:\cygwin\home\buildbot\slave\win-debug\build\WebKitBuild\Debug\lib\WebKit.exp
7>C:\cygwin\home\buildbot\slave\win-debug\build\WebKitBuild\Debug\bin\WebKit.dll : fatal error LNK1169: one or more multiply defined symbols found
7>Build log was saved at "file://C:\cygwin\home\buildbot\slave\win-debug\build\WebKitBuild\Debug\obj\WebKitLib\BuildLog.htm"

It's probably a dependency problem though because I don't see RenderBox.cpp being recompiled.