May 2006 - webkit-dev - lists.webkit.org

Ever seen intermittent WebKit build errors like these?
by Trey Matteson 01 Jun '06

01 Jun '06

I recently took a fresh pull of the WebKit repository, and when doing the initial build ran into all sort of weird, semi-reproducible errors (examples below). Often if I immediately retry the build, the same file will succeed, but then another similar weird error will crop up down the line. A few times the same file would persistently fail, but after a reboot the build would make progress. This is using Xcode 2.1 and 2.2.1, and I saw it on 10.4.2 and (mostly) 10.4.6. I even tried a fresh 10.4.6 install. I'm curious if anyone else has ever run into similar errors building these projects. It looks to me like some sort of hardware problem related to the disk storage, as if either the intermediate compiler products or the pages from the source files are getting corrupted (sometimes I get complaints about weird chars in the source). Along these lines, I tried switching hard disks, but saw the same issues. I tried both disks both internally and over Firewire, and FWIW this is a dual 1.25 GHz G4. On the other hand, I don't get any system hangs or panics, which would be more like other hardware problems I've had it the past. I don't really have any other odd symptoms besides these flakey WebKit builds. Sorry that this is all somewhat tangential, but any related info would be appreciated, in case you've seen similar phantoms. Thanks, trey /usr/bin/gcc-4.0 -x c++ -arch ppc -pipe -Wno-trigraphs -fno- exceptions -fno-rtti -fpascal-strings -fasm-blocks -g -O0 -Werror - Wnon-virtual-dtor -DSVG_SUPPORT -DXPATH_SUPPORT -fmessage-length=0 - mtune=G5 -fno-threadsafe-statics -I/Volumes/Whopper/symroots/ WebCore.build/Debug/WebCore.build/WebCore.hmap -Wall -W -Wcast-align - Wchar-subscripts -Wformat-security -Wmissing-format-attribute - Wpointer-arith -Wwrite-strings -Wno-format-y2k -Wno-unused-parameter - Wno-long-double -F/Volumes/Whopper/symroots/Debug -I/Volumes/Whopper/ symroots/Debug/include -IForwardingHeaders -I. -Ikhtml -Iicu -I/usr/ include/libxslt -I/usr/include/libxml2 -I/Volumes/Whopper/symroots/ Debug/DerivedSources/WebCore -I/Volumes/Whopper/symroots/ WebCore.build/Debug/WebCore.build/DerivedSources -include /Library/ Caches/com.apple.Xcode.501/SharedPrecompiledHeaders/WebCorePrefix- btosnkekqzqbbacwyfktbotyehgt/WebCorePrefix.h -c /Volumes/Whopper/ WebKit/WebCore/dom/Range.cpp -o /Volumes/Whopper/symroots/ WebCore.build/Debug/WebCore.build/Objects-normal/ppc/Range.o /Volumes/Whopper/WebKit/WebCore/dom/Range.cpp:1366: internal compiler error: Bus error Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://developer.apple.com/bugreporter> for instructions. {standard input}:10:FATAL:.abort detected. Assembly stopping. /usr/bin/gcc-4.0 -x objective-c -arch ppc -pipe -std=gnu99 -Wno- trigraphs -fobjc-exceptions -fpascal-strings -fasm-blocks -g -O0 - Werror -Wmissing-prototypes -DFRAMEWORK_NAME=WebKit - DENABLE_WEBKIT_UNSET_DYLD_FRAMEWORK_PATH - DREMOVE_SAFARI_DOM_TREE_DEBUG_ITEM -fmessage-length=0 -fobjc-direct- dispatch -fobjc-gc -mtune=G5 -I/Volumes/Whopper/symroots/WebKit.build/ Debug/WebKit.build/WebKit.hmap -Wall -W -Wcast-align -Wchar- subscripts -Wformat-security -Wmissing-format-attribute -Wpointer- arith -Wwrite-strings -Wno-format-y2k -Wno-unused-parameter -Wbad- function-cast -Wmissing-declarations -Wnested-externs -F/Volumes/ Whopper/symroots/Debug -F/System/Library/Frameworks/ ApplicationServices.framework/Frameworks -F/System/Library/Frameworks/ Carbon.framework/Frameworks -F/System/Library/Frameworks/ Quartz.framework/Frameworks -F/System/Library/PrivateFrameworks -I/ Volumes/Whopper/symroots/Debug/include -I/Volumes/Whopper/symroots/ Debug/usr/local/include -Iicu -I/Volumes/Whopper/symroots/ WebKit.build/Debug/WebKit.build/DerivedSources -include /Library/ Caches/com.apple.Xcode.501/SharedPrecompiledHeaders/WebKitPrefix- bmukbrhvffuffjanpqnzuyyyolzl/WebKitPrefix.h -c /Volumes/Whopper/ WebKit/WebKit/Misc/WebNSViewExtras.m -o /Volumes/Whopper/symroots/ WebKit.build/Debug/WebKit.build/Objects-normal/ppc/WebNSViewExtras.o /Volumes/Whopper/WebKit/WebKit/Misc/WebNSViewExtras.m: In function '- [NSView(WebExtras) _web_DragImageForElement:rect:event:pasteboard:source:offset:]': /Volumes/Whopper/WebKit/WebKit/Misc/WebNSViewExtras.m:211: error: unrecognizable insn: (insn/f 367 366 368 /Volumes/Whopper/WebKit/WebKit/Misc/ WebNSViewExtras.m:169 (parallel [ (set (mem:SI (plus:SI (reg/f:SI 1 r1) (const_int -103079215128 [0xffffffe7ffffffe8])) [0 S4 A8]) (reg:SI 28 r28)) (set (mem:SI (plus:SI (reg/f:SI 1 r1) (const_int -103079215124 [0xffffffe7ffffffec])) [0 S4 A8]) (reg:SI 29 r29)) (set (mem:SI (plus:SI (reg/f:SI 1 r1) (const_int -16 [0xfffffffffffffff0])) [0 S4 A8]) (reg/f:SI 30 r30)) (set (mem:SI (plus:SI (reg/f:SI 1 r1) (const_int -12 [0xfffffffffffffff4])) [0 S4 A8]) (reg:SI 31 r31)) ]) -1 (nil) (expr_list:REG_FRAME_RELATED_EXPR (parallel [ (set/f (mem:SI (plus:SI (reg/f:SI 1 r1) (const_int -103079215128 [0xffffffe7ffffffe8])) [0 S4 A8]) (reg:SI 28 r28)) (set/f (mem:SI (plus:SI (reg/f:SI 1 r1) (const_int -103079215124 [0xffffffe7ffffffec])) [0 S4 A8]) (reg:SI 29 r29)) (set/f (mem:SI (plus:SI (reg/f:SI 1 r1) (const_int -16 [0xfffffffffffffff0])) [0 S4 A8]) (reg/f:SI 30 r30)) (set/f (mem:SI (plus:SI (reg/f:SI 1 r1) (const_int -12 [0xfffffffffffffff4])) [0 S4 A8]) (reg:SI 31 r31)) ]) (nil))) /Volumes/Whopper/WebKit/WebKit/Misc/WebNSViewExtras.m:211: internal compiler error: in insn_default_length, at insn-attrtab.c:768 Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://developer.apple.com/bugreporter> for instructions. {standard input}:766:FATAL:.abort detected. Assembly stopping.

2 1

some feedback on the calcAbsoluteHorizontalValues patch
by Darin Adler 31 May '06

31 May '06

Begin forwarded message: > From: Allan Sandfeld Jensen <kde(a)carewolf.com> > Date: May 31, 2006 7:56:38 AM PDT > To: ap(a)opensource.apple.com > Cc: khtml-devel(a)kde.org > Subject: Re: [webkit-changes] [14351] trunk/WebCore > Reply-To: khtml-devel(a)kde.org > > On Saturday 13 May 2006 16:58, ap(a)opensource.apple.com wrote: > >> 2006-05-13 Sam Weinig <sam.weinig(a)gmail.com> >> >> Reviewed by Hyatt, landed by ap. >> >> WebCore: >> Patch for http://bugzilla.opendarwin.org/show_bug.cgi?id=7604 >> calcAbsoluteHorizontalValues() is being getting passed >> arguments >> in the wrong order in calcAbsoluteHorizontal() >> >> Cleans up the RenderBox code for absolutely positioned >> elements >> and adds new functions for replaced absolutely positioned >> elements. Now uses Length so that magic number -666666 for >> auto lengths is no longer used. >> >> * rendering/RenderBox.cpp: >> (WebCore::RenderBox::calcAbsoluteHorizontal): >> (WebCore::RenderBox::calcAbsoluteHorizontalValues): >> (WebCore::RenderBox::calcAbsoluteVertical): >> (WebCore::RenderBox::calcAbsoluteVerticalValues): >> (WebCore::RenderBox::calcAbsoluteHorizontalReplaced): Handle >> replaced case separately. >> (WebCore::RenderBox::calcAbsoluteVerticalReplaced): ditto. >> * rendering/RenderBox.h: > > I just merged this patch to KHTML, and noticed a regression: > > In the overconstrained case in calcAbsoluteVerticalReplaced() (step > 6) you > discard information about top, margin-top and margin-bottom. > > It should look like this: > / > *--------------------------------------------------------------------- > --*\ > * 6. If at this point the values are over-constrained, ignore > the value > * for 'bottom' and solve for that value. > > \*-------------------------------------------------------------------- > ---*/ > else { > m_marginTop = marginTop.width(containerHeight); > m_marginBottom = marginBottom.width(containerHeight); > topValue = top.width(containerHeight); > > // Solve for 'bottom' > // NOTE: It is not necessary to solve for 'bottom' because > we don't ever > // use the value. > } > > There is of course the same problem in calcAbsoluteVerticalReplaced > () where it > should look like: > / > *--------------------------------------------------------------------- > --*\ > * 6. If at this point the values are over-constrained, ignore > the value > * for either 'left' (in case the 'direction' property of the > * containing block is 'rtl') or 'right' (in case > 'direction' is > * 'ltr') and solve for that value. > > \*-------------------------------------------------------------------- > ---*/ > else { > m_marginLeft = marginLeft.width(containerWidth); > m_marginRight = marginRight.width(containerWidth); > if (containerDirection == LTR) { > leftValue = left.width(containerWidth); > rightValue = availableSpace - (leftValue + m_marginLeft > + m_marginRight); > } > else { > rightValue = right.width(containerWidth); > leftValue = availableSpace - (rightValue + m_marginLeft > + m_marginRight); > } > } > > Regards > `Allan

2 1

problem of compilation of webkit with S60 SDK
by Ronan Meneu 30 May '06

30 May '06

Hello, I have just tried to build webkit on my windows machine (svn revision 14621). I think everything is set up properly. I get the following logs : - - - - - Building under x: drive 6 archives were successfully processed. Synching \S60\LayoutTests with \epoc32\winscw\c\LayoutTests 0 LayoutTests files updated 0 LayoutTests files removed WARNING: Cannot determine the version of the RVCT Compiler. In file included from .\BLD.INF:1: S60Internals\S60.hrh:1: ..\..\epoc32\include\variant\Symbian_OS_v9.1.hrh: No suc h file or directory BLDMAKE ERROR: ERROR: cpp.exe returned non-zero exit status (8448) cpp.EXE -undef -nostdinc -+ -I ..\epoc32\include -I . -I .\ -I ".\S60I nternals" -include S60Internals\S60.hrh ".\BLD.INF" - - - - - The file Symbian_OS_v9.1.hrh is required, but does not seem to be part of the S60 SDK fo symbian OS, for C++; third edition. I have noticed that "..\..\epoc32\include\" is there, but does not contain any variant directory. Any ideas ? Thanks Ronan

3 3

Re: [webkit-dev] programmatic text selection
by Giacomo Luca Maioli 27 May '06

27 May '06

> On 5/26/06, Giacomo Luca Maioli <luca(a)tricheco.net> wrote: >> >> But Safari throws an error while processing the last method (the >> one which actually highlight text), so you should try: >> >> >> window.getSelection().setBaseAndExtent(selDiv, 0, selDiv, 1); >> >> So you'll only need to play around with this (strange and mostly >> unknown) method. > > I'm afraid Safari doesn't appear to support setBaseAndExtent() method > on a selection, either. > I just tested the code on both Firefox and Safari. You can find a working example here: http://www.tricheco.net/selection.html I'm using Safari Version 2.0.3 (417.9.3) and Firefox 1.5.0.3 on Mac and Windows.

6 9

Is WebKit's javascript subpar?
by Abhi Beckert 26 May '06

26 May '06

I've seen several blog posts recently (eg: http://rentzsch.com/code/dashcodeForAjaxAppDevelopment) that boldly state WebKit's "ajax support" is vastly inferior to FireFox. All of them have been very vague and haven't specified exactly where WebKit is lacking, so I thought I'd ask you guys: Is WebKit inferior, or is it just because FF is cross platform/has more users, and the companies are focusing on FF first, and other engines later? If these claims are unfounded, lets chop them off at the head.

14 20

programmatic text selection
by Sam Gendler 26 May '06

26 May '06

Folks, I've been all over the net without finding a definitive answer, so I thought I'd come to the source. Is there any way, in Safari/webkit, to programmatically select some text in javascript?

2 2

s60 build questions
by Mindy P 26 May '06

26 May '06

Thrilled to see the s60 sources available... Tried to build today and seem to have gotten around all the issues. It seems like the MemoryManager.h file was not included in a number of places it should have been? Or there was some other error preventing the correct inclusion? Either way, it did build and I have the Reindeer icon available in on the simulator, but it crashes as soon as I try to launch it. Anyone else seeing this? Any ideas as to what I may have done incorrectly?

4 8

Nokia releases S60 WebKit sources
by roland.geisler＠nokia.com 24 May '06

24 May '06

Hi, I am at the moment sitting at the World Wide Web Conference 2006 in Edinburgh where a few minutes ago Nokia announced to open source its S60 WebKit sources. We will release our code under the terms of the open source BSD License, and our source code will be made available through the WebKit Open Source Project site. The code that we are contributing includes many mobility components that are needed for a basic browser, including items such as a memory manager, mobile usability improvements and a reference user interface. In addition we have prepared a number of testing tools and documentation to get you as a developer quickly started. More information can be found from our open source site at http://webkit.opendarwin.org/. As we are moving forward we will be making improvements to tools and release further features. It has been exciting and at times challenging one and a half years for us to develop our new mobile Web browser for the S60 software platform based on Apple's and KDE's fantastic Safari and Konqueror browser code. In 2Q Nokia started shipping our first S60 3rd edition mobile phones with the new browser, and I am pleased to mention that our customers have been highly pleased with the good performance, compliance and feature set of our new mobile browser. I wanted to take the opportunity to thank the KDE community and the Apple Safari team for having built what is today WebCore and JavaScriptCore, a large part of your code is now running in our Nokia mobile phones. After having focused for the past months mostly on our first browser software release, we at Nokia felt that it is time to start looking ahead. With the decision of open sourcing our S60 WebKit code we hope to start a fruitful collaboration with the open source community and partner companies to further develop the Web Browser for S60 and proliferate the engine to other mobile software platforms. We believe that working in the open source will create an innovative ecosystem for mobile browser development that will benefit the mobile industry and browser industry at large. We at Nokia are excited to take this step, and we are looking forward to working together with many of you from the community and partner companies in the near future. If you would like to contact us, please feel free to send us email either directly, via the WebKit mailing list or meet us on the WebKit IRC channel. Best regards, Your Web Browser for S60 team at Nokia, Roland Geisler Head of Marketing & Strategy, S60 Browser Nokia PS: Please forward this email to any contributor whom I may have missed.

1 0

Bugs for S60 port
by bradley.morrison＠nokia.com 24 May '06

24 May '06

Hi there, As you may or may not be aware, a branch has been created for the S60 port of WebKit. Also, quite recently, a couple of new S60-related platforms were added to the WebKit product's platform list. We were wondering if it's OK with everyone to start migrating a bunch (say roughly 15/25) of errors related to the S60 port into Bugzilla? Some of these errors will have been fixed in the ToT (like [1]) but some may not be (like [2]). Of course, we're more than happy to assign the bugs to people working in the S60 branch! The plan is at this stage to create them against the WebKit product but differentiate them a little by having an S60-related platform value, but I'm not sure if that makes people's life hard. If so maybe I could talk to Joost or someone about potentially getting an S60WebKit product, but we'd rather stick with the WebKit product ... Many thanks, Bradley [1] http://bugzilla.opendarwin.org/show_bug.cgi?id=8959 [2] http://bugzilla.opendarwin.org/show_bug.cgi?id=8950

3 2

Re: [webkit-dev] Will WebKit build with Xcode 2.3?
by Eric Seidel 23 May '06

23 May '06

Thanks Jon! -eric On May 23, 2006, at 1:11 PM, Jon Shier wrote: > Just an FYI, but I just built ToT with 2.3 a few minutes ago > without a problem. I haven't noticed any behavior changes yet either. > > > > Jon

1 0