August 2009 - webkit-dev - lists.webkit.org

Some JSC API Suggestions
by Brian Barnes 14 Aug '09

14 Aug '09

A couple of "convenience" API suggestions. Again, I'm in the middle of a big switch over from Mozilla's SpiderMonkey to JSC, so I see a couple IMHO (and I've been wrong before!) API additions that might make it easier for new users. .... JSValueMakeObject(.....) I know you can cast to the JSObjectRef from JSValueRef (if it is indeed a value), but it's not immediately obvious from the API (I figured it out from the examples.) And it gives an equivalent to JSValueToObject, like all other types have. What to do if the value isn't an object is probably one reason it doesn't exist, but maybe that could be doc'd to return a JS null or exception could actually be part of the call. ... JSValueMakeArray(.....size_t argumentCount,JSValueRef *arguments.....) Another one that can obviously be done in code but would be nice as a convenience. Right now it takes a good 5-10 lines to build an array (complete with values, etc.) ... JSValueMakeException( .... const char *str ....) Just a way to make exception creating easy, and it could attach sourceURL, line, etc, there (this would also give you a cheap way to find out where you are in execution if you wanted to build some kind of specialized debugger, make an exception, get the line -- the same trick you do in Java.) It makes throwing exceptions a bit easier. Also, SpiderMonkey had one nice thing that could be used, which is sprintf like variable count exception functions, for instance: *exception=JSValueMakeException(ctx,"There is no %s by the name of %s",thing_str,name_str); That would make it very easy to build and throw exceptions (in only one line most of the time) for most any exception. Again, thanks for all the great work and listening to my suggestions, which might be dumb :) I also have a couple places where I think the documentation generated from the header files could be updated, but I'll just bug you guys with one thing at a time. [>] Brian

1 0

commit-queue Documentation
by Eric Seidel 13 Aug '09

13 Aug '09

Adam Barth announced last week that he'd wrapped bugzilla-tool in a shell-script and created the commit-queue. We're now running this queue every day, but not overnight yet. The commit-queue script is super-simple. It wakes up every 10 minutes and runs: bugzilla-tool bugs-to-commit | xargs -n1 bugzilla-tool land-patches --force-clean --commit-queue --quiet (see https://bugs.webkit.org/show_bug.cgi?id=28040) I've made an attempt at answering some of the commit-queue questions which came up today: https://trac.webkit.org/wiki/CommitQueue If folks have more questions/suggestions, I'd love to hear them. :) -eric

1 0

proposal to expose word breaker through JavaScript: range.expand()
by Xiaomei Ji 13 Aug '09

13 Aug '09

I proposed to expand a range to word boundary. Here <https://bugs.webkit.org/show_bug.cgi?id=27632>is the related webkit bug. Below is the spec. Appreciated any feedbacks. *Syntax*: expands the range to the 'unit' boundary. interface Range { void expand(in DOMString uint) } *Parameters*: unit: String that specifies the units to move in the range, using one of the following values: word -- expand the range to include completed words. A word is the smallest semantic form in one language. In languages use space to break word, such as English, a word is a collection of characters terminated by a space or punctuation. sentence -- expand the range to include completed sentences. A sentence is a collection of words terminated by punctuation. block -- expand the range to include completed paragraphs. document -- expand the range to include the whole document. *Use case*: To identify a semantic unit (such as a word) when user mouse over or mouse click on a page. A specific use case is dictionary, which shows the worddefinition when user mouse over or mouse click in a webpage. *Example*: This example returns the range containing the word user moused over or mouse clicked (not double-clicked). var range = document.caretRangeFromPoint<http://dev.w3.org/csswg/cssom-view/#dom-documentview-caretrangefrompoint>(event.clientX, event.clientY); range.expand('word'); *Reference*: Microsoft's spec<http://msdn.microsoft.com/en-us/library/ms536421(VS.85).aspx> . Thanks, Xiaomei On Thu, Jul 30, 2009 at 4:55 AM, Anne van Kesteren <annevk(a)opera.com> wrote: > Thanks for the reply! > > On Wed, 29 Jul 2009 20:17:13 +0200, Xiaomei Ji <xji(a)chromium.org> wrote: > > "word" is a keyword. Like Microsoft's > > spec<http://msdn.microsoft.com/en-us/library/ms536421%28VS.85%29.aspx>, > > a range could be extended to a 'character', a 'word', a 'sentence', or a > > 'line' etc. > > > > As to whether it should be a method in Document or in Range, it is open > > to discussion. > > I think putting it on Range similar to what Microsoft has done would be > better. Document is already pretty bloated. > > > >> And finally, use cases would help as well as a definition of "word > >> boundary" and how this works/won't work in an international context. > > > > A word is the smallest semantic form in one language. In languages use > > space to break word, such as English, a word is a collection of > characters > > terminated by a space or punctuation. In languages do not use space to > > break word, such as Chinese, word breaker is needed to break a word. > > > > The API should work for English, at least. Whether it works in an > > international context depends on whether the layout engine/browser > > supports correct word breaker in that language. > > I do think we should define this. > > > I think the main problem here is that we do not have an editor for a new > version of DOM Range. There are several other extensions that browsers have > implemented and emulated from each other that would be nice to document > clearly, but so far nobody has volunteered. > > (It could maybe be done as a separate draft as well, similarly to how CSSOM > View extends DOM Range, but updating DOM Range to today's standards for > specifications would be good I think.) > > > -- > Anne van Kesteren > http://annevankesteren.nl/ >

1 0

Leaks in layout tests
by Drew Wilson 13 Aug '09

13 Aug '09

I'm seeing a bunch of leaks when running even simple layout tests (i.e. run-webkit-tests --debug --leaks fast/js/typeof-syntax.html). This ring any bells for anyone?: start main dumpRenderTree(int, char const**) runTestingServerLoop() runTest(std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) -[NSRunLoop(NSRunLoop) runMode:beforeDate:] CFRunLoopRunInMode CFRunLoopRunSpecific MultiplexerSource::perform() URLConnectionClient::processEvents() URLConnectionClient::ClientConnectionEventQueue::processAllEventsAndConsumePayload(XConnectionEventInfo<XClientEvent, XClientEventParams>*, long) URLConnectionClient::_clientDidFinishLoading(URLConnectionClient::ClientConnectionEventQueue*) _NSURLConnectionDidFinishLoading -[NSURLConnection(NSURLConnectionReallyInternal) sendDidFinishLoading] -[WebCoreResourceHandleAsDelegate connectionDidFinishLoading:] WebCore::ResourceLoader::didFinishLoading(WebCore::ResourceHandle*) WebCore::SubresourceLoader::didFinishLoading() WebCore::Loader::Host::didFinishLoading(WebCore::SubresourceLoader*) WebCore::CachedCSSStyleSheet::data(WTF::PassRefPtr<WebCore::SharedBuffer>, bool) WebCore::CachedCSSStyleSheet::checkNotify() WebCore::HTMLLinkElement::setCSSStyleSheet(WebCore::String const&, WebCore::String const&, WebCore::CachedCSSStyleSheet const*) WebCore::CSSStyleSheet::checkLoaded() WebCore::HTMLLinkElement::sheetLoaded() WebCore::Document::removePendingSheet() WebCore::HTMLTokenizer::executeScriptsWaitingForStylesheets() WebCore::HTMLTokenizer::notifyFinished(WebCore::CachedResource*) WebCore::HTMLTokenizer::scriptExecution(WebCore::ScriptSourceCode const&, WebCore::HTMLTokenizer::State) WebCore::FrameLoader::executeScript(WebCore::ScriptSourceCode const&) WebCore::ScriptController::evaluate(WebCore::ScriptSourceCode const&) JSC::evaluate(JSC::ExecState*, JSC::ScopeChain&, JSC::SourceCode const&, JSC::JSValue) WTF::PassRefPtr<JSC::ProgramNode> JSC::Parser::parse<JSC::ProgramNode>(JSC::ExecState*, JSC::Debugger*, JSC::SourceCode const&, int*, JSC::UString*) JSC::Parser::parse(JSC::JSGlobalData*, int*, JSC::UString*) jscyyparse(void*) JSC::ParserArenaDeletable::operator new(unsigned long) WTF::fastMalloc(unsigned long) malloc malloc_zone_malloc

4 4

Help finding element at point
by Chris Mumford 13 Aug '09

13 Aug '09

Hi: I'm trying to figure out the DOM element at a certain (x,y) point. To do this I'm using Document::prepareMouseEvent() which seems to return what I expect - most of the time. Specifically <input type="text"> nodes are returned as <div>#text</div>. It appears to me as though I'm getting the internal representation of the text input element. How can I get the "real" element at a given location? Thanks.

6 6

Volunteers to run commit-queue?
by Adam Barth 12 Aug '09

12 Aug '09

Hi webkit-dev, I'm going to be in Canada next week for USENIX Security, so I won't be able to run the commit-queue script. If you'd like to try your hand at running the script, I've attached it to this email. I need to re-write it in python so we can check it into the WebKitTools/Scripts directory. The commit-queue isn't polished yet, so you should be prepared to do a little hand holding with the script. The three main issues to be aware of are as follows: 1) The script takes over the working copy to apply patches, build, and test. I've been running it from a dedicated working copy so I can be productive while it runs. 2) The script is tries to be conservative in actually committing to the repository. That means if someone else commits to the same top-level directory during the build / test cycle, the script won't actually pull the commit trigger. It will just try to land the patch again in the next cycle. We might want to experiment with being slightly more aggressive here. 3) If a patch is bad (doesn't apply cleanly, doesn't build, doesn't pass all the tests), the script won't r- the patch automatically. It will just diligently try to land the patch each cycle. You should pay some attention when a patch fails and update the bug appropriately. One clear next step is to update the tool to set the commit-queue- flag when a patch fails. Other than that, you just need to watch the builtbots to make sure you didn't destroy the tree. :) Adam P.S., Protip: If you want to see which bugs are in the queue, you can run ./WebKitTools/Scripts/bugzilla-tool bugs-to-commit or use this (slightly less precise) Bugzilla query: https://bugs.webkit.org/buglist.cgi?query_format=advanced&short_desc_type=n…

4 7

Hi All (Windows WebKit through COM, focus issue)
by Mike 12 Aug '09

12 Aug '09

I just joined this list, so hopefully I'm not intruding... I recently started work on making a lightweight browser for WebKit on Windows through COM+.NET, which actually has been a great success. I'm happy to report that I'm able to use the latest nightly build (r39553) from within a .NET Windows Form. I don't mind publishing the wrapper in the future to help further the development of WebKit, but I'm having an issue which is preventing me from moving forward... To sum it up the best I can, if I have a WebView on a form, and I call "setHostWindow" with that form's handle, I experience no behavioral issues. If I create a panel on the form, and I use that as the host handle, the WebView works as designed. But, if I create a tab control and add multiple tab pages, and use panels on those tabs to host the control, when the form is active, the WebView does not act as if it is active (it doesn't display the cursor for text boxes and other fields, and the safari theme for the scroll bar does not light up). The first tab page works properly, but the other tab pages display this behavior. Example of appropriate behavior on the first tab... form is active, control shows proper focus http://img49.imageshack.us/img49/9758/example1rc5.jpg The second tab, when it's active (and the form is active), the WebView control doesn't show proper focus, and the text box does not show the focus ring correctly either... http://img530.imageshack.us/img530/4653/example2tg6.jpg I think the easiest way to fix this problem, would be to allow manual control of the "active state" of the WebView control through the COM interface. Manually setting the state through events that fire on the Windows forms would be an incredibly easy task. But, I'm not sure how difficult it is to change the base code (if there's a lot of hoops to jump through), or whether or not it's actually appropriate to solve this problem in that manner... Either way, I would appreciate it a lot if someone got back to me. I've been trying to figure out how to get around this issue for the last couple days (even tried manually sending signals to the control via the windows API), but I haven't been able to get anything working yet. Thanks for reading, Mike

3 3

Building Win32/Release WebKit: linker error LNK1106
by yypprr＠sohu.com 11 Aug '09

11 Aug '09

Hi, All: I am a newer of WebKit. and when i try to build webkit on windows, i got this following error message when build the webcore project: Performing Pre-Link Event...Creating library...D:\Customer\Webkit\lib\WebCore.lib : fatal error LNK1106: invalid file or disk full: cannot seek to 0x2003875BProject : warning PRJ0018 : The following environment variables were not found:$(PRODUCTION) When I monitored the webcore.lib in the disk space, when "Creating library..." is in progress, WebKitBuild/lib/WebCore.lib is created and rapidly grows from a couple of megs to 128M, then to 256M, and at last to 512M, 1G. After LNK1106 error is displayed, the file is removed. Does anyone met the similar issue? Will be appreciated for any comments. Thanks Peng

4 7

Re: [webkit-dev] JSC: Callback functions returning exception, what should be the JSValueRef return?
by Brian Barnes 11 Aug '09

11 Aug '09

Darin Adler wrote: > On Aug 11, 2009, at 12:15 PM, Brian Barnes wrote: > >> Oliver says when I return with an exception, line and sourceURL, etc, >> properties will be added to it, and then I can use that later to pick >> up the line number. > > That’s right. > >> You now say this doesn't seem happen, which means it would be >> impossible for me to pickup the line number (as there's no API for it.) > > What I meant to say is that this will only happen if the exception is > an object. > >> Or is it a third item, which is I need to create from a specific >> prototype TO get the line numbers added? > > It just needs to be an object which can have properties set on it, as > opposed to another kind of JavaScript value that can’t. > > I prefer to answer questions on a public mailing list where everyone > else can see the answers. > > At this point you should try some things. I think you know enough > about how this should work in theory, and you can let us know if > things don’t work in practice. > > -- Darin Thank you, and I completely apologize for replying privately, I hit the "reply" instead of "reply all" and didn't notice this. I've bounced the reply to this list. [>] Brian

1 0

JSC: Callback functions returning exception, what should be the JSValueRef return?
by Brian Barnes 11 Aug '09

11 Aug '09

When I define a callback function (JSObjectCallAsFunctionCallback), and I have it thrown an exception, what should be the return value? The documentation says: "A JSObject that is the constructor's return value" ... but the other side -- JSObjectCallAsFunction -- states: "The JSValue that results from calling object as a function, or NULL if an exception is thrown or object is not a function." So should I also be returning NULL from my callback functions if I'm throwing an exception? If so, you should probably improve the documentation there. [>] Brian

3 8