Chromium

OS: Mac OS X 10.7
Browser: Google Chrome  18.0.1016.0 (Official Build 118622) canary

Findings: This is Mac specific issue. It's not seen on Win 7, Linux Ubuntu 10.04.

mike these look like CSS please unassign urself if they are not.

mike these look like css, please unassign urself if they are not.

Also occurs here in Chrome 17 OSX 10.6:
http://www.jethrolarson.com/projects/zenzoom/demo.html

Using the www.cuttherope.ie site on MacOS 10.6, I saw the following:

Chrome 18.0.1025.3 dev: BAD - custom cursor showed as system cursor, but did not disappear
Chrome 19.0.1035.0 canary: BAD - custom cursor disappears
WebKit Nightly r107100: OK - custom cursor displays normally.

So it's not a CSS bug, looks more like it's to do with rendering cursors at the system level.

Tested with 19.0.1035.0 (Official Build 120973) on Mac,Linux and Windows.
Link : http://www.cuttherope.ie/

Mac: The cursor disappears altogether and is visible only in certain conditions.
Linux : Except at links it is not showing the cursor as custom cursor.(after re-load the same page it is working fine)
Windows: Except at links it is not showing the cursor as custom cursor.(after re-load the same page it is working fine)


The same link in the FF is working fine in all the 3 OS.

We are seeing an issue that seems related, when hovering over GMaps on our site:

1. In Chrome 17, go to http://www.redfin.com/homes-for-sale#!lat=37.78008685844382&long=-122.50329423090898&market=sanfrancisco&sf=2&v=6&zoomLevel=12
2. Zoom in or out.
3. Is Hand pointer visible? If yes, repeat step #2.

Actual: Hand pointer vanishes until you click again on the map.
Expected: Hand pointer.

Using Chrome 17 on OS 10.7.3 I am able to reproduce this very easily, and we have had a number of people write in a bout it.

We have reports of this occuring on the circles page on Google+. We use CSS to install a custom cursor when dragging people tiles. b/5997726

Anyone know of a decent workaround for this?

Avi, you know about cursors, right?

[Empty comment from Monorail migration]

Yes, I am the Mac cursor man. I swear this used to work.

https://crbug.com/chromium/114517 might be related.

[Empty comment from Monorail migration]

[Empty comment from Monorail migration]

I'm sure this is almost no help, but I used bisect-builds to track the problem down to this webkit roll http://src.chromium.org/viewvc/chrome?view=rev&revision=108961

Trying revision 108963...
Revision 108963 is [(g)ood/(b)ad/(q)uit]: b
You are probably looking for build 108963.
WEBKIT CHANGELOG URL:
http://trac.webkit.org/log/trunk/?rev=99453&stop_rev=99399&verbose=on
CHANGELOG URL:
http://build.chromium.org/f/chromium/perf/dashboard/ui/changelog.html?url=/trunk/src&range=108959:108963
Built at revision:
http://src.chromium.org/viewvc/chrome?view=rev&revision=108963

My build kept churning out Chromium binaries that said "Aw Snap" on every page. But this will at least restrict the range to search:

Bisecting: 12 revisions left to test after this (roughly 4 steps)
[646f6f9101bbaf6621a30f124f5569378caca85c] Simplify NetscapePlugin::convertPoint and make it work (in theory) with transformed plug-ins https://bugs.webkit.org/show_bug.cgi?id=71699
dfab: ~/chromium/src/third_party/WebKit $ git bisect good
Bisecting: 6 revisions left to test after this (roughly 3 steps)
[aaf39526d806691892def48d6e42ef483be35a0b] WKPage.h should be in PrivateHeaders.
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 6 revisions left to test after this (roughly 3 steps)
[ff8909bb240060abbbff5d5f0b3698d8ec1d9249] REGRESSION(r99409): Broke transitions/clip-transition.html <http://webkit.org/b/71692>
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 6 revisions left to test after this (roughly 3 steps)
[74a032643aa7182c789a7757165b529da480fc0b] [Chromium] Rename WebView::scalePage to WebView::setPageScaleFactor to match WebCore https://bugs.webkit.org/show_bug.cgi?id=71485
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 6 revisions left to test after this (roughly 3 steps)
[e30fa9a1cc6f186e29be717fd4cbed0d9e9278c6] [EFL] Support requestAnimationFrame API https://bugs.webkit.org/show_bug.cgi?id=67112
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 6 revisions left to test after this (roughly 3 steps)
[5d124e298dd00480f754fa979f522c3ab0429e46] Rename NetscapePlugin::wantsWindowRelativeNPWindowCoordinates to NetscapePlugin::wantsPluginRelativeNPWindowCoordinates https://bugs.webkit.org/show_bug.cgi?id=71708
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 6 revisions left to test after this (roughly 3 steps)
[e86965e44677235cf4f15ac44777a7a355a74d6d] Unreviewed, rolling out r99432. http://trac.webkit.org/changeset/99432 https://bugs.webkit.org/show_bug.cgi?id=71709
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 6 revisions left to test after this (roughly 3 steps)
[da2af49abe35ff52f9a3226b066e39ad11fc254b] Need a way to allow a scheme access to Local Storage and Databases while Private Browsing is enabled. https://bugs.webkit.org/show_bug.cgi?id=71631
dfab: ~/chromium/src/third_party/WebKit $ git bisect good
Bisecting: 4 revisions left to test after this (roughly 3 steps)
[821ae106b6ef9aabb5072bab33c0f32d4abfb989] Don't use CSSOM's CSSValue.cssValueType internally in WebCore. <http://webkit.org/b/71679>
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 4 revisions left to test after this (roughly 3 steps)
[864847409166e09b7365076f4f6fd3b4cc4e41b2] NetscapePlugin::wantsWindowRelativeNPWindowCoordinates should return false on Mac https://bugs.webkit.org/show_bug.cgi?id=71707
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 4 revisions left to test after this (roughly 3 steps)
[a9738d8039fd9a7cbb4b730765fd0822a2d690f1] Add missing .in and .pl files to the Xcode project for easy access.
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
Bisecting: 4 revisions left to test after this (roughly 3 steps)
[949aea926a134358eed13dd90406c4f08276d497] [chromium] Mark perf/documnet-contians.html as slow.
dfab: ~/chromium/src/third_party/WebKit $ git bisect skip
There are only 'skip'ped commits left to test.
The first bad commit could be any of:
e86965e44677235cf4f15ac44777a7a355a74d6d
ff8909bb240060abbbff5d5f0b3698d8ec1d9249
864847409166e09b7365076f4f6fd3b4cc4e41b2
aaf39526d806691892def48d6e42ef483be35a0b
5d124e298dd00480f754fa979f522c3ab0429e46
e30fa9a1cc6f186e29be717fd4cbed0d9e9278c6
821ae106b6ef9aabb5072bab33c0f32d4abfb989
a9738d8039fd9a7cbb4b730765fd0822a2d690f1
74a032643aa7182c789a7757165b529da480fc0b
949aea926a134358eed13dd90406c4f08276d497
b3be099642b29e1e1a0a7aaa4da93b2b5a2355a3
We cannot bisect more!

To be clear, that's http://trac.webkit.org/log/trunk/?rev=99453&stop_rev=99447&verbose=on

Dan, yeah, there was a time during which renderers crashed upon loading. Sigh.

That's a pretty scary bisection result; none of those changes seem to be about cursor handling.

https://crbug.com/chromium/111027#c4: jethrolarson: it doesn't work on your website because your site is broken. In your http://www.jethrolarson.com/projects/zenzoom/zenzoom.css you specify:

.zoom{cursor:url(projects/zenzoom/zoomin.cur), url(zoomin.cur), pointer;}

which is a relative URL. That resolves to http://www.jethrolarson.com/projects/zenzoom/projects/zenzoom/zoomout.cur

which doesn't exist.

A simpler repro:

http://www.w3schools.com/cssref/playit.asp?filename=playcss_cursor&preval=url(smiley.gif),url(myBall.cur),auto

The problem is that it works... until it doesn't.

Different problem:

Note the smiley-face in http://www.w3schools.com/cssref/playit.asp?filename=playcss_cursor&preval=url(smiley.gif),url(myBall.cur),auto is blue. It should be yellow. When skia was switched on for the Mac, it looks like the cursor data being sent was changed. Dammit.

Splitting off from this as https://crbug.com/chromium/114598.

Re http://www.alessioatzeni.com/mac-osx-lion-css3/ : where are the CSS cursors?

Initial inklings are that for some reason, even though the CSS cursor is valid, we're getting empty cursor data passed in. Why? [shrug]

[Empty comment from Monorail migration]

OK... Cursors are coming out blank because a custom cursor is getting set that has no custom data. RenderWidgetHost::OnMsgSetCursor is getting a WebCursor of type 43 (custom) with no custom data or size. What? Still tracking.

RenderWidget::didChangeCursor is getting a WebCursorInfo with blank bitmap and hotspot (-1, -1). Grrr. What is going on here?

ChromeClientImpl::setCursor builds the WebCursorInfo from a WebCore::Cursor. That contains a WebCore::PlatformCursor, which has the image as a WebCore::Image.

In that image, m_data is nil. It looks like the image is completely uninitialized.

We're in EventHandler::handleMouseMoveEvent. Near the end:

    if (newSubframe) {
        [...]
    } else {
        [...]
        if (FrameView* view = m_frame->view()) {
            OptionalCursor optionalCursor = selectCursor(mev, scrollbar);
            if (optionalCursor.isCursorChange())
                view->setCursor(optionalCursor.cursor());
        }
    }

OptionalCursor is what you'd expect, a class with a bool and a Cursor.

It looks like EventHandler::selectCursor is returning a bogus cursor, but I can't see what's going on inside the current breakpoint. Setting a new one.

OK. This is what's going on.

Inside of EventHandler::selectCursor, it looks to see if the CSS style has a cursor set:

    if (style && style->cursors()) {
        const CursorList* cursors = style->cursors();
        for (unsigned i = 0; i < cursors->size(); ++i) {
            CachedImage* cimage = 0;
            StyleImage* image = (*cursors)[i].image();
            if (image && image->isCachedImage())
                cimage = static_cast<StyleCachedImage*>(image)->cachedImage();
            if (!cimage)
                continue;
            IntPoint hotSpot = (*cursors)[i].hotSpot();
            // Limit the size of cursors so that they cannot be used to cover UI elements in chrome.
            IntSize size = cimage->imageForRenderer(renderer)->size(); // ***1***
            if (size.width() > 128 || size.height() > 128)
                continue;
            if (!cimage->errorOccurred()) // ***2***
                return Cursor(cimage->imageForRenderer(renderer), hotSpot); // ***3***
        }
    }

Right now, I'm repro-ing by loading maps.google.com. There is one item in the |cursors| list. Indeed, the type is a CachedImage. Inside of CachedImage, errorOccurred() returns false, but there is no m_image. So Image::nullImage() is returned from CachedImage::imageForRenderer (line 1)

Its size is 0x0, so it passes the tests, and since no error occurred (line 2), a bogus cursor is created (line 3).

WebKit issue; creating bug to continue discussion.

Continuing discussion at https://bugs.webkit.org/show_bug.cgi?id=78834 .

[Empty comment from Monorail migration]

I'm still waiting for my build to finish, but I think japhet has posted a fix for CachedResource.cpp that may fix this problem too. See https://crbug.com/chromium/111218

I patched that in, and I'm not seeing this fixed :( Verifying further...

My build has finished; avi is right. japhet's patch doesn't fix this bug.

japhet's new patch, though, does. Yay!

Here's a build containing japhet's new sci.txt patch from https://crbug.com/chromium/111218 http://dl.dropbox.com/u/8497522/chrome-mac-2.zip

This should be fixed with the patch in WebKit r108100. Waiting for a roll.

WebKit http://trac.webkit.org/changeset/108100

[Empty comment from Monorail migration]

What's the timeline on this getting resolved in stable? It's effecting ALL maps API sites at this all point and we've gotten a number of reports about it.

This issue has been closed for some time. No one will pay attention to new comments.
If you are seeing this bug or have new data, please click New Issue to start a new bug.

[Empty comment from Monorail migration]

[Empty comment from Monorail migration]

[Empty comment from Monorail migration]

This issue was migrated from crbug.com/chromium/111027?no_tracker_redirect=1

[Monorail mergedwith: crbug.com/chromium/113358]
[Monorail mergedinto: crbug.com/chromium/111218]
[Monorail components added to Component Tags custom field.]