vector features aren't rendered in IE

[tschaub]

To see the problem in action:

• Open the spherical-mercator.html example in IE6.
• Draw a polygon (see blue sketch with draw with temporary intent)
• Finish drawing a feature and see it disappear.

[ahocevar]

Same in IE7 and IE8.

[ahocevar]

The attached patch fixes the reported issue, and brings a couple of other improvements:

• Aet style.visibility to "visible" in postDraw(). This fixes the reported issue.
• Always start with an initial origin of (0,0) and use offset. This reduces the chance that we run out of significant digits for large coordinate figures and fixes the  snap-split.html example.
• Use toFixed() everywhere (and not just in some places, as it was before) for values that we use as coordinates. This just makes things more consistent.

Tests needed some modifications because we now need to mock up an offset, which was set in initialize() before, but is set in setExtent() now, which we don't call for the draw* tests.

Tests pass and both the spherical-mercator and the snap-split examples work in IE7 and IE8. Please review.

[tschaub]

Nice work figuring this out.

One minor comment from the optimization department: for truncating positive numbers to integers, (num | 0) runs a good bit faster than (num.toFixed()) (5x faster on FF, 3x faster on IE8, 2x faster on IE6).

[tschaub]

Replying to tschaub:

Nice work figuring this out. One minor comment from the optimization department: for truncating positive numbers to integers, (num | 0) runs a good bit faster than (num.toFixed()) (5x faster on FF, 3x faster on IE8, 2x faster on IE6).

Meant "truncating numbers" instead of "truncating positive numbers".

[tschaub]

I also notice  one place where node.style.top and node.style.left are set without appending "px".

I'm pretty sure this isn't intentional. If not, it seems like the fix fits with the others above.

[tschaub]

vml fixes

[tschaub]

Above patch includes the changes mentioned (using bitwise truncation and appending "px" to style). Examples are looking good. Will verify tests pass shortly.

[ahocevar]

Thanks tschaub for the bitwise truncation hint. And the missing "px" was indeed unintentional, and fits in well.

[tschaub]

All looks good. I verified that tests pass and examples work in IE6 and IE8.

Thanks for tracking down the issues. These are some nice fixes. Please commit.

[ahocevar]

fixed a typo (should read + "px" instead of | "px")

[ahocevar]

(In [10039]) Fix initial offset and visibility issues in VML renderer; also make sure that all floats are converted to integers (previously using toFixed(), now using (num | 0)). p=tschaub,me, r=tschaub,me (closes #2465)

[ahocevar]

fillColor likes to be set before visibility

[ahocevar]

[10039] introduced a regression (see http://openlayers.org/pipermail/users/2010-February/016141.html), which is fixed by the above patch.

Examples mentioned in this thread still work, tests still pass. Please review.

[crschmidt]

Looks good, tests pass, please commit.

[ahocevar]

(In [10049]) fixed regression with broken polygon fills. r=crschmidt (closes #2465)

[bartvde]

I've got a good suspicion that this ticket caused the !Layer.Vector tests to fail in IE (I've tested with version 6), see also: ticket:2538

Tim, Andreas, are you able to look into this?

[bartvde]

Ok, this fixes is partly, however one value is now -16px instead of -17px but this could be becausing of rounding issues.

Index: tests/Layer/Vector.html
===================================================================
--- tests/Layer/Vector.html     (revision 10119)
+++ tests/Layer/Vector.html     (working copy)
@@ -542,6 +542,7 @@
                         "graphicYOffset correctly set");
         }
         if (layer.renderer.CLASS_NAME == 'OpenLayers.Renderer.VML') {
+                layer.renderer.offset = {x: 0, y: 0};
                 feature.style = customStyle1;
                 layer.drawFeature(feature);
                 t.eq(root.firstChild.style.width,

[tschaub]

I don't see a patch to review here.

[tschaub]

For clarification, the Layer/Vector.html test passes at r10038 and not at r10039 in IE.

[tschaub]

My interpretation is that graphicXOffset and graphicYOffset are working in IE and that the tests are not good. I think a clearer test of symbolizer offsets would be to draw a graphic with zero offset, retrieve the position, draw the graphic with some known offset, and confirm that the difference in position is consistent with the provided offsets.

[bartvde]

I'll try and write a new test like Tim suggests tomorrow morning.

[bartvde]

update the IE test as Tim suggested

[bartvde]

I've updated the test, the test now passes in IE6. Please review.

[elemoine]

Looks good. Confirming that Layer.html now passes in IE8.

[bartvde]

(In [10141]) fix up Layer.Vector tests in IE which were broken as of r10039, r=elemoine (closes #2465)