1093327 - Incorrect orient="auto" behavior for SVG <polyline> and <polygon> markers

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Daniel Trebbien]

Test case: http://jsfiddle.net/agxewL2g/

In Firefox 33.0.2 on Mac OS 10.10 and Windows 8.1 Pro, as well as Firefox Nightly 36.0a1 (2014-11-03) on Mac, the auto-orientation of <polyline> start markers and <polygon> start & end markers is not correct.

The specification of orient 'auto' has changed slightly between SVG 1.1 and the latest draft from October 10, 2014 of SVG 2:
http://www.w3.org/TR/SVG11/painting.html#OrientAttribute
https://svgwg.org/svg2-draft/painting.html#OrientAttribute

However, Firefox' behavior is incorrect according to either wording.

If I have done my math right, here is an updated JSFiddle showing the tangent vectors and angle bisector for the two <polygon>s:
http://jsfiddle.net/agxewL2g/1/
According to SVG 1.1, the X axis of the arrow head marker should be aligned with the light red angle bisector vectors.

Results for other browsers:
- IE9 and IE11 correctly implement the wording of SVG 1.1
- Safari 8.0 has no problem with the <polyline> markers. For <polygon>, Safari appears to be implementing the SVG 2 wording; however, Safari is doing something weird with the end marker.
- Chrome 38.0.2125.111 / Blink 537.36 (@183466) also has no problem with the <polyline> markers. For <polygon>, Chrome appears to be implementing the SVG 2 wording; however, Chrome aligns the end marker X axis with the direction out of the initial/final vertex when it should align the end marker "in the direction of the path", which would be the direction into the vertex, I think.

Comment 1

[Daniel Trebbien]

Attachment: Firefox Nightly 36.0a1 (2014-11-03) Screen Shot.png

Comment 2

[Daniel Trebbien]

Attachment: Test case

Comment 3

[Daniel Trebbien]

Opera 10.63, 11.11, 11.51, 11.62, 12.10, and 12.16 correctly implement the wording of SVG 1.1.

Comment 4

[Robert Longson [:longsonr]]

Daniel, have you tried older versions of Firefox? Did this work in some previous version? If so would you be willing to use http://mozilla.github.io/mozregression/ to find out when things broke?

Comment 5

[Daniel Trebbien]

Hi Robert,

Using BrowserStack, I found that Firefox 4 through 24 correctly implement the wording of SVG 1.1, like IE and Opera (Presto).  Things broke in Firefox 25.

Sure, I can use mozregression.  I will let you know what I find...

Comment 6

[Daniel Trebbien]

Last good revision: b44898282f21 (2013-07-12)
First bad revision: 32c3ccd8946c (2013-07-13)
Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=b44898282f21&tochange=32c3ccd8946c


Maybe the regression was introduced by the patches for Bug 879659?

Comment 7

[Robert Longson [:longsonr]]

Attachment: marker.txt

Thanks for the testcase and regression range. Very helpful.

Comment 8

[Robert Longson [:longsonr]]

https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=69a9d71cdd90

Comment 9

[Robert Longson [:longsonr]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/96c0b7cb6e93

Comment 10

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/96c0b7cb6e93