374442 - marker-end on <path> doesn't work when set by script

Status: RESOLVED WORKSFORME

(Core :: SVG, defect)

Description

[Declan Naughton]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a3pre) Gecko/20070312 Minefield/3.0a3pre
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a3pre) Gecko/20070312 Minefield/3.0a3pre

. See test case.

Reproducible: Always

Steps to Reproduce:
1.
2.
3.

Comment 1

[Declan Naughton]

Attachment: test case

Comment 3

[Jesse Ruderman]

Attachment: fixed testcase

The original testcase tries to reference the element 'p' before the element appears.  With incremental XML parsing, that obviously won't work.

Comment 4

[Jesse Ruderman]

Reopening from dup because bug 372572 WFM, but I can reproduce this bug (on Mac).

Comment 5

[Robert Longson [:longsonr]]

Use setAttributeNS see http://jwatt.org/svg/authoring/ 

Comment 6

[Jesse Ruderman]

The namespace set by setAttribute(), null, is correct.  Twiddling the attribute in DOM Inspector makes the arrow appear, so this is clearly a bug in Gecko's SVG handling, not a bug in the testcase.

Comment 7

[Jonathan Watt [:jwatt]]

While using the NS versions of the element functions is definitely necessary, using the NS versions of the attribute functions hasn't been (only for attributes in the null namespace of course). In fact I think it's unlikely we could ever now require the NS methods for attributes in the null namespace (even if we wanted to) given that we've supported the NS-less methods for so long, so I should probably change that page.

Anyway, as Jesse says, the use of setAttribute instead of setAttributeNS doesn't seem to be the problem here.

Comment 8

[Robert Longson [:longsonr]]

Attachment: working version of testcase

In Jesse's testcase the style system tells us that there is no marker-end attribute on the path. In this testcase the style system tells us that there is an attribute. If there is a bug it must be in the style system.

Comment 9

[Robert Longson [:longsonr]]

Comment on attachment 285579 [details]
working version of testcase

Ahh I see why mine works. Save Page As stuck the missing attribute on it. Oops. Mea Culpa.

Comment 10

[Robert Longson [:longsonr]]

If you make a function out of the script and call it on load it works. Although we get a setAttribute call the style system seems not to be set up and returns that marker-end has no style so we don't draw it.

Comment 11

[John Resig]

This seems like a good candidate for 1.9.1 (an obvious bug that should be resolved).

Comment 12

[Martin Czuchra]

We're using such markers as arrowheads in dynamic workflow visualizations. This bug really is a pain for our application, as we can't fall back to the initialization phase. A fix would be really nice. Meanwhile we're trying to work around this somehow.

Comment 13

[Takeshi Kurosawa :takenspc]

Worksforme on latest mozilla-central nightly build.
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.1a2pre) Gecko/2008072902 Minefield/3.1a2pre

-> WORKSFORME