If an mfenced element contains line breaks, the delimiters and separators are lost

[dpvc]

If an <mfenced> element contains a line break, the delimiters and separators are lost. Also, the separators and delimiters are not used as possible breakpoints.

[fred-wang]

These != reftests verify that the delimiters and separators are not lost:

MathMLToDisplay/Presentation/GeneralLayout/mfenced/issue255-1a.html
MathMLToDisplay/Presentation/GeneralLayout/mfenced/issue255-1b.html
MathMLToDisplay/Presentation/GeneralLayout/mfenced/issue255-1c.html
MathMLToDisplay/Presentation/GeneralLayout/mfenced/issue255-1d.html

This != reftest verifies that delimiters or separators are used as possible breakpoints:

MathMLToDisplay/Presentation/GeneralLayout/mfenced/issue255-2.html

=> In testsuite (see my MathJax-test fork)

[dpvc]

The issue255 branch of my fork of MathJax should handle your first four test cases. I don't yet have the delimiters and separators being valid breakpoints, so more work needs to be done, but that may not happen in this release, so I've marked it as ready for release plus to be addressed later.

[fred-wang]

OK, I verified the first four tests in Chrome, Opera and Firefox.

The fifth test case seems to pass for me too, although the break always happen after a demiliter/separator not after the <mspace>'s of the test page. Perhaps that's what is supposed to be fixed?

[fred-wang]

=> Ready for Release

[dpvc]

The current implementation of linebreaking in mfenced elements doesn't consider the delimiters or separators as potential breakpoints (that needs to be fixed). So any breaks occur because of the children in the mfenced. Since mspace is a valid breakpoint, you can still break before that, and so that is the linebreaking you are seeing in your test. Had you used something like <mi>x</mi> instead of mspace's, you would not have gotten any breaking.

I'm wondering if a better test for this would be to compare the <mfenced> to the corresponding <mrow> with <mo> elements in the right places. E.g., compare

<math>
  <mfenced open="(" separators="+" close=")">
    <!-- ( -->
    <mi>x</mi>
    <!-- + -->
    <mi>x</mi>
    <!-- + -->
    <mi>x</mi>
    <!-- + -->
    <mi>x</mi>
    <!-- + -->
    <mi>x</mi>
    <!-- + -->
    <mi>x</mi>
    <!-- ) -->
  </mfenced>
</math>

to

<math>
  <mrow>
    <mo>(</mo>
    <mi>x</mi>
    <mo>+</mo>
    <mi>x</mi>
    <mo>+</mo>
    <mi>x</mi>
    <mo>+</mo>
    <mi>x</mi>
    <mo>+</mo>
    <mi>x</mi>
    <mo>+</mo>
    <mi>x</mi>
    <mo>)</mo>
  </mrow>
</math>

which should be rendered the same.

[dpvc]

OK, I've added code to allow line breaks at the delimiters and separators (I thought of a straight-forward way to do it so decided to go ahead with it now rather than wait until later).

[fred-wang]

OK, I've added your testcase and the files seem to match by eye.

MathMLToDisplay/Presentation/GeneralLayout/mfenced/issue255-3.html

I've also modified issue255-2.html to use mpadded instead of mspace.

I will upload these tests after my next commit.

=> Ready for release.

[dpvc]

Note that my second code block above had a </mfenced> that should have been an </mrow>. I've edited it here, but if you just copy and pasted, you will want to make the change. Sorry about that.

I've also updated issue255 to fix a problem that caused lines to be too short (due to edits I made in the earlier commit, which I have rolled back).

[fred-wang]

Note that my second code block above had an </mfenced> that should have been an </mrow>. I've edited it here, but if you just copy and pasted, you will want to make the change. Sorry about that.

Yes, don't worry I noticed that when I asked emacs to indent the code.

I've also updates issue255 to fix a problem that caused lines to be too short (due to edits I made in the earlier commit, which I have rolled back).

OK. I think you can merge it to develop, anyway.