471711 - reftest bidi-006*.html fails. We appear to be messing with the width of a <p> element

Status: RESOLVED INACTIVE

(Core :: Layout: Text and Fonts, defect)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

There are two tests which fail(bidi-006.html and bidi-006-j.html).  These are displaying a red box to the right of the area where it is supposed to be displaying a simple box surrounding some text.

These tests only fail on maemo, and not the linux desktop version of fennec.  These tests were created as a result of bug 387653.

I suspect we are interpreting the width incorrectly in the css where we set:
/* borders should be covered by overhang */
p#test {
  border-left: 3em solid red;
  border-right: 3em solid red;
  width: -moz-max-content;
  width: intrinsic;
}

What happens is this html displays a red box for P to the right of the area:
  <p id="test">&#x5e0;&#x5b8;</p>
  <p id="overhang"><span></span>&#x5e0;<span></span></p>


You can find these tests in layout/reftests/bidi/bidi-006*.html

Comment 1

[Jonathan Kew [:jfkthame]]

If the device doesn't have a font that provides the Hebrew characters needed, we'll end up drawing missing-glyph boxes instead, and I wouldn't expect the diacritic to be handled correctly -- we don't know how to draw a "zero-width diacritic missing glyph". That may well explain the failure here.

See also bug 471714.

Comment 2

[Joel Maher ( :jmaher ) (UTC -8)]

moving to test component

Comment 3

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Attaching the test output (which includes data URLs of the images) would aid diagnosis.

Comment 4

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: clip from reftest log with the bidi-006* failures

This should give the failed and reference images for the bidi-006 tests.  I took this from the log files where it exibits the problem.  If there is another format that would be helpful I am happy to get the data.

Comment 5

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Yeah... so we do in fact show missing glyph boxes, and the characters do not combine, which means that (a) the border sticks out because the character with the combining mark is wider than without and (b) the reference has the combining mark missing glyph, which is completely covered up by the span in the test.

So, yes, this failure does seem to be due to missing fonts.

Comment 6

[Joel Maher ( :jmaher ) (UTC -8)]

installing the hebrew fonts for the nokia maemo device resolved the problem.  If we don't see a problem with displaying the missing glyph boxes, then we can resolve this issue and I will update the requirements for reftests on maemo to add hebrew fonts.

Comment 7

[Joel Maher ( :jmaher ) (UTC -8)]

aki: this is related to a handful of failures found in the reftests.  Installing hebrew fonts on the device (a one time setup) allows these tests to pass.  Is that something we can do to the tinderbox maemo devices?

Comment 8

[Aki Sasaki (not active)]

I can do that.
Which package is it?

Comment 9

[Aki Sasaki (not active)]

Installed on maemo-n810-01 through 13.