468218 - implement @font-face { src:local() } on Linux

Status: RESOLVED FIXED

(Core :: Graphics, defect, P4)

Comment 1

[Karl Tomlinson (:karlt)]

I guess this blocks 1.9.1 for platform parity reasons, at least.

Comment 2

[Karl Tomlinson (:karlt)]

Attachment: patch

I'll do a little more testing and see if I can put together some reftests.

This patch also changes the creation of patterns for gfxDownloadedFcFontEntry from lazy to immediate.  Fonts are only downloaded when they are expected to be used, so lazy creation of patterns was an unnecessary complication.

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(In reply to comment #2)
> I'll do a little more testing and see if I can put together some reftests.

That sounds a little challenging, but I can imagine you might be able to put something together using nsIFontEnumerator:
http://mxr.mozilla.org/mozilla-central/source/gfx/idl/nsIFontEnumerator.idl
You might actually want to combine that with the reftest-inside-mochitest stuff that bz put together:
http://mxr.mozilla.org/mozilla-central/search?string=WindowSnapshot
(That would let you find a 'font-family: foo' that causes the text to appear different from 'font-family: bar', and then check that you get the expected results with src:local(foo) and src:local(bar) in various combinations, including, perhaps, fun with other descriptors.)

But if you had other ideas for how to test, don't let this one stop you...

Comment 4

[Karl Tomlinson (:karlt)]

Attachment: reftest

Thanks for your suggestions, David.  My plans weren't quite that fancy.

This just tries a set of fonts, at least one of which is expected to be present on almost every system.

Comment 5

[Karl Tomlinson (:karlt)]

Attachment: patch v1.0.1

Corrected a wrong variable name being used by non-OpenType fonts (detected by the reftest).

Comment 6

[Karl Tomlinson (:karlt)]

Attachment: patch v1.0.2

correction to the wrong variable correction

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

For what it's worth, there's an existing test that does the font-enumerator half of comment 3, in layout/base/tests/test_bug394057.html .

Comment 8

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 354480 [details] [diff] [review]
patch v1.0.2

+    if (mPatterns.Length() != 0) {
         // Remove back reference to this font entry and the face in case
         // anyone holds a reference to the pattern.

Shouldn't we assert here that mPatterns.Length() == 1? It seems to me that a gfxDownloadedFcFontEntry never has more than one pattern, is that true? If so, should be documented.

Otherwise the patch looks good.

Comment 9

[Karl Tomlinson (:karlt)]

Attachment: patch v1.1

(In reply to comment #8)
> Shouldn't we assert here that mPatterns.Length() == 1? It seems to me that a
> gfxDownloadedFcFontEntry never has more than one pattern, is that true? If so,
> should be documented.

Yes, that's right.  Added assertion, as well as a comment in InitPattern().

Comment 10

[Karl Tomlinson (:karlt)]

Fixed on m-c:
http://hg.mozilla.org/mozilla-central/rev/b66168eafccb

In testsuite:
http://hg.mozilla.org/mozilla-central/rev/073903c476f3

Comment 11

[Karl Tomlinson (:karlt)]

Fixed on 1.9.1:
http://hg.mozilla.org/releases/mozilla-1.9.1/rev/3fac3f86737c
http://hg.mozilla.org/releases/mozilla-1.9.1/rev/a9818a7bd1f3

Comment 12

[Karl Tomlinson (:karlt)]

Forgot some new files in the reftest on 1.9.1:
http://hg.mozilla.org/releases/mozilla-1.9.1/rev/9152e412a70f