130473 - font-size-adjust doesn't work on Linux and Mac

Status: RESOLVED FIXED

(Core Graveyard :: GFX, enhancement)

Description

[Kai Lahmann (is there, where MNG is)]

..just another missing CSS2-Feature

Comment 1

[Boris Zbarsky [:bzbarsky]]

Excuse me, but where is it missing exactly?

Comment 2

[Hixie (not reading bugmail)]

Actually, we support font-size-adjust on most platforms.

Since you didn't give a particular platform or a particular area of Mozilla
(e.g., layout, DOM), nor a testcase, I'm marking this WORKSFORME.

Please reopen it if you can clarify what you mean by "another missing
CSS2-Feature", preferably with a testcase showing the problem.

Comment 3

[Kai Lahmann (is there, where MNG is)]

eh.. here (linux, 0.9.9) it does exactly nothing...test: http://www.richinstype.com/test/fonts/fontsizeadjust.php

Comment 4

[Hixie (not reading bugmail)]

Ok, that would be because it only works on Windows. (Please don't mark a bug
All/All unless you know for a fact that it is cross-platform...)

This bug shall henceforth be known as the "font-size-adjust doesn't work on
Linux" bug.

BTW, the "rich in style" tests aren't exactly the best out there. Here are some
better font-size-adjust tests, if I dare say so myself:
   http://www.hixie.ch/tests/adhoc/css/fonts/size-adjust/

Comment 5

[Hixie (not reading bugmail)]

rbs: who should own this bug? What needs doing? Is it a duplicate?

Comment 6

[rbs]

shanjian, perhaps?

Reading the implementation I did in the Windows version while working on bug
74186 can be very instructive as to what is needed. Its supports touched several
font-related functions, but a nice side-result was more modularity in the
creation of font handles. For Linux, now that FreeType2 (and others) are being
added, the support will need to be propagated everywhere.

It is not a duplicate, I think.

Comment 7

[Roy Yokoyama]

->Linux i18n : shanjian

Comment 8

[Kai Lahmann (is there, where MNG is)]

1.2? This would mean some pages render compleatly different on 1.0/Windows and 1.0/Linux....

Comment 9

[Boris Zbarsky [:bzbarsky]]

Kai, they already do, since typical fonts installed are different, typical
resolutions users use are different, etc.

Comment 10

[Greg K.]

Reconfirmed using FizzillaCFM/2002071208. Setting All/All.

Comment 11

[Boris Zbarsky [:bzbarsky]]

Do we support this on MacOS 9?

Comment 12

[Madhur Bhatia]

does not seem to be supported on macos X, too.

Another testcase u can look at :
http://www.meyerweb.com/eric/css/tests/css2/sec15-02-04b.htm

Comment 13

[Felix Miata]

I went to URL in comment  4. In Linux 20020722 1.1b, all except 5 behave as
expected. 5 is all a bunch of overlapping mush. In contrast, with OS/2
2002082608 trunk, every test fails miserably. Both Linux and OS/2 handled the
test at URL in comment 12 just fine.

Comment 14

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

Attachment: testcase

Comment 15

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

In trunk builds, this works for me on Windows, Linux, and Mac (allowing for a little bit of variation on the third test -- although nowhere near as much as showed up before font-size-adjust was implemented).

Comment 16

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

Yes, without the relative size issue, the new gfx supports on the three platforms.

Comment 17

[Mats Palmgren (inactive)]

The second "Text" on the first line in the attached testcase has a slightly
larger font size than the other four.  Firefox 20070517 on Ubuntu edgy.

Comment 19

[Chris Lawrence]

I'm attaching three testcases that display some odd variations between font sizes under Linux at least.  The "correct" font-size-adjust value for Calibri appears to be around 0.472, but when set in trunk it makes the text too small; setting font-size-adjust to around 0.53 to 0.55 has about the same effect as font-size-adjust: none, but there are odd variations: with 0.53, the h1 text is right but body text is too small, while with 0.54 h1 is too big but body text is too small(!).   At 0.55 h1 is too big but body text is right.

The "correct" values are derived from this website and seem to comport with those in the CSS 2 and 3 specifications for Verdana etc.:
http://brunildo.org/test/xheight.pl?family=Calibri&size=100pt&string=xipbM%C8&famil2=DejaVu+Sans

These results are with Pango enabled; I will test without Pango enabled and see if the same problem appears.

Build ID: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9pre) Gecko/2008041610 Minefield/3.0pre

Comment 20

[Boris Zbarsky [:bzbarsky]]

Pango is always used on trunk, no?

Comment 21

[Chris Lawrence]

Attachment: Sample web page to show problems with font-size-adjust

All three blocks of text (each div containing identical h1 and p) should be the same size.

Comment 22

[Chris Lawrence]

Attachment: Second sample web page to show problems with font-size-adjust

This one uses Verdana instead of Calibri; the 1st and 2nd divs look about the same, but the third one ends up too small.  Part of the problem is that FF/Gecko may be adjusting based on the first font it actually finds rather than the first font in the list; that would at least explain the third box discrepancies observed in both tests.

Comment 23

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

Per http://www.w3.org/Style/css2-updates/REC-CSS2-19980512-errata.html#x21 font-size-adjust applies to the first choice font as well.  (This helps in a number of ways -- it prevents authors from accidentally specifying the wrong value because they have the first choice font and giving users who don't have it bad result, and for other reasons described in http://lists.w3.org/Archives/Public/www-style/1999Feb/0032.html .)  So I think the testcase should actually be smaller.

I think a good way to think about font-size-adjust is that it's a backwards-compatible way to specify the font size by specifying the size of the lowercase letters instead, using a multiplier for backwards-compatibility with browsers that don't support font-size-adjust but do support font-size.  This is how I described it in http://developer.mozilla.org/en/docs/CSS:font-size-adjust .

So I think displaying the third paragraph in that testcase smaller than the others is correct.  That said, I don't think it belongs in this bug, either, since I think this bug is fixed.

Comment 24

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

http://www.w3.org/TR/2002/WD-css3-fonts-20020802/#font-size-adjust has that change already applied.

Comment 25

[Chris Lawrence]

It would appear that this problem is a somewhat different (but related) bug.  font-size-adjust should definitely be applied to the "base" font too, but it appears its effect is off with X11 at least.  I'll put together a reduced test case and file a new report.

On my Mac running nightly trunk builds, incidentally, the testcase behavior appears to be correct (modulo a pixel or so, probably due to .01 disagreements between the right value for f-s-a under Linux and Macintosh due to the different font renderers), suggesting it's a problem with calculating the font's x-height (or some other dimension) under X11/Pango/something - and this propagates through to substitutions for other fonts too.

Comment 26

[Chris Lawrence]

New bug filed as bug 429605.

Comment 27

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

So I'm going to mark this fixed, since it was fixed by code that was added during the switch to cairo.

We should still have tests for it...

Comment 28

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

Attachment: reftests (work in progress)

Here's some work in progress for reftests.

They fail on Linux, though (or may be random) because we're rounding the x-height and the em-height to integers before doing the computation.  I'm not sure whether we do that on other platforms; it could mean that the tests would need to be marked as random across platforms.