83250 - gfx/gtk TryFamily() can pick up wrong fonts on Mozilla Mail

Status: VERIFIED FIXED

(Core :: Internationalization, defect)

Description

[Masaki Katakai]

I have mentioned about this problem before,
I didn't find the exact reproduce case at that time.

But I found the exact reprodue case that

1. specify -sun-gothic- font for japanese
2. open Mozilla mail
3. View japanese message
  -> daewoo-gothic-ksc5601.1987-0 is used

The daewoo-gothic-ksc5601.1987-0 font is for korean,
not for japanese.

It seems that TryFamily("gothic") is called on
Mozilla mail and TryFamily() does not care
font language (we haven't put fixes into TryFamily()).

I'm not sure why  TryFamily("gothic") is called
only on Mozilla Mail. Not sure it is correct
behavior or not.

Second, should we consider language also in  TryFamily()?

FindFont(a/0x0061), nsFontMetricsGTK.cpp 3630
    FindStyleSheetSpecificFont, nsFontMetricsGTK.cpp 3212
        familyName = gothic, nsFontMetricsGTK.cpp 3223
        TryFamily daewoo-gothic-ksc5601.1987-0, nsFontMetricsGTK.cpp 3170
        load font daewoo-gothic-ksc5601.1987-0, nsFontMetricsGTK.cpp 2675
        TryFamily dt-gothic-jisx0208.1983-0, nsFontMetricsGTK.cpp 3170
        load font dt-gothic-jisx0208.1983-0, nsFontMetricsGTK.cpp 2675
        TryFamily dt-gothic-jisx0201.1976-0, nsFontMetricsGTK.cpp 3170
        load font dt-gothic-jisx0201.1976-0, nsFontMetricsGTK.cpp 2675

Comment 1

[kill this account]

Technically the CSS spec does not mention encoding but clearly this is an
important element to consider.

It seems reasonable to have TryFamily consider font lang groups.

Katakai, could you work on this?

Comment 2

[Masaki Katakai]

OK, I'll take this.

Comment 3

[Andreas Becker]

Changing QA contact to [e-mail] for now.

Comment 4

[Masaki Katakai]

Attachment: proposed patch

Comment 5

[Masaki Katakai]

I've attached the patch. Brian, can you check?

In TryFamily(), I just call TryLangGroug() with "*-familyname-*-*"
pattern, which will be expanted to "-*-familyname-<langgroup's registry>
-<langgroup's encoding>".

gFamilies is still needed in nsFontMetricsGTK::FamilyExists(), which
is called by upper side and it's static. There is no way to get
language group here, so I did't make any changes here.

nsFontMetricsGTK::FamilyExists() still searches as "*-familyname-*-*"
pattern (without language registry and encoding) but it would be OK
because nsFontMetricsGTK::FamilyExists() returning TRUE means
the family may exist for the language group.

In nsFontMetricsGTK::TryFamily(), the followng check would not
be necessary. What do you think?

  nsFontFamily* family = FindFamily(aName);
  if (family) {

Comment 6

[kill this account]

the current code calls TryFamily() which a name like "times new roman"
(FindStyleSheetSpecificFont) so it does not have any info that would help
determine a lang group. Using the doc's/elements's lang group, mLangGroup,
is good.

If this search fails should the code search for a font of this name
regardless of lang group?

Comment 7

[Masaki Katakai]

Attachment: how about this one?

Comment 8

[Masaki Katakai]

Thanks for comments, Brian.

Yes, I think we should do that because family
name can be specified by user and user expect
different font of charset encoding.

1. try family name with doc's/elements's lang group
2. if fails, search the family name without lang
   group

I've attached the patch. Can you check?

Comment 9

[kill this account]

r=[e-mail]

Comment 10

[Masaki Katakai]

Blizzard,

Requesting sr= for trunk.

Please refer the last my comment of 2001-06-19 20:26 how we
change the TryFamily() with considering language group as
top priority.

Comment 11

[Christopher Blizzard (:blizzard)]

sr=blizzard

Comment 12

[Masaki Katakai]

patch checked in

Comment 13

[Masaki Katakai]

Verified on 2002011614 trunk build.