Re: 2.3.1; traveling; request for bug help

[Owen Taylor <otaylor redhat com>]

Thanks a lot Noah for doing this. It's really helpful! If I was
going to make one quibble though, it might be a little nicer if it was 
split up by area so that I could go through it a bit more
efficiently :-)

I've marked through possible milestone assignments below in square
brackets. L/H are Low/High priorities. I don't think it would hurt
to create the 1.6 API Freeze and 1.6.0 milestones now, though we
need to make a pass through all the future bugs before we start
the 1.6 cycle to sort out other bugs that belong there.

Regards,
						Owen

> 1.2.5 milestone
> ===============
> ? 71033 Scale attr needs to ensure that font size > 0
>   appears to be waiting on information from the reporter (Havoc); not
>  touched in ages
I wouldn't really say that it is waiting for info from Havoc,
my comment was really more of a "how the heck should this work".
I think it would be better to simply make size-0 fonts work.

Retitled, moved to 1.4.0, priority low. [1.4.0L]

> + 73119 The characters gets overlapped in GtkEntry widget.
>   already fixed for all backends except win32, patch needed

Can be left on 1.4.0, but it really is going to depend on
when a Win32 person gets to it.  [1.4.0]

> + 96314 hangu-x shaper : needs Jamo normalization
>   added comment that this may be a dup of a glib bug, 100456

It might be best to retitle to:
"make hangul shaper use GLib normalization functions" and
mark as blocking on the two glib bugs,but I don't think it's really a
duplicate of anything, unless maybe 95708 (since the hangul-x shaper
is gone, and the bug is about factoring code out and sharing it.)
[1.6.0]

> + 110521 can't find bold and/or italic fonts on win32
>   patch and decent amount of discussion, plus it sounds important
Hmm, patch doesn't really correspond to my suggestion, but 
I'll let Tor make the final call on this.
[1.4.0]

> + 118108 indic-ot: wrong ordering for kannada length marks
>   patch, not reviewed
> 
> + 118299 Treat BENGALI LETTER A as consonant
>   dependency of 113551 which looks promising
> 
> + 118301 ba + halant applied to wrong character
>   dependency of 113551 which looks promising
Waiting for an Indic expert, or someone who wants to:

A) Become an Indic expert
B) Take responsibility for feeding changes that we make to the
    Indic code back to the ICU project.

[1.4.0]

> + 118592 Length overruns when skipping glyphs
>   patch that Owen has at least glanced at
I haven't actually glanced at the latest version. It's fairly
straightforward stuff, but a lot of things to review. And I'd
really like to see an extension of to GPOS as well.

[1.4.0]

> + 120195 there are some functions in pangowin32-fontmap.c that might maybe
>   should be static
>   no patch, but should be very easy

Yep. [1.4.0]

> * 121543 Scale attribute semantics
>   patch by Owen tested and confirmed to work

Added an "OK to commit comment" [1.2.5H]

> 1.4 API milestone
> =================
> - 59071 Ellipsization support for PangoLayout
>   no patch; Owen estimates "a week or so" of work
Definitely a punt. [1.6 API Freeze]

> - 64538 pango_layout_set_justify() doesn't work
>   doesn't look like this will be finished in time; Damon Chaplin was
>   working on it, but announced some time ago that it wouldn't be ready for
>   1.4 IIRC
Probably needs to be punted. Sigh. [1.6 API Freeze]

> * 70451 Automatic paragraph direction according to Unicode BiDi algorithm
>   4th iteration patch, must be almost ready to commit

I have a bunch of stuff sitting around, need to finish this up. [1.4 API
Freeze]

> ? 99235 API for getting available sizes for a face
>   no patch, but may be doable in the time frame
Certainly not that much work, the main thing you'd have to do is
figure out how the API reports "font is scalable and can occur
at any size" [1.4 API FreezeL]

> ? 101380 API to find find weak cursor at given X location
>   no patch yet; no idea how hard this one is
I don't think it's a lot of lines or hard, but it does require
some understanding of rather internal stuff. Basically what would be
involved is taking:

pango_layout_line_x_to_index

moving it to a private copy, and adding conditionalization for
weak/strong. [1.4 API Freeze]

> - 101406 Shape marks with base characters
>   no patch yet; comments suggest non-triviality
Definitely non-trivial. But I'm not quite ready to punt, since 
it will involve rebreaking semi-public API that we are already
breaking in 1.4 to fix. [1.4 API Freeze]

> * 105553 There should be a way to disable fallback.
>   3rd iteration patch, works well for me
Your patch, as far as I've looked at it, is a very good attempt
at implementing this API, but I'm really wondering if we are
going down a rat-hole here ... there's far too much complexity leaking
around.

The problem I think we are trying to solve here is how, when you
get a PangoFontFace, to load that font, and *only* that font; 
going to great lengths to put some reasonable interpretation
on "Serif" or "Helvetica,Arial" is not really necessary.

So, you probably should have just told me to go get lost on my
comments after the first patch :-); unfortunately, doing this
right has some major interaction with the current state of the
font/shaper selection code, which is entirely broken because
I forgot how fontconfig works when I wrote it. [1.4 API Freeze]

> + 106624 Sample string approach to approximate_char_width and CJK
>   this appears to be a small change
Think there are still an outstanding question here ... basically,
how does a program determine the grid size for a dual width
font. [1.4 API Freeze?]

> + 108728 Pango should offer a way to filter monospaced fonts
>   patch for fontconfig parts needs trivial polishing; 
>   win32 version required (tml?)
Added some comments. [1.4 API Freeze]

> ? 114237 Add squiggly-red underline type
>   no patch, but some mention of it, possibly even a patch, by Trever Adams
>   on gtk-devel-list
This was by Nicolas Setton; the Pango ptach looks good exceptfor
some minor nits. [1.4 APIfreeze]

> + 117282 Pango can't render Arabic accents
>   there is a mostly finished patch
Waiting on me to figure out how to handle bug 121060.[1.4 API Freeze]

> - 121060 Ligature Second Anchor Point in OpenType GPOS Table
>   no real patch, discussion looks open-ended [1.4 API Freeze]

I know how I want to handle this one now from an API interface and have
some sort of implementation sitting on my hard drive. [1.4 API Freeze]

> 1.4.0 milestone
> ===============
> - 57375 gtk_text_iter_starts/ends_sentence works wrong with '!', '?', '\n'
>   I marked this as a dup of 97545, but I'm not 100% sure that was right; if
>   not, please feel free to reopen it.
Good enough.

> - 61838 pango-ot, arabic-xft, markings shaping handling
>   there is a patch, but Owen suggests it needs reworking, and besides, it's
>   very old now; bug hasn't been touched in a year, is it still relevant?

Marked as a duplicate.

> ? 66626 Mishandling of indent
>   no patch, no idea how hard it is
Not really hard, but not entirely trivial. Probably easiest to add
some sort of flag to PangoLayoutLine indicating that this is the
first line of a paragraph. [1.4.0L]

> * 68435 Update to fribidi 0.10
>   patch; this is needed for high unicode planes support

Maybe you can get Behdad to take a look as a second pair of eyes? If
not, I'd just go ahead and commit. [1.4.0]

> - 71414 pango configure lacks --disable-modules
>   no patch, medium amount of work, comments say it's not that important
Afternoon of work. Can safely be punted. [1.6.0]

> ? 85745 Unicode LINE SEPARATOR character does not cause line breaks in
>   rendered text
>   not sure if this can be subsumed by 97545; if not I suspect it may not be
>   that hard to fix; if so, let's dup it
No relation to pango_break() - it's a separate issue. I'd guess it's
less than an afternoon  of work. [1.4.0L]

> ? 89449 Better handling of precomposed/combining forms in Hebrew shaper
>   no patch
I think this gets subsumed into the general problem of "how do we make
pango rendering independent of normalization form." [1.6.0]

> - 89541 Iteration by characters, clusters broken
>   no patch, only "Start of a test program"
This one is definitely a bit of work. But important. Probably should
be left on 1.4.0 for now. [1.4.0]

> - 95043 GTK doesn't deal well with arbitrarily named font styles
>   no patch; may require further fontconfig changes
May require emulating the correct behavior on top of fontconfig; I want
to keep the idea that all faces of a font are selectable along CSS3
lines. [1.4.0L]

> + 95708 enhancing Hangul shaper (Xft) with Oxxx/Nxxx fonts
>   many iterations of a patch and much discussion; however, it's a big patch
>   and no extensive review from Owen
Needs a reviewer, any reviewer. [1.4.0]

> - 96300 Hangul shaper needs a generic font-based render_func selection
>   mechanism as found in Thai shaper
>   no patch, Jungshik says "I suggest this bug has to be put off until..."
>   some other bugs are fixed
A WONTFIX, I think. 

> ? 97545 Make pango_default_break follow Unicode TR #29
>   new test program, but no patch; the function is rather scary, I doubt
>   this will make it in
Yeah, I think this is probably best punted for now, though it would
certainly be nice to find someone to take this one on. [1.6.0]

> - 98011 Use FontMap resolution in pango_ft2_font_get_face()?
>   no patch, appears to require further discussion
Low priority, can wait. [1.6.0]

> ? 98982 Optimization of Win32 pango rendering
>   patch, indecisive discussion, hasn't been touched for a year
I think this is probably a RESOLVED/INCOMPLETE. Though I'll leave it up
to Tor to make the call.

> - 100584 Improve FT_Face caching
>   no patch, appears to require further thought

Can be punted, yes. [1.6.0]

> ? 101079 opentype font suppor for diacritics for Latin/Greek/Cyrillic letters
>   there is a patch, not reviewed; output doesn't look great; might not be a
>   bug in the patch but elsewhere
Need to get it to the point where we are sure that it is doing good
before we commit. [1.6.0]

> * 101081 Non-BMP (plane 1 thru plane 16) characters are not supported
>   patches work for me
Added an OK to commit for the parts that are still relevant. [1.4.0H]

> * 101856 PangoFT2 is unable to use fonts that have no Unicode mapping table
>   patch from Sven which he claims is good to apply
Added an "OK to commit" [1.4.0H]

> + 103938 feature request: better handling of invalid combinations
>   patch that requires some minimal tweaking (it handles the 2-cell
>   combining characters wrong)
Added a query.

> + 104151 testfonts.c not in pango-1.2.0.tar.gz
>   no patch, but this looks like something not too difficult that should be
>   done for the next big release
Waiting on the Win32 people.

> - 104672 Optimization for non-centered single lines.
>   no patch, looks like more discussion is required, and it's "just" an
>   optimization ;)
Can be punted, priority is low. [1.6.0]

> ? 109900 pango_[ft2|win32|x]_render_layout_line implementations don't
>   handle all attributes
>   patch already committed for ft2, patch needed for win32

> ? 112467 Hangul opentype module : Un series fonts
>   only "v 0.1 patch (not working)"
Looks like something for later to me.

> - 113452 pango_xft_font_get_metrics() uses wrong screen
>   guessing that since Owen did only the "easy part", there remains a hard
>   part which hasn't been attempted yet
Yeah, if you read through it, there is a piece that blocks on 105553 -
your disable fallbacks bug. [1.6.0]

> + 113551 Bugs in the Bengali rendering system of Pango.
>   this is just a tracker bug, but the extensive comments and patches on
>   this bug suggest that most of the work has been done
I doubt we'll get this completely finished for 1.4. [1.4.0 for now]

> ? 114431 underlines in *_draw_layout_line()
>   no patch, but similar functionality already implemented in gdkpango.c
Probably will have to be punted. A few hours of work. [1.4.0]

> ? 118302 ZWJ and ZWNJ not passed to indic shaper
>   "The code is now in Pango to pass ZWJ to the Indic shaper... However,
>   ..., what the indic shaper does with the ZWJ is not quite correct."
>   no patch for the latter
Another bug waiting for an indic expert. Not big, but takes some
research and a bit of rewriting. [1.4.0]

> + 120353 Approximate digit width should be max of all digits
>   no patch yet, but sounds easy

Yeah,should be easy. [1.4.0]