90385 - Linux: how to print using FreeType

Status: VERIFIED DUPLICATE of bug 144663

(Core :: Internationalization, defect)

Description

[kill this account]

assumming Mozilla is modified to use FreeType we will need to determine
how printing will work with FreeType

Comment 1

[kill this account]

My current plan is to generate Postscript and send it to Ghostscript for
which will rendering to the specific piece of hardware (the printer).

To maintain the best visual quality I plan to embedd the TrueType data in
the Postscript (AKA: Type42).

To provide for a large character set I plan to use CID keying.

To make the output as compact as possible I plan to do "incremental glyph
loading" (only send the needed glyphs).

The name for a Postscript CID keyed Type42 font is Type11.

I have "hand" built a Type11 font with incremental glyph loading.

I plan to start with the OpenOffice psprint code and add Type11 to it.

Comment 2

[Roland Mainz]

bstell:
Is there any PostScript printer which understands "Type42" ?

Comment 3

[kill this account]

My plan is to send the data to Ghostscript not a Postscript printer.

Comment 4

[kill this account]

For the first step I'm going to build a demo.

For the demo I will aim for functionality and will not spend time on things 
I am fairly sure are doable. The demo should be able to print a Japanese
page but will have the following issues that will still need to be addressed:

 hardcoded the font name (should come from CSS / prefs)

 always load the one font in the begin document (should load fonts as needed)

 always send the glyph data everytime a char is output (should keep track of
 which glyphs have been sent so we only send them once)

 assume the fonts supports all the needed chars

 send the all the Unicode CID ranges (only need to send those where the font
 has chars)

When the demo is working we can discuss what to do next.

Comment 5

[Yuying Long]

-> nsbeta1

Comment 6

[kill this account]

Yuying: thanks for the nsbeta1 vote. I also really want to get this in. I 
doubt however, I will have code ready  before the freeze for 1.0.

Comment 7

[Jungshik Shin]

Just FYI....
The ultimate solution to bug 121879 and
bug 126602 (and many other printing issues,
needless to say) appears to be FT printing.
 

Comment 8

[kill this account]

could you expand on what 'FT printing' means

Comment 9

[Frank Tang]

nsbeta1- . This is a post m1.0 feature . 

Comment 10

[kill this account]

Attachment: sample postscript output

sample postscript with the hmtx table

Comment 11

[kill this account]

Attachment: sample postscript without the hmtx table

Comment 12

[kill this account]

Attachment: tar/gz of demo program

Comment 13

[kill this account]

There are several problems:

 Currently Ghostscript can has problems reading the metrics
 from the incrementaly loaded glyphs.

 Ghostscript seems to crash on certain fonts/glyphs.

The demo code is working but far from beautiful.

My Postscript is still limited so I'm probably doing many things wrong. 

Comment 14

[kill this account]

Attachment: postscript displayed via ghostview

Comment 15

[kill this account]

Attachment: without hmtx table

Comment 16

[kill this account]

Attachment: tar/gz; modified OpenOffice fontsubstting code

This bit puts the OpenOffice fontsubsetting code in
mozilla/gfx/src/ps/OpenOffice. I'm not sure where it should go right now.

Comment 17

[kill this account]

Attachment: DEMO HACK to add truetype printing to moz

I just wanted to get something started so I hacked in the gfs/src/ps dir to 
make moz use truetype to print. This is a total hack and needs lots of rework.

It requires the modified OpenOffice printing code.
It is fairly buggy. It crashes Ghostscript.

Comment 18

[kill this account]

Attachment: tar/gz; modified OpenOffice fontsubstting code

this sort of works (still buggy)

Comment 19

[kill this account]

Attachment: patches to Ghostscript (6.53)

Comment 20

[kill this account]

Attachment: DEMO HACK to add truetype printing to moz (only output font once)

Comment 21

[kill this account]

Attachment: add string width measuring code

Comment 22

[kill this account]

Attachment: patches to Ghostscript (6.53)

Comment 23

[kill this account]

Attachment: tar/gz; modified OpenOffice fontsubstting code

Comment 24

[kill this account]

Attachment: minor tweaks

this still needs:

1) code to find the correct font (currently hard coded to use '/tmp/font.ttf')
2) code to allow the existing AFM code to co-exist with this code
3) support for multiple fallback fonts

Comment 25

[kill this account]

Attachment: tar/gz; minor fixes to OpenOffice fontsubstting code

substitute '-' for spaces in family name

Comment 26

[kill this account]

Attachment: Print code now uses font catalog and prefs. Still needs lots of work.

Separated FreeType2 access code (nsFreeType) from code that draws to screen
(nsFreeTypeX11).

Printing code now uses (a separate copy of) the font catalog.

Printing code now falls back to font prefs if CSS font not found.

This still needs lots of work.

Comment 27

[kill this account]

Attachment: png of www.mozilla.org/start using a font I found on the web (displayed via GhostView)

the font came from this site: http://fonts.tom7.com/fonts98.html

Comment 28

[kill this account]

Attachment: gz; postscript for www.mozilla.org/start in png

(note: requires patched Ghostscript; see attachments)

Comment 29

[kill this account]

Attachment: png of www.yahoo.co.ja printout (displayed via GhostView)

Comment 30

[kill this account]

Attachment: gz; postscript for www.yahoo.co.ja

(Note: requires patches to Ghostscript; see attachments)

Comment 31

[kill this account]

Attachment: png; simplified chinese

Comment 32

[kill this account]

Attachment: postscript for simplified chinese png

Comment 33

[Masaki Katakai]

Can Sun's china team verify Bstell's PS can be viewed on Solaris DPS?

Comment 34

[kill this account]

Attachment: png; simple text with lsb

Comment 35

[kill this account]

Attachment: postscript for simple_lsb.png

Comment 36

[kill this account]

Can Sun China try applying the patches to Ghostscript 6.53 and 
ghostscript/ghostview/print the postscript examples?

Comment 37

[Pete Zha]

Brian, No problem. We will do it later.
I just compiled mozilla with freetype enabled last week. And I will try your
patch this week.

Comment 38

[Pete Zha]

Brian, I have built ghostscript on my workstation with your patch. And all the
ps files you uploaded work fine for me.

Comment 39

[Masaki Katakai]

Pete, how about Solaris DPS? Does it work also?

Comment 40

[Pete Zha]

My platform is SunOS 5.9 s81_51 sun4u sparc SUNW,Sun-Blade-100
And I used gcc to build ghostscript.
Masaki:I will try DPS later.

Comment 41

[kill this account]

> I have built ghostscript on my workstation with your patch. And all the
> ps files you uploaded work fine for me.

very cool ;)

Just curious: do you use ghostview (gv) to display the output?

You might also want to try building the "tar/gz; minor fixes to OpenOffice 
fontsubstting code" (attachment 75871 [details]) 'stand alone'. You could then generate
your own postscript.

Comment 42

[Pete Zha]

Katakai, Can not use DPS to view those ps files. Does DPS need upgrade?
Brain, I'm using gs to view them.

Comment 43

[kill this account]

I finally figured a way to handle the "gsave/grestore removes the incremental
glyph definitions": 

 1) Wrap the gsave calls to push a hash on a stack of inc-glyph hashes.

 2) Everytime a glyph to be drawn check if the bit in the font's bit array is
    set. If the bit is not set then set the bit, add an entry for the
    font/glyphID to the current (top of stack) inc-glyph hash, and output the
    outline info.

 3) Wrap the grestore call to pop the current inc-glyph hash and clear the
    bit in the font's bit array for the font/glyphID pairs that are found in the
    hash

Comment 44

[kill this account]

Attachment: tar/gz: perl code to open/inspect a TrueType font

Comment 45

[Masaki Katakai]

All,

> My plan is to send the data to Ghostscript not a Postscript printer.

We shouldn't assume the destination for printing is only ghostscript.
It would not be acceptable on non-Linux system which do not have ghostscript. If
the system has the ghostscript but if user needs to
update to the latest to use Mozilla, it would not be acceptable too.

It's OK if we consider only power-users of Linux. They can configure
their ghostscript and Mozilla easily by their needs. But it does not
work for normal enduser and Linux/Unix distributors.

It would be OK if the first version of prototype is only Type42 and
requires ghostscript patch, but the final version of this support
should have Type1 or Type3 embedded fonts also, to support
existing PostScript printers and existing ghostscript.

I think it would be required,

- output is Type42, Type1 or Type3 embedded fonts
- Should have UI for selection of these types on preferences dialog
- Should have preference to control the default value

If there is a distribution who doesn't want to update ghostscript
but wants to support Mozilla FreeType2 printing, the distribution
can define "type3" or "type1" in preference as e.g. "freetype2.
print.outputformat".

Comment 46

[kill this account]

> We shouldn't assume the destination for printing is only ghostscript.
> It would not be acceptable on non-Linux system which do not have 
> ghostscript. 

Could you list the systems where Linux/Unix Mozilla/Netscape works
but Ghostscript is not available?

> If the system has the ghostscript but if user needs to
> update to the latest to use Mozilla, it would not be acceptable too.

Could you state why you think it not acceptable to ask the user
to update Ghostscript?

> It would be OK if the first version of prototype is only Type42 and
> requires ghostscript patch, but the final version of this support
> should have Type1 or Type3 embedded fonts also, to support
> existing PostScript printers and existing ghostscript.

How do you plan to support CJK fonts with existing printers?

Do you plan to implement font subsetting to handle the 8 bit font limit?

Comment 47

[kill this account]

Attachment: tar/gz; OpenOffice code with fix for composite glyph layout bug

Comment 48

[kill this account]

Attachment: patches to Ghostscript (6.53)

no longer need fix for composite glyphs

Comment 49

[kill this account]

*** This bug has been marked as a duplicate of 144663 ***

Comment 50

[Yuying Long]

Mark as verified due bug 144663 handles all the free type print issues.