Bug 436099

Summary: Missing letter(s?) in Acrobat 5
Product: Ghostscript Reporter: Jack Moffitt <jack>
Component: PDF WriterAssignee: L. Peter Deutsch <ghost>
Status: NOTIFIED WONTFIX    
Severity: normal    
Priority: P2    
Version: master   
Hardware: All   
OS: All   
Customer: Word Size: ---

Description Jack Moffitt 2001-06-25 05:57:53 UTC
Originally reported by: rjvbertin@users.sourceforge.net
See the attached archive. The .ps file has been created 
with StarOffice 5.2 under windows 95, printing to the 
"Acrobat Distiller 4" printer (to file) via the adobeps 
4.4.1 driver. When viewed in gs7 (with gsview 4.0), 
everything seems to be OK. Viewed in Acrobat 5, 
however, an icircumflex is missing from the line 
"Surus-k n, ens  d'Umma" on the 2nd page. (There may be 
more missing but I didn't verify this). The icircumflex 
is there when I view with gs7.00.
When I convert the .ps file with Acrobat Distiller 4 
(result included too; the c43-distiller4.pdf), viewing 
seems OK in Acrobat 4, but there are substantial misses 
around that same position in Acrobat Reader 5.

I think this is not a problem with ghostscript, but I 
thought it a good idea to mention it anyway as the .ps 
file has been created using truetype fonts that *have* 
posed me considerable problems in conjunction with 
adobeps 4.4.1. If needed, I am willing to upload the 
fonts in question, or check 
http://rjvbertin.free.fr/Assyr/index.html .

RJV Bertin
Comment 1 Jack Moffitt 2001-06-26 05:42:50 UTC
Comment originally by rodw@users.sourceforge.net
Logged In: YES 
user_id=251410

In my experience with Acrobat in a professional printing 
environment, truetype is ALWAYS an issue. Most problems go 
away if you tell the printer driver to download truetype 
fonts as outlines. This converts all TT fonts to Type 1 
Postacript fonts and Distiller (and more importantly, the 
output device gets it right. On most of my RIPS, truetype 
fonts from Acrobat 4.0 cause a PS error. Might help your 
problem.
Comment 2 Jack Moffitt 2001-06-26 07:07:03 UTC
Comment originally by rjvbertin@users.sourceforge.net
Logged In: YES 
user_id=93983

Well, I could try to switch back to using outlines, of 
course... The readme that comes with adobeps (I since 
discovered that they're at version 4.51!) explains that it 
is usually better to use Type42 if the RIP can handle it. 
And indeed, the considerable problems that I mentioned (a 
font becoming useless even in Windows after 1x printing!) 
disappeared when I changed to sending as Type42. I have been 
able to isolate 2 glyphs in the particular font that are the 
cause, but I don't succeed in "repairing" them (or the 
hinting).

You talk about printing environment. I think this, and 
generating pdf, are slightly different issues. For a nice 
pdf, it is probably desirable to include truetype fonts as 
truetype. gs 7 is capable of that now, luckily - I find the 
quality (on-screen at least!) has significantly improved 
because of this. When printing from acrobat, one can then 
always configure the printerdriver such that optimal 
printing results...

BTW, I did succeed in "repairing" the icircumflex such that 
I now get correct behaviour. The interesting bit is that I 
know that this same glyph did not always display correctly 
under Mac OS 7.x ...

Thanks for the suggestion anyway.
Comment 3 Dan Coby 2001-06-26 12:07:39 UTC
Comment originally by dancoby@users.sourceforge.net
Logged In: YES 
user_id=12908

What exactly, if anything, would you like us to do?  If I 
read your report correctly, Ghostscript was not involved in 
the creation of the files involved.

Note:  We have found that the presence of kerning 
information in an embedded font causes Acrobat 5.0 to not 
display text.  This problem does not exist in Acrobat 4.0.
Comment 4 Jack Moffitt 2001-06-26 12:19:03 UTC
Comment originally by rjvbertin@users.sourceforge.net
Logged In: YES 
user_id=93983

I dunno :) Yes, ghostscript was involved in the creation! I 
checked with Distiller to see whether this was a 
ghostscript-particular problem, or a more general problem. I 
mentioned it here because I thought that it might be of 
interest for those working on the pdf writer (there may be 
workarounds?).

The note on kerning is interesting. As I mentioned in my 
previous post, in my case the problem disappeared when I 
recreated the offending glyph (composite, apparently with 
additional hinting).
Comment 5 Dan Coby 2001-06-26 14:52:29 UTC
Comment originally by dancoby@users.sourceforge.net
Logged In: YES 
user_id=12908

>Yes, ghostscript was involved in the creation!  I checked 
with Distiller to see whether this was a ghostscript-
particular problem, or a more general problem. 

Perhaps I am missing something.  Your report says:

1.>  See the attached archive.

No archive was attached.  I also have not found anthing 
pertaining to this problem sent to any of our standard bug 
reporting mail boxes.

2.>  The .ps file has been created with StarOffice 5.2 
under windows 95, printing to the  "Acrobat Distiller 4" 
printer (to file) via the adobeps 
4.4.1 driver.

Ghostscript not mentioned in this process.

3.>  When viewed in gs7 (with gsview 4.0), everything seems 
to be OK.

Here Ghostscript is used to view the file (not create) and 
everything works as desired.

4.>  Viewed in Acrobat 5, however, an icircumflex is 
missing from the line "Surus-k n, ens  d'Umma" on the 2nd 
page. (There may be more missing but I didn't verify this).

Adobe Acrobat 5.0

5.>  The icircumflex is there when I view with gs7.00.

Views correctly using Ghostscript.  No mention of file 
being created by Ghostscript.

6.>  When I convert the .ps file with Acrobat Distiller 4 
(result included too; the c43-distiller4.pdf), viewing 
seems OK in Acrobat 4, but there are substantial misses 
around that same position in Acrobat Reader 5.

Adobe Acrobat Distiller 4 and Reader 4 and 5.0.
Comment 6 Dan Coby 2001-06-26 17:33:09 UTC
Comment originally by dancoby@users.sourceforge.net
Logged In: YES 
user_id=12908

>Yes, ghostscript was involved in the creation!  I checked 
with Distiller to see whether this was a ghostscript-
particular problem, or a more general problem. 

Perhaps I am missing something.  Your report says:

1.>  See the attached archive.

No archive was attached.  I also have not found anthing 
pertaining to this problem sent to any of our standard bug 
reporting mail boxes.

2.>  The .ps file has been created with StarOffice 5.2 
under windows 95, printing to the  "Acrobat Distiller 4" 
printer (to file) via the adobeps 
4.4.1 driver.

Ghostscript not mentioned in this process.

3.>  When viewed in gs7 (with gsview 4.0), everything seems 
to be OK.

Here Ghostscript is used to view the file (not create) and 
everything works as desired.

4.>  Viewed in Acrobat 5, however, an icircumflex is 
missing from the line "Surus-k n, ens  d'Umma" on the 2nd 
page. (There may be more missing but I didn't verify this).

Adobe Acrobat 5.0

5.>  The icircumflex is there when I view with gs7.00.

Views correctly using Ghostscript.  No mention of file 
being created by Ghostscript.

6.>  When I convert the .ps file with Acrobat Distiller 4 
(result included too; the c43-distiller4.pdf), viewing 
seems OK in Acrobat 4, but there are substantial misses 
around that same position in Acrobat Reader 5.

Adobe Acrobat Distiller 4 and Reader 4 and 5.0.
Comment 7 Jack Moffitt 2001-06-27 12:28:48 UTC
Comment originally by rjvbertin@users.sourceforge.net
Logged In: YES 
user_id=93983

Ok, sorry, I wasn't clear: I discovered the problem in a pdf 
file created with gs7. I guess I have a problem talking to 
people as I do to my computers >8-

Re 1:) I am very positive I did attach an archive, however. 
I do seem to remember having seen some message about the 
size of the attachment, but it didn't remain visible long 
enough to register. If the attachment is indeed not there 
because of a size limit, it would be good if it were 
indicated, with alternatives!

If anyone wants to look into this, I'll send the archive if 
I still can find it.
Comment 8 Jack Moffitt 2001-06-27 15:08:35 UTC
Comment originally by jackiem@users.sourceforge.net
Logged In: YES 
user_id=206537

Can you submit the postscript file you are testing with? 
We'll need it if we're going to track down the problem.
Comment 9 Jack Moffitt 2001-06-28 12:30:36 UTC
Comment originally by rjvbertin@users.sourceforge.net
Logged In: YES 
user_id=93983

There should be a zipped postscript file uploaded with this 
message.
Comment 10 Jack Moffitt 2001-06-28 12:34:54 UTC
Comment originally by rjvbertin@users.sourceforge.net
Logged In: YES 
user_id=93983

&^%$@#$()! Only 256000 bytes allowed, I now saw. Well, then 
this one should pass. Only the 1st 2 pages. Created from 
StarOffice 5.2, using adobeps 4.51 with the Distiller 4 ppd.
Comment 11 L. Peter Deutsch 2001-07-26 22:51:17 UTC
Comment originally by lpd@users.sourceforge.net
Logged In: YES 
user_id=8861

Sorry, I don't understand the exact sequence of operations
to reproduce the problem. I converted c43.ps using the
current ps2pdf, and the result was displayed correctly with
Acrobat Reader 4 for Linux. Does the problem only occur with
AR5? Or do I need special options for ps2pdf? Please advise.
Comment 12 Jack Moffitt 2001-07-27 05:35:15 UTC
Comment originally by rjvbertin@users.sourceforge.net
Logged In: YES 
user_id=93983

1) convert using (current) ps2pdf, without any options.
2) view in Acrobat 4: result should be correct (cf. my
initial posting; verified)
3) View in Acrobat 5: problem should be there as mentioned.

In other words, as far as I can tell, this problem is indeed
specific to AR5. (I no longer have AR4 on a Win9x machine,
and AFAIK AR5 does not exist for Linux.) Do you need any
more specifics?
Comment 13 L. Peter Deutsch 2001-08-13 19:21:59 UTC
Comment originally by lpd@users.sourceforge.net
Logged In: YES 
user_id=8861

Thanks for the clarification. I (lpd) only run Linux, so I
can't investigate any further: I've reassigned the problem
to jackiem.
Comment 14 L. Peter Deutsch 2001-10-18 19:28:57 UTC
Comment originally by lpd@users.sourceforge.net
Logged In: YES 
user_id=8861

This is probably the same problem as #434707. We'll review
this one after we have a fix for that one.
Comment 15 L. Peter Deutsch 2002-01-24 13:28:29 UTC
Comment originally by lpd@users.sourceforge.net
Logged In: YES 
user_id=8861

It appears that this is *not* the same problem as #434707,
at least not entirely, now that I've investigated the
situation in more depth.

When I convert the c43.ps file with ps2pdf, Acrobat Reader
(AR) 4 on Linux and AR5 on Windows display it with the
i-circumflex of "Surus-kin" missing.

When I convert the file with Acrobat Distiller (AD) 5 on
Windows, AR5 on Windows displays the file with the *entire
rest of the line* missing, starting at the i-circumflex.

I looked at the file very carefully.  The i-circumflex
character is glyph #196 in font MSTT3190f18bf7tS00.  This
glyph is a composite glyph, made up of glyph #7 (dotless i)
and glyph #116 (circumflex).  The "flags" for the glyph #116
component have the value 0x102, indicating that the
component position is given as X and Y coordinates rather
than using point numbers, but that the coordinates each only
occupy a single (2's complement) byte rather than the more
customary 16 bits.  No other glyph in this font uses
this "flags" setting, and I think AR simply handles it
wrong.

I have reported this bug to Adobe.

Making ps2pdf work around this bug is somewhat awkward: it
will have to modify the actual glyph description, inserting
2 bytes and adjusting some lengths.  Unfortunately, I see no
alternative to doing this.  I will look into this next.
Comment 16 L. Peter Deutsch 2002-01-24 14:58:48 UTC
Comment originally by lpd@users.sourceforge.net
Logged In: YES 
user_id=8861

I've modified pdfwrite to always produce 16-bit X and Y
offsets for composite glyphs. The resulting file now
displays correctly with AR4 on Linux, but AR5 on Windows
still omits the i-circumflex.

I'll continue to explore this problem.
Comment 17 L. Peter Deutsch 2002-01-24 16:15:58 UTC
Comment originally by lpd@users.sourceforge.net
Logged In: YES 
user_id=8861

I was wrong; AR4 on Linux displays the converted file
properly even before the modification. So the modification
did not improve the situation at all.

I will wait for Adobe to get back to me about the Acrobat
bugs.
Comment 18 Jack Moffitt 2002-01-25 05:25:40 UTC
Comment originally by rjvbertin@users.sourceforge.net
Logged In: YES 
user_id=93983

I see I submitted an interesting problem  :)
I remember that I have at 
other problems with this same glyph (icircumflex), even with 
displaying it under some versions of the MacOS. If I remember correctly, 
I "solved" these by recreating the glyph (I use FontLab 3x). I'll see if I 
can retrieve the document that was used to create c43.ps, to see what 
happens now.

BTW: there is an AR update (5.05) that contains a 
substantial number of bug fixes!

RenE
Comment 19 L. Peter Deutsch 2002-01-25 08:50:07 UTC
Comment originally by lpd@users.sourceforge.net
Logged In: YES 
user_id=8861

Acrobat Reader 5.0.5 for Windows also replaces the
icircumflex with a narrow blank.
Comment 20 Jack Moffitt 2002-02-19 12:57:17 UTC
Comment originally by rjvbertin@users.sourceforge.net
Logged In: YES 
user_id=93983

Update. I just updated to ghostscript 7.04 .
Essentially, nothing has changed concerning the pdf
generation: the resulting pdf is shown correctly in acroread
4.05/linux, but not in reader 5.05/WIndows (shows narrow
space indeed).

HOWEVER, when I try to render c43.ps via gv, gs core dumps
(both the dynamically and the statically linked version).
This doesn't happen with v. 7.00, nor when I run gs directly
(resulting in an enormous window).

When I try to view the file under Windows via gsview 4.2,
well... Basically, I get a huge window, and a severe
out-of-memory-and-disk situation ("fit-window-to-page"
option on). Without that option, gs/gsview stubbornly keeps
giving me the 1:1 (or thereabouts) view instead of the
smaller. I know this is not the thread for these issues, but
is this a known issue?
Comment 21 L. Peter Deutsch 2002-04-10 00:07:46 UTC
Comment originally by lpd@users.sourceforge.net
Logged In: YES 
user_id=8861

It seems clear that there is a problem with this font that
has nothing to do with Ghostscript, since Acrobat Reader
won't display the output of Acrobat Distiller correctly
either. Therefore, we are closing the report.

Please report the other problems to their appropriate place.