"restore" should restore resource category? (Re: [gs-devel] GS 6.53/7.05 unable to process CJK PS files printed from acroread)

[Ray Johnston]

Dear mpsuzuki,

I think that the problem is that the Adobe AR5 Procset is expecting that
global memory is unaffected by a restore, as it would be if the code
were running in a printer under a job server.

The job server saves the state before starting a job, then restores it
afterwards.

Ghostscript does not run files as jobs under a job server by default
so the ct_BadResourceImplementation? test is being performed at save
level 0. At save level 0, global memory is restored, including the
dictionary where the Resource category is kept.

To make Ghostscript run a file as if it were under a job server, you
can use:

	-c "false 0 startjob pop" -f file.ps

Where 'file.ps' is the file to be run.

When I test the code snippet you sent (with the comments removed),
the ct_BadResourceImplementation? gets set to true if I do this
startjob before running the code.

This may be the simplest solution to the problem. Please let me know.

Regards,
Ray
____________________________________________________________________

mpsuzuki at hiroshima-u.ac.jp wrote:
> 
> Dear Mr. Anthony and GS developers,
> 
> >http://anthony.homelinux.net/~foka/gs/c_hkscs-page3-printed-by-acroread-5.0.5.ps.gz
> 
> As I've written before, font errors are caused by missing-FID trouble.
> 
> PDF-to-PS convertor in AcrobatReader 5.x for UNIX inserts several small
> programs to check implementation of PS printer, and some of them is a bit
> nasty.
> 
> I suppose AR5 for UNIX inserts several emulation procedures into his
> output, but there might be some cases which suitable emulation procedures
> are not enabled.
> 
> If anybody have PS printer using (incomplete) clones of PS Level2 or
> Level3 renderer, please check how /ct_BadResourceImplementation? is
> defined and Anthony's sample is correctly rendered /or not.
> 
> In Anthony's case, implementation checker of AR5 (in MakeOCF utils)
> takes GS as "broken resource implementation", and it wipes out
> predefined CIDFont support codes from dictionary.  As a result, GS's
> CIDFont support is disabled, and GS cannot add FID to CIDFont dictionary.
> 
> Following is the checker AR5 for UNIX inserts, taken from Mr. Anthony's sample.
> ----------------------------------------------------------------------
> %%BeginResource: procset Adobe_CoolType_Utility_MAKEOCF 1.14 0
> %%Copyright: Copyright 1987-2001 Adobe Systems Incorporated.
> %%Version: 1.14 0
> 
> systemdict
>   /languagelevel known
>     dup
>       { currentglobal false setglobal } { false }ifelse
>     exch
>       userdict
>         /Adobe_CoolType_Utility 2 copy known
>           {
>             2 copy get dup maxlength 25 add dict copy
>           } {
>             25 dict
>           } ifelse
>           put
>         Adobe_CoolType_Utility begin
>           /ct_Level2? exch def
> %          /ct_Clone?
> %            1183615869
> %            internaldict
> %              dup /CCRun known not
> %                exch /eCCRun known not
> %                  ct_Level2?
> %                and
> %              or
> %            def
> 
>           /ct_BadResourceImplementation?
>             ct_Level2? {
>               save currentglobal true setglobal
>               /Generic /Category findresource dup length
>                 dict
>                 copy dup
>                   /InstanceType /dicttype put
>                   /egiremaevets exch
>                   /Category defineresource
>                 pop setglobal
>                 /foo 1 dict dup
>                   /s 3 put/egiremaevets
>                   defineresource
>                   pop
>                 restore
> %               mark {
>                 /bar 1 dict
>                   dup
>                     /a 7 put
>                     /egiremaevets defineresource
> %              } stopped {
> %                cleartomark
> %                true
> %              } {
> %                cleartomark false
> %              } ifelse
> %              currentglobal true setglobal
> %
> %              /egiremaevets /Category undefineresource setglobal
>             } {
>               false
>             } ifelse
>           def
> --------------------------------------------------------------------------
> 
> /ct_BadResourceImplementation? is set to true when the implementation
> is marked as "bad", and false when no problem (Distiller is marked as
> "false").
> 
> The testing procedure is:
> [1] if the system is PS Level 1, always
>     /ct_BadResourceImplementation? = false.
>     (I'm not sure whether this is reasonable)
> 
> [2] if the system is PS Level 2 or later,
>     following test is executed.
> 
>     [2-1] make the vanilla VM snapshot by "save"
> 
>     [2-2] create /egiremaevets resource category
> 
>     [2-3] add /foo dictionary to /egiremaevents resource
> 
>     [2-4] execute "restore"
> 
>     [2-5] try to add /bar dictionary to /egiremaevents resource
> 
>         [2-5-a] if ok, /ct_BadResourceImplementation? = false
> 
>         [2-5-b] if failed, /ct_BadResourceImplementation? = true
> 
> In gs6.53, procedure [2-4] wipe out /egiremaevents category itself,
> so [2-5] is failed, and marked as "bad implementation".
> 
> In Distiller, procedure [2-4] wipes out /foo in /egiremaevents,
> but empty /egiremaevents is left in resource category. So, [2-5]
> is succeeded.
> 
> When I modify [2-5-b] to make always /ct_BadResourceImplementation? = false,
> gs6.53 could render Mr. Anthony's sample PDF.
> 
> There might be several fixes:
> 
> (A) make GS's resource management similar to Distiller (difficult?)
> 
> (B) by some options, restrict GS functionalities to real Level 1 (difficult?)
> 
> (C) duplicate FID adding procedure to default /DefineResource in gs_res.ps
> 
> (D) add missing-FID fallback functionalities to zbuildfont9 etc
> 
> which is best and which is easiest?
> 
> Regards,
> 
> mpsuzuki
> 
> _______________________________________________
> gs-devel mailing list
> gs-devel at ghostscript.com
> http://www.ghostscript.com/mailman/listinfo/gs-devel