cybergraphics testers required for AFPL Ghostscript 8 port

[whoosh777]

My port of AFPL Ghostscript 8.00 is essentially complete,

however the cybergraphics viewers (there are approx. 6 of them) need

testing. Once this testing is done

and any bugs encountered fixed the docs will be written

and the product released: announcements here and on

other Amiga sites.

The viewers (for viewing postscript and pdf files) are:

1. 24 bit rgb wb viewer,
2. 8 bit gray wb viewer,
3. 1 bit b/w wb viewer,
4. 24 bit rgb custom screen viewer,
5. 8 bit gray custom screen viewer,
6. 1 bit b/w custom screen viewer.

You may be wondering why custom screen and wb viewers:

custom screens allow lower resolutions for faster rendering.

You may wonder why bother with 8 bit gray or 1 bit b/w when

you have 24 bit rgb: 1 bit b/w is much much faster image render.

8 bit is also faster (I think!) its certainly faster for

printing as colour conversion can be done via look up table.


Now the equivalent viewers for AGA are tested + fully functioning,

HAM8 is used for viewing 24bit rgb which gives pseudo true colour.



The first 3 people to volunteer by replying to this message get the

task: that way I cannot be accused of bias in selecting testers.

Perhaps noone will volunteer, I hope not!


If anyone volunteers beyond 3 then some of them could do second tier testing:

ie once bugs are reasonably fully fixed with the first 3 then I can

send the program to the next volunteers. Having more than 3 testers at a time

is logistically too difficult as an email can easily take an hour to write,

:it shouldn't but it does.


In addition email me with your contact email address + details of

your system: eg amount of RAM, CPU, which graphics card, etc



The most important thing to test + debug is the 24 bit rgb wb viewer.

Note these viewers are completely new + fully written by me,

totally different from any viewers on any previous Ghostscript Amiga ports.


Some people use cybergraphics with 15 or 16 bit screens, testing is

required to be certain that the viewers functions correctly for

15, 16 and 24 bit wb screens.


This port also contains a fully tested + debugged improved version

of the Ghostscript turboprint driver written by Turboprint's author.


:the original Turboprint driver recompiled AS-IS would not print a

single pixel! However after a lot of testing + debugging it is

fully back in action, producing stunning commercial quality printouts

via Turboprint.



Requirements for testing: a 68k Amiga with cybergraphics, perhaps

state what card you have in your message in case it helps get

volunteers with different cards.



In case you dont know what Ghostscript is, it is

for viewing + printing  postscript and pdf files. It will also

convert them to jpeg, png, bmp, fax, ... formats. Essentially it is

an open source cross-platform word processor back end.

:AFPL ensure that the code understands the current up to date

versions of Postscript and PDF.


:its also completely free,


If you wish to test, please try and download and install the pre-release

from http://www.whoosh777.pwp.blueyonder.co.uk

the version I will send is 1000 times as good as the pre-release!


The pre-release sucks in comparison!!



One other thing, if anyone out there has a colour printer with

official AmigaOS driver for that specific printer, please contact me.

:I have written a 4096 colour (12 bit plane) device for Ghostscript printing via

AmigaOS (vs 24 bit rgb Turboprint). This needs testing, my printer is

a Star LC-10, which unfortunately only prints via Epson emulation

which doesnt understand colour. :Only one tester required for this.



This is an unusual posting, but I thought what better way to find testers?

Also I thought the classified's were inappropriate for this.

I'm interested to see what pulling power amiga.org has...



whoosh777@blueyonder.co.uk


PS: this port should have been done 10 years ago, better late than never!

[cecilia]

hey, good luck! i'd voluteer but my cybergfx amiga is 3000 miles away. and needs a new HD in any case  :-D
looking forward to the release!

[Dragster]

Hello Whoosh... I'm happy to test.

What's your email? Anyway, you can see my system specs on my signature

I'm running Cybergraphx4 (with lates patcgh applied).

regards,

Daniel

[whoosh777]

I havent figured out how to quote from messages so I'm doing this

by cut-and-pasting to files from the forum,



Dragster wrote:

===========================================

Hello Whoosh... I'm happy to test.

What's your email? Anyway, you can see my system specs on my signature

I'm running Cybergraphx4 (with lates patcgh applied).

regards,

Daniel

==============================================

:my email is whoosh777@blueyonder.co.uk,

Thanks for volunteering, you are at the top of the list being first to reply!

BTW what depth do you use for your WB?

also what does it say if you type

version cybergraphics.library full

?

I had better point out that I dont have a graphics card myself,

my setup is 50Mhz 68030 A1200 with 2 + 4 + 8 meg.

I will prepare a recompile with instructions and binmail this to you,

this should reach you by the end of today. This will be an

unoptimised 68020 fpu recompile. Development is speeded up by using

unoptimised recompiles.




Cecilia wrote:

=====================================================

hey, good luck! i'd voluteer but my cybergfx amiga is 3000 miles away. and needs a new HD in any case
looking forward to the release!


=====================================================

Thanks for wishing me luck!

At least it sounds like you are interested in using the product,

if you are away from your cybergfx amiga when the product is released,
you will be able to use HAM8 to view PS + PDF docs in
pseudo-18 bit colour == 262144 colours. This is fully functioning so
doesnt need testing!

With the HAM8 viewer the initial
render is slower as there is a lot of per-pixel computational overhead
especially to avoid HAM8 distortions when scrolling the view. Various
tricks are used that minimize HAM8 artefacts to very few pixels.

Once that render is complete its just as fast as eg 8 bit gray.


=======================================================

One of the problems with creating s/w is that a project can continue forever

with an eternal sequence of "just one more feature", to deal with this

problem I decided to do various things:

1. Have a no-more-new-features deadline: select a day beyond which only
   existing features will be worked upon.

2. Have a programming deadline, try to complete the programming by a
   certain day. I chose 30th June, I think it may go beyond this a bit
   but not much. Any going beyond this deadline will be for fixing
   bugs and problems. The main programming is already complete,
   this current testing is to fix details that eg prevent images appearing
   on cybergraphics.

3. Interesting ideas for new features that occur after 1. will have to wait till
   after product release no matter how interesting they are!


Project completion is a very difficult thing to do, once you begin moves to

complete you get this demoralised feeling!



I think also that a completed project is always imperfect,

imperfection is an integral part of completion.




whoosh777

[Kronos]

Would you mind someone testing it against CGX_V5 ? :-D

O.k. I know not 68k, and not even an Amiga, but I would like to make a try, and older
(6.5 ?) GhostScripts work just fine here, regardless if 68k or WOS.

[SilvrDrgn]

whoosh777 wrote:
My port of AFPL Ghostscript 8.00 is essentially complete, however the cybergraphics viewers (there are approx. 6 of them) need testing.

I may be able to do some testing, if I have time, on my A4000T with CV-PPC gfx card and CyberGraphX (latest patches).  Email me.

[whoosh777]

Kronos wrote:

=========================================================

Would you mind someone testing it against CGX_V5 ?

O.k. I know not 68k, and not even an Amiga, but I would like to make a try,
and older (6.5 ?) GhostScripts work just fine here, regardless if 68k or WOS.

=========================================================

We can give it a try!

What happens if you type

version cybergraphics.library full

?

Can you explain briefly what CGX_V5 is??

(I am puzzled by your statement that its not an Amiga)

I assume WOS=WarpOS, I know absolutely nothing about WarpOS,
can you explain in a few lines what this is??

:is it a re-implementation of OS3.? or something else??

(I know the A500 and A1200, and have used OS1.2,1.3,2.0,3.0,3.9,)

If the viewers dont function, then from the shell output +
description of how the program malfunctioned I probably
can get it to function.

Unlike some (all?) previous ports my code is deliberately uni-tasking
which should avoid certain potential probs. Multitasking code can
be very tricky to debug, as you dont have eg repeatability.
:its also easy to get wrong a protocol for inter-task communication.
(:there are always unforeseen subtleties).

However through interleaving it creates a good illusion of being
multitasking! (a bit like games use interleaving to have lots of creatures
moving around "in parallel", think of Lemmings!)
In fact through interleaving you can scroll + render + print seemingly
in parallel.



With AGA (on my setup at least) there is a bug with SetRGB32CM():
it doesnt set the lower 4 bits of the blue component.
ie if you set a 24 bit palette
colour with SetRGB32CM() and then read it with GetRGB32() and
then compare the upper 8 bits of each 32 bit component, blue is wrong
in the lower 4 bits!

To fix this I had to hack the correct values into memory directly,
this may throw an emulation. With the hacked correct colours
AGA looks much nicer. This hack is not relevant to cybergraphics.

For AGA emulation I may need to switch off this hack.


SilvrDrgn wrote:

=========================================================

I may be able to do some testing, if I have time,

on my A4000T with CV-PPC gfx card and CyberGraphX (latest patches). Email me.

=========================================================

Thanks for volunteering!

I'll send both of you binmails with instructions by the end of today.

A 4th person also emailed me in parallel but didnt post a message due to

an uncertainty. So possibly there will be 4 testers for this stage.

Any further volunteers will be contacted on a first come first serve

basis for second tier testing or if anyone in the existing list is

unable to test for any reason. If someone isnt able to test, then

I'll go down the list and contact the next person!

This thread literally contains a FIFO queue!



whoosh777

[realstar]

He probably means he is using Cybergraphics5 on a
Pegasos system under MorphOS which features the
latest version of CyberGrafx and is an Amiga
compatible system that runs on PPC.  See this:

http://www.pegasosppc.com

They are a great alternative to the classic
Amiga systems and there exists a very active
developer community too.  I also own one of
these machines and would recommend testing
your software on the Pegasos since it features
great compatibility with classic 68K software
and native PPC MorphOS programs.

[whoosh777]

realstar wrote:

================================================

He probably means he is using Cybergraphics5 on a
Pegasos system under MorphOS which features the
latest version of CyberGrafx and is an Amiga
compatible system that runs on PPC.

============================================

that makes sense, I remembered Kronos saying
he had a Pegasos in some recent thread.

But why did he say WOS, or was that a typo?

(did he mean MOS??)

=========================================

See this:

http://www.pegasosppc.com

They are a great alternative to the classic
Amiga systems and there exists a very active
developer community too. I also own one of
these machines and would recommend testing
your software on the Pegasos since it features
great compatibility with classic 68K software
and native PPC MorphOS programs.

===============================================

Pegasos are doing all the things we wish Amiga Inc
were doing and the boss is actually interested
in desktop Amigas!

No boss of Commodore-Amiga ever used Amiga's
what's the world coming to!

I remember David Pleasance of Commodore UK launching
an in house initiative to make staff familiarize
themselves with using Amigas!

Also OS and hardware are both being done in-house
which used to be one of Amiga's unique-selling-points.

I wish Eyetech were a bit more generous acknowledging
Pegasos's merits, eg the April chip thing proved
their ability to improvise around an external problem.

Everything Pegasos does is wrong by default according
to Eyetech.

----------------------------------

As long as "cybergraphics.library" is present
it should be alright. cybergraphics forces one to
 to write code more cleanly as the implementation
details of the library really are hidden.

Thus code has a much better chance of porting than with
AGA.

It will be interesting to see whether the program
functions as-is on Pegasos.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%

QUESTION:

Does Pegasos emulate the entire 68k series
or just eg 68020?

Ditto Amithlon?

Presumably MMU code will cause probs??

As long as it emulates the output of gcc2.95.3-4
then its OK.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%

As JIT-compilation is 75% of native speed its almost
not worth doing a separate recompile for G3 upwards:
pages are going to render like instantly on a G3,
the bottleneck is going to be the printer speed.
ie if you made the prog twice as fast on a G3
printing a page might take 10 seconds instead of 11!

(probably)

:I could look into doing a separate recompile both for
Pegasos and A1, however this should wait till after the
release to avoid further delays.

BTW 3 problems have decided to emerge right now with
my port: I noticed some very minor problem: resolution
of one page of a pdf file was not getting reset to
default values for the next page when it was supposed to.
I corrected this, however this led to 2 new probs:

Ghostscript decided to decrease the width of its bitmap
by 1 pixel without telling me,
this has thrown my code which always checks
widths before memory transfers in order to avoid
memory trashing. Also the above correction is
causing a final phantom page to happen resulting
in an abandoned viewer screen on program exit.

:As I write this I have such an abandoned (1-bitplane)
screen in the background. resource tracking would have
zapped that abandoned viewer. One can always hack
resource tracking though by putting in an extra layer
above OS-calls...

So I am in a panic rushing around trying to fix the
problem. I can see a quick fix by putting in an extra
field which will get set to prevent the phantom page
from opening! The bitmap shrinkage can be fixed by
making some temporary array grow by at least 1 byte.

:I will get this recompile to the volunteers today
I promise!!

(By today I mean today UK time, which may be tomorrow or
yesterday by the testers time, one tester is in Mexico
and another in Australia so normal meanings of time
have gone out the window. So dont try to sue me for
lying till another 3 hours have passed.)

Each change to the prog eats up another 25 minutes:
5 minutes to recompile the source file and 20 minutes
to relink the prog.

aaaarggggghhhhh!

whoosh777

[Kronos]

Each change to the prog eats up another 25 minutes:
5 minutes to recompile the source file and 20 minutes
to relink the prog.

aaaarggggghhhhh!

whoosh777

Reminds me of the time when I tried to use the C++ modul of SAS-C6.5 on a 030 at 32mhz .....

[whoosh777]

Kronos said:
 
==============================================
 
 Reminds me of the time when I tried to use the C++ modul of
SAS-C6.5 on a 030 at 32mhz .....

==============================================

not far off:

this is C with gcc on an 030 at 50Mhz!

SAS-C compilers are superb though, much better than gcc.
Porting Ghostscript with Lattice is very tricky though,
Lattice doesnt like the Ghostscript source code!

:I tried, but had to give up as there were much too many
error messages, reams and reams of errors.


Those problems I mentioned, each problem I fixed made even
worse ones appear. Eventually I had to reinstate the
original problem which now seemed quite harmless!

Finally at approx. 2:45am I sent off recompiles to 4 testers,
including you I think, unless I emailed it to your evil twin.

I asked him if he was you,

He was explaining what CGX_5 was, so I presume he was
you?

If he wasnt you, then please email me at whoosh777@blueyonder.co.uk,
how will I know its you and not your evil twin??

We have an authentication problem!

Put some keyword in your email and I will ask you here in the forum
if that was you, otherwise anyone could pretend to be Kronos.

Thats the only way we can resolve this, unless we bring in PGP,

I'm having now to turn people away as emails keep coming in
wishing to test. I'm putting people on a waiting list!

So this idea for getting Cyber testers was almost too successful!

I am very scared of the response from the testers as I think
the initial 24 bit Cyber WB viewer test is going to fail horribly!

To do a reasonable port of Ghostscript you have to become a
student of pathology!

crashes, trashed images, frozen mice pointers, mouse pointers vanishing,
software failure numbers, printer.device divide by zero crashes,
asynchronous IO crashes, crashes on exit, crashes on entry, abandoned windows,
bang! boom! fireworks all the time!


whoosh777