[9fans] suicide of the new /bin/gs

[9fans] suicide of the new /bin/gs

[Heiko Dudzus]

[-- Attachment #1: Type: text/plain, Size: 626 bytes --]

Some days ago there was a new ghostscript binary on sources.  I'm
having trouble with it:

| term% lp -dstdout goedel.dvi > goedel.ps
| term% lp goedel.dvi
| gs 3244: suicide: sys: fp: invalid operation fppc=0xd52cd status=0xb8a1 pc=0x000d4fc8
| cat: can't open /tmp/gsp3237: '/tmp/gsp3237' file does not exist
| term% yesterday -b -20070304 /386/bin/gs
| bind /n/dump/2007/0304/386/bin/gs /386/bin/gs
| term% lp goedel.dvi
| term% 

Printing to stdout works, printing to paper doesn't.  Using ps2pdf
also ends up in a gs suicide.

I attached a listing of stk(). I can send a snap(4) file on request.

Heiko

[-- Attachment #2: gs.stk --]
[-- Type: text/plain, Size: 1584 bytes --]

term% acid 3244
/proc/3244/text:386 plan 9 executable

/sys/lib/acid/port
/sys/lib/acid/386
acid: stk()
gs_screen_currentpoint(penum=0xdfffdbb4,ppt=0xdfffdb78)+0x7f /sys/src/cmd/gs/src/gshtscr.c:561
gx_ht_process_screen_memory(penum=0xdfffdbb4,pgs=0x300b78,phsp=0x64f268,accurate=0x0,mem=0x2ecd64)+0x4c /sys/src/cmd/gs/src/gsht.c:263
gs_sethalftone_prepare(pht=0x64f21c,pgs=0x300b78,pdht=0x64f288)+0x252 /sys/src/cmd/gs/src/gsht1.c:201
zsetcolorscreen(i_ctx_p=0x318c34)+0x1c8 /sys/src/cmd/gs/src/zht1.c:78
interp(pi_ctx_p=0x2ea80c,pref=0xdfffe0d8,perror_object=0xdfffe340)+0x898 /sys/src/cmd/gs/src/interp.c:1512
gs_call_interp(pref=0xdfffe31c,pi_ctx_p=0x2ea80c,pexit_code=0xdfffe348,perror_object=0xdfffe340,user_errors=0x1)+0x105 /sys/src/cmd/gs/src/interp.c:488
gs_interpret(pi_ctx_p=0x2ea80c,perror_object=0xdfffe340,pref=0xdfffe31c,user_errors=0x1,pexit_code=0xdfffe348)+0x59 /sys/src/cmd/gs/src/interp.c:446
gs_main_interpret(minst=0x2ea638,pref=0xdfffe31c,user_errors=0x1,pexit_code=0xdfffe348,perror_object=0xdfffe340)+0x45 /sys/src/cmd/gs/src/imain.c:214
gs_run_init_file(minst=0x2ea638,pexit_code=0xdfffe348,perror_object=0xdfffe340)+0x12d /sys/src/cmd/gs/src/imain.c:542
gs_main_init2(minst=0x2ea638)+0xf1 /sys/src/cmd/gs/src/imain.c:352
swproc(pal=0xdfffe65c,arg=0xdfffefc8,minst=0x2ea638)+0x2e8 /sys/src/cmd/gs/src/imainarg.c:270
gs_main_init_with_args(argv=0xdfffef54,argc=0x8,minst=0x2ea638)+0x292 /sys/src/cmd/gs/src/imainarg.c:207
main(argc=0x8,argv=0xdfffef54)+0x3b /sys/src/cmd/gs/src/gs.c:47
_main+0x26 /sys/src/ape/lib/ap/386/main9.s:12
acid: 

Re: [9fans] suicide of the new /bin/gs

[Joel C. Salomon]

On 3/7/07, Heiko Dudzus <heiko.dudzus@gmx.de> wrote:
> Some days ago there was a new ghostscript binary on sources.  I'm
> having trouble with it:

Seconded; converting lecture notes from PostScript to PDF is now failing.

Re: [9fans] suicide of the new /bin/gs

[Heiko Dudzus]

> On 3/7/07, Heiko Dudzus <heiko.dudzus@gmx.de> wrote:
>> Some days ago there was a new ghostscript binary on sources.  I'm
>> having trouble with it:
> 
> Seconded; converting lecture notes from PostScript to PDF is now failing.

I see no recent changes to the gs sources in the replica log, but
there were changes to ape. I guess that's causing the problem.(?)

Most of these changes were specific to the arm architecture.  The only
one that could matter for /386/bin/gs seems to be in

1173024004 0 c sys/src/ape/lib/ap/plan9/kill.c - 664 sys sys 1173022765 911

However, I'm not very sure about these guesses and I don't know how to
fix this.  :/

Heiko

Re: [9fans] suicide of the new /bin/gs

[lucio]

> However, I'm not very sure about these guesses and I don't know how to
> fix this.  :/

Hm, I submitted that patch after Andrey suggested it and I thought it
had worked successfully for what it was intended (see previous mail
about APE failures).  Perhaps it ought to have been tested more
carefully and perhaps it was more pervasive than I thought.  It may be
advisable to revert it.

++L

Re: [9fans] suicide of the new /bin/gs

[Heiko Dudzus]

lucio@proxima.alt.za wrote:
> Hm, I submitted that patch after Andrey suggested it and I thought it
> had worked successfully for what it was intended (see previous mail
> about APE failures).  Perhaps it ought to have been tested more
> carefully and perhaps it was more pervasive than I thought.  It may be
> advisable to revert it.

Sorry, it seems like /sys/src/ape/lib/ap/plan9/kill.c is not the
problem.  I reverted it, then recompiled libap, then recompiled gs and
it still suicides.

At the moment I have no clue why gs got broken.  (Correct me if it
isn't considered broken).

Heiko

Re: [9fans] suicide of the new /bin/gs

[erik quanstrom]

is this problem with the same gs and new ape or
a problem with new gs and new ape, i.e. two variables?

the backtrace seemed to indicate that it was a gs change.

- erik

Re: [9fans] suicide of the new /bin/gs

[Heiko Dudzus]

erik quanstrom wrote:
> is this problem with the same gs and new ape or
> a problem with new gs and new ape, i.e. two variables?

The former, I guessed, because I'm not aware of a changed gs source.
(Found nothing in replica log for (at least) a year)

Heiko