Re: zsh-3.0-pre2 cores on irix 5.3 (?)

Re: zsh-3.0-pre2 cores on irix 5.3 (?)

[Bart Schaefer]

On Jul 7, 12:50pm, Andreas Koenig wrote:
} Subject: Re: zsh-3.0-pre2 cores on irix 5.3 (?)
}
} I started investigating, and reduced the bug (see below)
} 
}   > 1. Do you start getting the `alloc_stackp != 0' error immediately upon
}   >    starting zsh?  Or only after doing some completions?  Or only after
}   >    the periodic function starts firing?
} 
} Immediately after starting and then every $PERIOD seconds.
} 
} Here's my current .zshrc:
} 
} PERIOD=1800
} periodic () { . $HOME/.zlogout }
} 
} Here's my .zlogout:
} 
} echo "In .zprofile 1"
} #dirs >! $HOME/.zdirhistory
} #echo -n ^[]1\;$USERNAME@$HOST\a^[]2\;$USERNAME@$HOST $$\a
} 
} You see the lines it had before commented out. It turns out that ANY
} command in the file triggers _one_ alloc_stackp message.

Turns out that "source" or "." inside any shell function will trigger
that BUG warning, which isn't really a bug at all -- it's doshfunc()
calling loop() recursively, so the allocation stack depth reflects the
depth of recursion.

zagzig<3> echo echo foo > foo
zagzig<4> blat() { . foo }
zagzig<5> blat
foo
BUG: alloc_stackp != 0 in loop()
zagzig<6>

The for(;;) in loop() continues before reaching the DPUTS() call on any
blank or commented line, which is why I didn't see it when I first tried
to reproduce it, and which is why you see it once for each command.

Try this patch.

*** Src/init.c.orig	Fri Jul  5 10:57:45 1996
--- Src/init.c	Sun Jul  7 09:07:57 1996
***************
*** 128,134 ****
  	    if (toplevel)
  		noexitct = 0;
  	}
! 	DPUTS(alloc_stackp, "BUG: alloc_stackp != 0 in loop()");
  	if (ferror(stderr)) {
  	    zerr("write error", NULL, 0);
  	    clearerr(stderr);
--- 128,134 ----
  	    if (toplevel)
  		noexitct = 0;
  	}
! 	DPUTS(alloc_stackp > locallevel, "BUG: alloc_stackp > locallevel in loop()");
  	if (ferror(stderr)) {
  	    zerr("write error", NULL, 0);
  	    clearerr(stderr);

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern

New male in /home/schaefer:
>N  2 Justin William Schaefer  Sat May 11 03:43  53/4040  "Happy Birthday"

Re: zsh-3.0-pre2 cores on irix 5.3 (?)

[Andreas Koenig]

>>>>> Bart Schaefer writes:

 bart> Try this patch.

 bart> *** Src/init.c.orig	Fri Jul  5 10:57:45 1996
 bart> --- Src/init.c	Sun Jul  7 09:07:57 1996


Thanks Bart, the patch--of course--does the trick.

andreas

Re: zsh-3.0-pre2 cores on irix 5.3 (?)

[Zoltan Hidvegi]

> Turns out that "source" or "." inside any shell function will trigger
> that BUG warning, which isn't really a bug at all -- it's doshfunc()
> calling loop() recursively, so the allocation stack depth reflects the
> depth of recursion.

The allocation stack can be deeper than locallevel in loop() when source
is used inside a command substitution or when source is called from a
trap.  So the patch below is more correct that Bart's patch.  Note that
in the main top level loop alloc_stackp should always be zero.

Zoltan

rcsdiff -qc -kk -r2.23 -r2.24 Src/init.c
*** Src/init.c
--- Src/init.c	1996/07/07 16:03:48	2.24
***************
*** 101,106 ****
--- 101,109 ----
  loop(int toplevel)
  {
      List list;
+ #ifdef DEBUG
+     int oasp = toplevel ? 0 : alloc_stackp;
+ #endif
  
      pushheap();
      for (;;) {
***************
*** 128,134 ****
  	    if (toplevel)
  		noexitct = 0;
  	}
! 	DPUTS(alloc_stackp, "BUG: alloc_stackp != 0 in loop()");
  	if (ferror(stderr)) {
  	    zerr("write error", NULL, 0);
  	    clearerr(stderr);
--- 131,137 ----
  	    if (toplevel)
  		noexitct = 0;
  	}
! 	DPUTS(alloc_stackp != oasp, "BUG: alloc_stackp changed in loop()");
  	if (ferror(stderr)) {
  	    zerr("write error", NULL, 0);
  	    clearerr(stderr);

Re: zsh-3.0-pre2 cores on irix 5.3 (?)

[Zefram]

>+ #ifdef DEBUG
>+     int oasp = toplevel ? 0 : alloc_stackp;
>+ #endif

Huh?  If we're at the top level, alloc_stackp is 0 anyway.  Even if
there are bugs, it's 0 at this point -- the beginning of the function.
Or can loop() be called more than once with toplevel!=0?

-zefram

Re: zsh-3.0-pre2 cores on irix 5.3 (?)

[Zoltan Hidvegi]

> >+ #ifdef DEBUG
> >+     int oasp = toplevel ? 0 : alloc_stackp;
> >+ #endif
> 
> Huh?  If we're at the top level, alloc_stackp is 0 anyway.  Even if
> there are bugs, it's 0 at this point -- the beginning of the function.
> Or can loop() be called more than once with toplevel!=0?

run_init_scripts() is called before loop(1) which can mess up alloc_stackp.
Also debug tests should notice even seemingly impossible bugs.  Signal
traps can always cause surprises.  E.g. in execcmd() _very_ weird things
happened before introducing execsave()/execrestore() in dotrap().  For
example variables suddenly changed between two instructions.  Of course
this can still happen any time a global or static variable is used but now
it does not happen as often as before.

Zoltan

Re: zsh-3.0-pre2 cores on irix 5.3 (?)

[Andreas Koenig]

Thanks, Bart and Zoltan,

with the two chunks of patches you sent me, the core is gone, but I
get

k@anna:/usr/sources/shells/zsh-3.0-pre2% exec Src/zsh
BUG: alloc_stackp != 0 in loop()
BUG: alloc_stackp != 0 in loop()


instead. Let me know, if need additional info,
andreas

Re: zsh-3.0-pre2 cores on irix 5.3 (?)

[Zoltan Hidvegi]

Bart wrote:
> On Jul 6,  1:52pm, Andreas Koenig wrote:
> } Subject: zsh-3.0-pre2 cores on irix 5.3 (?)
> }
> } As I had to rebuild my whole irix 5.3 from both CDROM and backup
> } tapes, this may be a due to a misconfiguration of my
> } machine. Nonetheless, recompiling beta19 works, pre2 results in a
> } core.
> } 
> } exec.c: In function `execcmd':
> } exec.c:1318: warning: format argument is not a pointer (arg 3)
> 
> That (and the core dump) are almost certainly from the DPUTS() typo
> that someone (Wayne?) already reported:

That patch from Wayne also containd a patch to signals.c.  If int and pid_t
are different it may be the problem.  For me, with these patches it works:

digo ~ % uname -a
IRIX digo 5.3 11091812 IP22 mips
digo ~ % echo $ZSH_VERSION 
3.0-pre2

Zoltan


--- Src/signals.c	1996/07/04 19:40:08	2.12
+++ Src/signals.c	1996/07/05 20:52:22	2.13
@@ -464,7 +464,7 @@
 	    Job jn;
 	    Process pn;
             pid_t pid;
-	    int *procsubpid = &cmdoutpid;
+	    pid_t *procsubpid = &cmdoutpid;
 	    int *procsubval = &cmdoutval;
 	    struct execstack *es = exstack;
 

Re: zsh-3.0-pre2 cores on irix 5.3 (?)

[Bart Schaefer]

On Jul 6,  1:52pm, Andreas Koenig wrote:
} Subject: zsh-3.0-pre2 cores on irix 5.3 (?)
}
} As I had to rebuild my whole irix 5.3 from both CDROM and backup
} tapes, this may be a due to a misconfiguration of my
} machine. Nonetheless, recompiling beta19 works, pre2 results in a
} core.
} 
} exec.c: In function `execcmd':
} exec.c:1318: warning: format argument is not a pointer (arg 3)

That (and the core dump) are almost certainly from the DPUTS() typo
that someone (Wayne?) already reported:

--- zsh-3.0-pre2/Src/zsh.h	Fri Jul  5 10:57:49 1996
+++ zsh-3.0-pre2-fix/Src/zsh.h	Fri Jul  5 11:25:44 1996
@@ -1307,7 +1307,7 @@
 
 #ifdef DEBUG
 # define DPUTS(X,Y) if (!(X)) {;} else \
-			fprintf(stderr, "%s\n", X), fflush(stderr)
+			fprintf(stderr, "%s\n", Y), fflush(stderr)
 # define MUSTUSEHEAP(X) if (useheap) {;} else \
 		fprintf(stderr, "BUG: permanent allocation in %s\n", X), \
 		fflush(stderr)


} #3  0x45db7c in loop (toplevel=0) at init.c:131

That line is this call:

	DPUTS(alloc_stackp, "BUG: alloc_stackp != 0 in loop()");

So apparently our allocation problems aren't all over with, yet.

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern

New male in /home/schaefer:
>N  2 Justin William Schaefer  Sat May 11 03:43  53/4040  "Happy Birthday"

zsh-3.0-pre2 cores on irix 5.3 (?)

[Andreas Koenig]

As I had to rebuild my whole irix 5.3 from both CDROM and backup
tapes, this may be a due to a misconfiguration of my
machine. Nonetheless, recompiling beta19 works, pre2 results in a
core.

Excerpts from the compilation:


zsh configuration
-----------------
zsh version           : 3.0-pre2
host operating system : irix5.3
source code location  : .
compiler              : gcc
compiler flags        :  -Wall -Wno-implicit -Wmissing-prototypes -DDEBUG -g
binary install path   : /usr/local/bin
man page install path : /usr/local/man
info install path     : /usr/local/info



exec.c: In function `execcmd':
exec.c:1318: warning: format argument is not a pointer (arg 3)
exec.c: In function `execrestore':
exec.c:2590: warning: format argument is not a pointer (arg 3)



init.c: In function `loop':
init.c:131: warning: format argument is not a pointer (arg 3)



lex.c: In function `lexsave':
lex.c:85: warning: format argument is not a pointer (arg 3)
lex.c: In function `gettok':
lex.c:516: warning: format argument is not a pointer (arg 3)
lex.c: In function `gettokstr':
lex.c:994: warning: format argument is not a pointer (arg 3)
lex.c: In function `parsestr':
lex.c:1158: warning: format argument is not a pointer (arg 3)
lex.c: In function `parse_subst_string':
lex.c:1203: warning: format argument is not a pointer (arg 3)



signals.c: In function `unqueue_signals':
signals.c:388: warning: format argument is not a pointer (arg 3)
signals.c: In function `handler':
signals.c:444: warning: format argument is not a pointer (arg 3)
signals.c:467: warning: initialization from incompatible pointer type
signals.c:492: warning: assignment from incompatible pointer type



subst.c: In function `paramsubst':
subst.c:669: warning: format argument is not a pointer (arg 3)



text.c: In function `getredirs':
text.c:485: warning: format argument is not a pointer (arg 3)
text.c:490: warning: format argument is not a pointer (arg 3)



zle_tricky.c: In function `doexpansion':
zle_tricky.c:1192: warning: format argument is not a pointer (arg 3)
zle_tricky.c: In function `doexpandhist':
zle_tricky.c:3817: warning: format argument is not a pointer (arg 3)
zle_tricky.c: In function `getcurcmd':
zle_tricky.c:3901: warning: format argument is not a pointer (arg 3)



k@anna:/usr/sources/shells/zsh-3.0-pre2% Src/zsh
zsh: segmentation fault (core dumped)  Src/zsh
k@anna:/usr/sources/shells/zsh-3.0-pre2% gdb Src/zsh core
GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.15.1 (mips-sgi-irix5.3), Copyright 1995 Free Software Foundation, Inc...
Core was generated by `zsh'.
Program terminated with signal 11, Segmentation fault.
Reading symbols from /usr/lib/libcurses.so...done.
Reading symbols from /usr/lib/libc.so.1...done.
#0  strlen () at strlen.s:15
strlen.s:15: No such file or directory.
(gdb) bt
#0  strlen () at strlen.s:15
#1  0xfabe3bc in _doprnt () at doprnt.c:1588
#2  0xfac7ffc in fprintf () at fprintf.c:45
#3  0x45db7c in loop (toplevel=0) at init.c:131
#4  0x461258 in source (s=0x1003bcf0 "/usr/people/k/.zlogout") at init.c:780
#5  0x42a170 in bin_dot (name=0x1001b31c ".", argv=0x1001b3b8, 
    ops=0x7fffaa68 "", func=0) at builtin.c:4641
#6  0x410f7c in execbuiltin (args=0x1001b314, bn=0x100015f0) at builtin.c:189
#7  0x43b8dc in execcmd (cmd=0x1001b2f4, input=0, output=0, how=2, last1=2)
    at exec.c:1604
#8  0x43622c in execpline2 (pline=0x1001b36c, how=2, input=0, output=0, 
    last1=0) at exec.c:781
#9  0x43516c in execpline (l=0x1001b358, how=2, last1=0) at exec.c:627
#10 0x4349b8 in execlist (list=0x1001b348, dont_change_job=1, exiting=0)
    at exec.c:508
#11 0x4402ec in doshfunc (list=0x10020418, doshargs=0x0, flags=0, 
    noreturnval=1) at exec.c:2420
#12 0x49f1d0 in preprompt () at utils.c:581
#13 0x45d960 in loop (toplevel=1) at init.c:110
#14 0x45d6e8 in main (argc=1, argv=0x7fffaf14) at init.c:77
Current language:  auto; currently asm
(gdb) 




Hope, that helps,
let me know, if you need additional info,
andreas