Weird behavior (gawk 3.0.4, 3.0.6, 3.1.0)

[Peter Seebach]

Hi!  I can't figure out whether this is yours or ours.  gawk (3.0.4 in our
current distribution) periodically does very surprising things.  e.g.:

---
$ ./gawk '/^foo$/ {print $1}' < /usr/share/dict/web2
cystoepiplocele
foo
---

This behavior is sporadic; some runs are fine, others fail.  I've tried gawk
3.0.4, 3.0.6, and 3.1.0.  Other programs seem to run fine, and the machine
has ECC memory, so I'm *assuming* it's not hardware; furthermore, we had a
brief rash of this happening on a number of boxes.

When I compiled gawk 3.0.4 with -DDEBUG=1, it failed:
$ ./gawk '/^foo$/ { print $1 }' /usr/share/dict/web2
gawk: fatal: internal error line 207, file: node.c

This failure is repeatable.

3.1.0 doesn't fail with that.  Output with -DDEBUG=1 is enclosed.  I can't
figure out how this can be happening.  I did try linking with Electric Fence,
and I get exactly the same results, so I'm uninclined to suspect a simple
malloc problem; I would have expected a different failure mode.

Any help/feedback/advice would be appreciated.  Oddly, this *doesn't* seem
to happen on all of the machines I have access to, even ones where all the
libraries and binaries involved md5 the same.  I don't really think this is
necessarily yours, but I'm a bit at sea here, and I don't know my way through
gawk well enough to debug it.

dfamust:
 0:BEGLINE 1:f 2:CAT 3:o 4:CAT 5:o 6:CAT 7:ENDLINE 8:CAT 9:END 10:CAT
 node: 0:BEGLINE
  in:
  is: ""
  left: ""
  right: ""
 node: 1:f
  in: "f"
  is: "f"
  left: "f"
  right: "f"
 node: 2:CAT
  in: "f"
  is: ""
  left: ""
  right: "f"
 node: 3:o
  in: "o"
  is: "o"
  left: "o"
  right: "o"
 node: 4:CAT
  in: "fo"
  is: ""
  left: ""
  right: "fo"
 node: 5:o
  in: "o"
  is: "o"
  left: "o"
  right: "o"
 node: 6:CAT
  in: "foo"
  is: ""
  left: ""
  right: "foo"
 node: 7:ENDLINE
  in:
  is: ""
  left: ""
  right: ""
 node: 8:CAT
  in: "foo"
  is: ""
  left: ""
  right: ""
dfaanalyze:
 0:BEGLINE 1:f 2:CAT 3:o 4:CAT 5:o 6:CAT 7:ENDLINE 8:CAT 9:END 10:CAT
node 0:BEGLINE
 nullable: no
 firstpos: 0:BEGLINE
 lastpos: 0:BEGLINE
node 1:f
 nullable: no
 firstpos: 1:f
 lastpos: 1:f
node 2:CAT
 nullable: no
 firstpos: 0:BEGLINE
 lastpos: 1:f
node 3:o
 nullable: no
 firstpos: 3:o
 lastpos: 3:o
node 4:CAT
 nullable: no
 firstpos: 0:BEGLINE
 lastpos: 3:o
node 5:o
 nullable: no
 firstpos: 5:o
 lastpos: 5:o
node 6:CAT
 nullable: no
 firstpos: 0:BEGLINE
 lastpos: 5:o
node 7:ENDLINE
 nullable: no
 firstpos: 7:ENDLINE
 lastpos: 7:ENDLINE
node 8:CAT
 nullable: no
 firstpos: 0:BEGLINE
 lastpos: 7:ENDLINE
node 9:END
 nullable: no
 firstpos: 9:END
 lastpos: 9:END
node 10:CAT
 nullable: no
 firstpos: 0:BEGLINE
 lastpos: 9:END
follows(1:f): 3:o
follows(3:o): 5:o
follows(5:o): 7:ENDLINE
follows(9:END):

Compiling pattern: ^foo$

Compiled pattern: 
0:      /begline
1:      /exactn/3/f/o/o
6:      /endline
7:      end of pattern.
7 bytes used/32 bytes allocated.
re_nsub: 0      regs_alloc: 0   can_be_null: 0  newline_anchor: 1
no_sub: 0       not_bol: 0      not_eol: 0      syntax: 13b04d

Compiling pattern: 

+

Compiled pattern: 
0:      /exactn/1/

3:      /dummy_failure_jump to 9
6:      /on_failure_jump to 15
9:      /exactn/1/

12:     /maybe_pop_jump to 6
15:     end of pattern.
15 bytes used/32 bytes allocated.
re_nsub: 0      regs_alloc: 0   can_be_null: 0  newline_anchor: 1
no_sub: 0       not_bol: 0      not_eol: 0      syntax: 13b04d
foo
yachtsmanship