Re: memory leak changing IGNORECASE (gawk 3.1.1)

[Ian D]

Here is more info, I have a traceback.

=>[1] _kill(0x0, 0x6, 0x0, 0xffffffff, 0x6ecc0, 0xff1b5798), at 0xff21c534
  [2] abort(0xff23a000, 0x2090bc8, 0x2087088, 0x0, 0x1, 0x208f584), at
0xff1b587
c
  [3] xmalloc(0x400, 0x4, 0x2087128, 0xff23a000, 0x239d8, 0xff1c272c), at
0x23ee
c
  [4] build_state(0x1, 0x208f584, 0x0, 0x4, 0x0, 0x35478), at 0x286c4
  [5] dfaexec(0x8d80c, 0x6d000, 0x8d881, 0x1, 0xffbebc4c, 0xa), at 0x29a48
  [6] research(0x208f558, 0x8d808, 0x0, 0x79, 0x1, 0xa), at 0x3636c
  [7] re_parse_field(0x0, 0x6d06c, 0x79, 0x72568, 0x208f558, 0x2b34c), at
0x2be1
8
  [8] get_field(0x4, 0x0, 0x6b000, 0x0, 0x0, 0x0), at 0x2c718
  [9] r_get_lhs(0x8c6b8, 0x0, 0x1, 0x9f9e8, 0x239b0, 0x35478), at 0x48db4
  [10] r_tree_eval(0x7a4c0, 0x0, 0x7a4c0, 0x20870e0, 0x0, 0x208ce34), at
0x4658c
  [11] match_op(0x7a510, 0x60, 0x58800, 0xff23a000, 0x239d8, 0xff1c272c), at
0x4
8ff4
  [12] r_tree_eval(0x7a510, 0x0, 0x1, 0xce, 0x239b0, 0x35478), at 0x46744
  [13] r_tree_eval(0x7a510, 0x0, 0x0, 0x17, 0x6b000, 0x1751c), at 0x46734
  [14] r_tree_eval(0x7a560, 0x0, 0x7a588, 0x6b000, 0x6b000, 0xa), at 0x46850
  [15] interpret(0x7a588, 0x8, 0x88, 0x6b000, 0x6b000, 0x36328), at 0x46154
  [16] interpret(0x1, 0x45540, 0x6b000, 0x0, 0x0, 0x0), at 0x4520c
  [17] interpret(0x7a740, 0x8, 0x90, 0x6b000, 0x6b000, 0xff1c18ac), at
0x4557c
  [18] interpret(0x1, 0x45540, 0x6b000, 0x0, 0x0, 0x0), at 0x4520c
  [19] interpret(0x7d378, 0x8, 0x79, 0x6b000, 0x6b000, 0x8d808), at 0x4557c
  [20] interpret(0x1, 0x45130, 0x0, 0xce, 0x6b000, 0x6d800), at 0x4520c
  [21] interpret(0x1, 0x2ebbc, 0x0, 0x17, 0x6b000, 0x1751c), at 0x4519c
  [22] do_input(0x6a800, 0x0, 0x72928, 0xc8, 0x0, 0x1), at 0x2ebdc
  [23] main(0x6a800, 0x6a800, 0x32c94, 0x6d918, 0x0, 0x0), at 0x336e8

----- Original Message -----
From: "Ian D" <address@hidden>
To: <address@hidden>
Sent: Saturday, October 12, 2002 10:08 PM
Subject: memory leak changing IGNORECASE (gawk 3.1.1)


> I have a complex script that runs through a series of regex tests for
every
> record read.  The regex expressions are parsed from a config file and
stored
> in variables.  For each expression tested, IGNORECASE is set to a variable
> that is 0 or 1 depending on whether the match should ignore case.  After
the
> test, IGNORECASE is set back to 0.  When running this script, memory usage
> continually increases until swap space is exhausted.  I then commented out
> the first IGNORECASE and retained the second, and I have the same problem.
> Here is a snippet, inside of a for loop (for each fpatt in an array),
which
> is invoked for every record read.
>
>                         #IGNORECASE=icase
>                         fieldok=fnot ? ($fnum !~ fpatt) : ($fnum ~ fpatt)
>                         IGNORECASE=0
>
> But if you comment out IGNORECASE=0, the memory is static, even for
millions
> of input records.
>
> The error I get is "gawk: cmd. line:139: (FILENAME=- FNR=312911) fatal:
> Memory exhausted."  I then modified source and determined that the line
> printing this error is dfa.c:193.  Line 139 is the IGNORECASE=0 line.
>
> I have seen references in the ChangeLog similar to this problem, but
> apparently it isn't fixed.  Any ideas on how to fix this is appreciated.
>
> Thanks,
> Ian
>