make-regexp error memory leak

make-regexp error memory leak

[Kevin Ryde]

[-- Attachment #1: Type: text/plain, Size: 305 bytes --]

When make-regexp throws an error it leaks memory, eg

	(while #t (false-if-exception (make-regexp "[")))

I'm looking at the changes below for the cvs and for 1.6, I think they
get the mallocated counting right.

        * regex-posix.c (scm_make_regexp): Free rx on error, to avoid memory
        leak.


[-- Attachment #2: regex-posix.c.free-cvs.diff --]
[-- Type: text/plain, Size: 648 bytes --]

--- regex-posix.c.~1.62.~	2003-04-07 08:05:25.000000000 +1000
+++ regex-posix.c	2004-07-04 18:56:03.000000000 +1000
@@ -1,4 +1,4 @@
-/*	Copyright (C) 1997, 1998, 1999, 2000, 2001 Free Software Foundation, Inc.
+/*	Copyright (C) 1997, 1998, 1999, 2000, 2001, 2004 Free Software Foundation, Inc.
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
@@ -188,6 +188,7 @@
 		    cflags & ~REG_NOSUB);
   if (status != 0)
     {
+      scm_gc_free (rx, sizeof(regex_t), "regex");
       scm_error (scm_regexp_error_key,
 		 FUNC_NAME,
 		 scm_regexp_error_msg (status, rx),

[-- Attachment #3: regex-posix.c.free-16.diff --]
[-- Type: text/plain, Size: 677 bytes --]

--- regex-posix.c.~1.53.2.3.~	2002-03-14 15:26:15.000000000 +1000
+++ regex-posix.c	2004-07-04 18:57:30.000000000 +1000
@@ -1,4 +1,4 @@
-/*	Copyright (C) 1997, 1998, 1999, 2000, 2001 Free Software Foundation, Inc.
+/*	Copyright (C) 1997, 1998, 1999, 2000, 2001, 2004 Free Software Foundation, Inc.
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -210,6 +210,8 @@
 		    cflags & ~REG_NOSUB);
   if (status != 0)
     {
+      free (rx);
+      scm_done_free (sizeof (regex_t));
       scm_error (scm_regexp_error_key,
 		 FUNC_NAME,
 		 scm_regexp_error_msg (status, rx),

[-- Attachment #4: Type: text/plain, Size: 143 bytes --]

_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: make-regexp error memory leak

[Rob Browning]

Kevin Ryde <user42@zip.com.au> writes:

> @@ -210,6 +210,8 @@
>  		    cflags & ~REG_NOSUB);
>    if (status != 0)
>      {
> +      free (rx);
> +      scm_done_free (sizeof (regex_t));
>        scm_error (scm_regexp_error_key,
>  		 FUNC_NAME,
>  		 scm_regexp_error_msg (status, rx),

Hmm.  It seems like you probably need a little more rearrangement
since scm_regexp_error_msg still needs the value.

-- 
Rob Browning
rlb @defaultvalue.org and @debian.org; previously @cs.utexas.edu
GPG starting 2002-11-03 = 14DD 432F AE39 534D B592  F9A0 25C8 D377 8C7E 73A4


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: make-regexp error memory leak

[Kevin Ryde]

Rob Browning <rlb@defaultvalue.org> writes:
>
> It seems like you probably need a little more rearrangement
> since scm_regexp_error_msg still needs the value.

Oops.  I ran it, obviously it was happy to look at freed memory :).
I changed to get the error message first.


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: make-regexp error memory leak

[Rob Browning]

Kevin Ryde <user42@zip.com.au> writes:

> Rob Browning <rlb@defaultvalue.org> writes:
>>
>> It seems like you probably need a little more rearrangement
>> since scm_regexp_error_msg still needs the value.
>
> Oops.  I ran it, obviously it was happy to look at freed memory :).
> I changed to get the error message first.

OK, once you get it working well, it should probably go into 1.6.

-- 
Rob Browning
rlb @defaultvalue.org and @debian.org; previously @cs.utexas.edu
GPG starting 2002-11-03 = 14DD 432F AE39 534D B592  F9A0 25C8 D377 8C7E 73A4


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: make-regexp error memory leak

[Kevin Ryde]

Rob Browning <rlb@defaultvalue.org> writes:
>
> OK, once you get it working well, it should probably go into 1.6.

Yep.  I made some changes.

There's probably enough bits and pieces to have a release from the 1.6
soon.  Especially since some of the fixes are actual wrong results for
certain cases (logbit, round, ...).


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: make-regexp error memory leak

[Rob Browning]

Kevin Ryde <user42@zip.com.au> writes:

> Rob Browning <rlb@defaultvalue.org> writes:
>>
>> OK, once you get it working well, it should probably go into 1.6.
>
> Yep.  I made some changes.
>
> There's probably enough bits and pieces to have a release from the 1.6
> soon.  Especially since some of the fixes are actual wrong results for
> certain cases (logbit, round, ...).

Definitely.  There have been a lot of improvements.  I was waiting to
make sure we'd cleared up all the bugs on Marius' release-critical
list.  That's done now, so I'll be gearing up for a 1.6.5 release
soon.

However, I want to finish and commit my srfi-4 work first since it
fixes a number of bugs.  For example, try storing a large number in a
64-bit vector -- the printed representations are wrong.

-- 
Rob Browning
rlb @defaultvalue.org and @debian.org; previously @cs.utexas.edu
GPG starting 2002-11-03 = 14DD 432F AE39 534D B592  F9A0 25C8 D377 8C7E 73A4


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: make-regexp error memory leak

[Kevin Ryde]

Rob Browning <rlb@defaultvalue.org> writes:
>
> For example, try storing a large number in a
> 64-bit vector -- the printed representations are wrong.

Ah, a bit dodgy that.

If you're hunting dodginess, it looks like make-s16vector may assume a
short is 16-bits when converting the "fill" parameter from an SCM.  A
system with a bigger short probably gets values silently truncated
instead of provoking range errors.  Similarly u16, and s32 and u32
with "int".


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: make-regexp error memory leak

[Rob Browning]

Kevin Ryde <user42@zip.com.au> writes:

> Ah, a bit dodgy that.
>
> If you're hunting dodginess, it looks like make-s16vector may assume a
> short is 16-bits when converting the "fill" parameter from an SCM.  A
> system with a bigger short probably gets values silently truncated
> instead of provoking range errors.  Similarly u16, and s32 and u32
> with "int".

Yep, thanks, though I *think*, I've probably already fixed those.  I
something similar to this for each one:

SCM_DEFINE (scm_make_s16vector, "make-s16vector", 1, 1, 0,
            (SCM n, SCM fill),
	    "Create a newly allocated homogeneous numeric vector which can\n"
	    "hold @var{len} elements.  If @var{fill} is given, it is used to\n"
	    "initialize the elements, otherwise the contents of the vector\n"
	    "is unspecified.")
#define FUNC_NAME s_scm_make_s16vector
{
  SCM uvec;
  size_t count = scm_num2size (n, 1, s_scm_make_s16vector);
  uvec = make_uvec (1, FUNC_NAME, SCM_UVEC_S16, count);

  if (SCM_UNBNDP (fill))
    memset (SCM_UVEC_BASE (uvec), 0, count * sizeof (int_s16));
  else
  {
    int_s16 * p;
    int f = scm_num2int (fill, 2, FUNC_NAME);

#if SIZEOF_INT > 2
    SCM_ASSERT_RANGE (2, fill,
                      (f >= (int_s16) -32768) && (f <= (int_s16) 32767));
#endif
    
    p = (int_s16 *) SCM_UVEC_BASE (uvec);
    while (count-- > 0)
      *p++ = f;
  }
  return uvec;
}
#undef FUNC_NAME


-- 
Rob Browning
rlb @defaultvalue.org and @debian.org; previously @cs.utexas.edu
GPG starting 2002-11-03 = 14DD 432F AE39 534D B592  F9A0 25C8 D377 8C7E 73A4


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel