From: Roman Shterenzon <romanbsd@...>
Date: 2008-10-25T23:09:22+09:00
Subject: [ruby-core:19507] Possible memory leak when using $KCODE='u'

--0-400979337-1224943369=:71518
Content-Type: text/plain; charset=us-ascii

Good day,

While trying to write some indexing application, I've found that my application runs out of memory.
I followed the instructions on http://blog.evanweaver.com/articles/2008/02/05/valgrind-and-ruby/ and below you can see my findings.
I suspect that there is some memory leak in regexp which occurs when setting $KCODE='u'
I narrowed down this problem to this code:

#!/usr/bin/ruby
$KCODE='u'

class TestMem
  def testme
        page = '<html><head><title>Test</title></head><body></body></html>'
        page.match(%r{<title>(.*?)</title>}i)
  end
end
t = TestMem.new
t.testme

Below is the output of valgrind (when running against my app, it shows larger amounts of leaked memory):
==22373== 8 bytes in 1 blocks are definitely lost in loss record 1 of 30
==22373==    at 0x4006AEE: malloc (vg_replace_malloc.c:207)
==22373==    by 0x80B8835: ruby_re_compile_pattern (regex.c:2480)
==22373==    by 0x80AA112: make_regexp (re.c:643)
==22373==    by 0x80AB426: rb_reg_initialize (re.c:1376)
==22373==    by 0x80AB559: rb_reg_new (re.c:1410)
==22373==    by 0x809953F: ruby_yyparse (parse.y:2003)
==22373==    by 0x809AC4D: yycompile (parse.y:2624)
==22373==    by 0x809AF83: rb_compile_file (parse.y:2716)
==22373==    by 0x80C20BC: load_file (ruby.c:967)
==22373==    by 0x80C1C40: proc_options (ruby.c:843)
==22373==    by 0x80C2694: ruby_process_options (ruby.c:1236)
==22373==    by 0x8054B61: ruby_options (eval.c:1539)

..snip.. (Skipping, since all of reports list memory that was assigned to bufp structure in regex.c: ruby_re_compile_pattern)

==22373== 1,024 bytes in 1 blocks are definitely lost in loss record 26 of 30
==22373==    at 0x4006AEE: malloc (vg_replace_malloc.c:207)
==22373==    by 0x80B8723: ruby_re_compile_pattern (regex.c:2467)
==22373==    by 0x80AA112: make_regexp (re.c:643)
==22373==    by 0x80AB426: rb_reg_initialize (re.c:1376)
==22373==    by 0x80AB559: rb_reg_new (re.c:1410)
==22373==    by 0x809953F: ruby_yyparse (parse.y:2003)
==22373==    by 0x809AC4D: yycompile (parse.y:2624)
==22373==    by 0x809AF83: rb_compile_file (parse.y:2716)
==22373==    by 0x80C20BC: load_file (ruby.c:967)
==22373==    by 0x80C1C40: proc_options (ruby.c:843)
==22373==    by 0x80C2694: ruby_process_options (ruby.c:1236)
==22373==    by 0x8054B61: ruby_options (eval.c:1539)

This doesn't happen if I remove the $KCODE='u'
Is it a real memory leak?

Thank you in advance,
--Roman


      
--0-400979337-1224943369=:71518
Content-Type: text/html; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable

<html><head><style type=3D"text/css"><!-- DIV {margin:0px;} --></style></he=
ad><body><div style=3D"font-family:'times new roman', 'new york', times, se=
rif;font-size:12pt"><div>Good day,</div><div><br></div><div>While trying to=
 write some indexing application, I've found that my application runs out o=
f memory.</div><div>I followed the instructions on=A0http://blog.evanweaver=
.com/articles/2008/02/05/valgrind-and-ruby/ and below you can see my findin=
gs.</div><div>I suspect that there is some memory leak in regexp which occu=
rs when setting $KCODE=3D'u'</div><div>I narrowed down this problem to this=
 code:</div><div><br></div><div><div>#!/usr/bin/ruby</div><div>$KCODE=3D'u'=
</div><div><br></div><div>class TestMem</div><div>=A0=A0def testme</div><di=
v>=A0=A0 =A0 =A0 =A0page =3D '&lt;html>&lt;head>&lt;title>Test&lt;/title>&l=
t;/head>&lt;body>&lt;/body>&lt;/html>'</div><div>=A0=A0 =A0 =A0 =A0page.mat=
ch(%r{&lt;title>(.*?)&lt;/title>}i)</div><div>=A0=A0end</div><div>end</div>=
<div>t =3D
 TestMem.new</div><div>t.testme</div><div><br></div><div>Below is the outpu=
t of valgrind (when running against my app, it shows larger amounts of leak=
ed memory):</div><div><div>=3D=3D22373=3D=3D 8 bytes in 1 blocks are defini=
tely lost in loss record 1 of 30</div><div>=3D=3D22373=3D=3D =A0 =A0at 0x40=
06AEE: malloc (vg_replace_malloc.c:207)</div><div>=3D=3D22373=3D=3D =A0 =A0=
by 0x80B8835: ruby_re_compile_pattern (regex.c:2480)</div><div>=3D=3D22373=
=3D=3D =A0 =A0by 0x80AA112: make_regexp (re.c:643)</div><div>=3D=3D22373=3D=
=3D =A0 =A0by 0x80AB426: rb_reg_initialize (re.c:1376)</div><div>=3D=3D2237=
3=3D=3D =A0 =A0by 0x80AB559: rb_reg_new (re.c:1410)</div><div>=3D=3D22373=
=3D=3D =A0 =A0by 0x809953F: ruby_yyparse (parse.y:2003)</div><div>=3D=3D223=
73=3D=3D =A0 =A0by 0x809AC4D: yycompile (parse.y:2624)</div><div>=3D=3D2237=
3=3D=3D =A0 =A0by 0x809AF83: rb_compile_file (parse.y:2716)</div><div>=3D=
=3D22373=3D=3D =A0 =A0by 0x80C20BC: load_file (ruby.c:967)</div><div>=3D=3D=
22373=3D=3D =A0 =A0by 0x80C1C40: proc_options (ruby.c:843)</div><div>=3D=3D=
22373=3D=3D =A0 =A0by 0x80C2694:
 ruby_process_options (ruby.c:1236)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0=
x8054B61: ruby_options (eval.c:1539)</div><div><br></div><div>..snip.. (Ski=
pping, since all of reports list memory that was assigned to bufp structure=
 in regex.c:=A0ruby_re_compile_pattern)</div><div><br></div><div><div>=3D=
=3D22373=3D=3D 1,024 bytes in 1 blocks are definitely lost in loss record 2=
6 of 30</div><div>=3D=3D22373=3D=3D =A0 =A0at 0x4006AEE: malloc (vg_replace=
_malloc.c:207)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x80B8723: ruby_re_com=
pile_pattern (regex.c:2467)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x80AA112=
: make_regexp (re.c:643)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x80AB426: r=
b_reg_initialize (re.c:1376)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x80AB55=
9: rb_reg_new (re.c:1410)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x809953F: =
ruby_yyparse (parse.y:2003)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x809AC4D=
: yycompile (parse.y:2624)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x809AF83:=
 rb_compile_file (parse.y:2716)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x80C=
20BC: load_file
 (ruby.c:967)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x80C1C40: proc_options=
 (ruby.c:843)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x80C2694: ruby_process=
_options (ruby.c:1236)</div><div>=3D=3D22373=3D=3D =A0 =A0by 0x8054B61: rub=
y_options (eval.c:1539)</div><div><br></div><div>This doesn't happen if I r=
emove the $KCODE=3D'u'</div><div>Is it a real memory leak?</div><div><br></=
div><div>Thank you in advance,</div><div>--Roman</div></div></div></div><di=
v style=3D"position:fixed"></div></div><br>=0A=0A=0A=0A      </body></html>
--0-400979337-1224943369=:71518--