From @moritz

This innocently looking script on Rakudo #​18 sends the GC into an infinite
recursion and thus provokes a segmentation fault​:

use v6;
my $str = "a\nbc\nd";
say $str.subst(/^^/, '# ', :g);

The backtrace looks like this​:
(gdb) bt
#​0 0x00007f95a5c86e55 in mark_special (interp=0x176e010, obj=0x25a8180)
  at src/gc/mark_sweep.c​:460
#​1 0x00007f95a5c84633 in Parrot_gc_mark_PObj_alive (interp=0x176e010,
  obj=0x25a8180) at src/gc/api.c​:162
#​2 0x00007f95a5cf725b in mark_context (interp=0x176e010, ctx=0x1fd0250)
  at src/sub.c​:91
#​3 0x00007f95a5dbdb8f in Parrot_Sub_mark (interp=0x176e010, pmc=0x25a81b0)
  at ./src/pmc/sub.pmc​:505
#​4 0x00007f95a5dc0b82 in Parrot_Coroutine_mark (interp=0x176e010,
  pmc=0x25a81b0) at ./src/pmc/coroutine.pmc​:264
#​5 0x00007f95a5c86e57 in mark_special (interp=0x176e010, obj=0x25a81b0)
  at src/gc/mark_sweep.c​:460

and so on, repeating the first five lines all over again.

Cheers,
Moritz

--
Moritz Lenz
http://perlgeek.de/ | http://perl-6.de/ | http://sudokugarden.de/

From @Whiteknight

On Sun Jun 21 10​:44​:09 2009, mlenz@​physik.uni-wuerzburg.de wrote​:

This innocently looking script on Rakudo #​18 sends the GC into an infinite
recursion and thus provokes a segmentation fault​:

use v6;
my $str = "a\nbc\nd";
say $str.subst(/^^/, '# ', :g);

The backtrace looks like this​:
(gdb) bt
#​0 0x00007f95a5c86e55 in mark_special (interp=0x176e010, obj=0x25a8180)
at src/gc/mark_sweep.c​:460
#​1 0x00007f95a5c84633 in Parrot_gc_mark_PObj_alive (interp=0x176e010,
obj=0x25a8180) at src/gc/api.c​:162
#​2 0x00007f95a5cf725b in mark_context (interp=0x176e010, ctx=0x1fd0250)
at src/sub.c​:91
#​3 0x00007f95a5dbdb8f in Parrot_Sub_mark (interp=0x176e010,
pmc=0x25a81b0)
at ./src/pmc/sub.pmc​:505
#​4 0x00007f95a5dc0b82 in Parrot_Coroutine_mark (interp=0x176e010,
pmc=0x25a81b0) at ./src/pmc/coroutine.pmc​:264
#​5 0x00007f95a5c86e57 in mark_special (interp=0x176e010, obj=0x25a81b0)
at src/gc/mark_sweep.c​:460

and so on, repeating the first five lines all over again.

Cheers,
Moritz

I've seen bugs like this in other places before. What happened in those
other examples was an infinite subroutine recursion, which eventually
triggered the GC to cleanup all the mess (Sub PMC -> context -> Sub ->
context -> Sub, etc). Could you go up to the top of the stack trace and
take a look for a pattern like that?

--
Andrew Whitworth
a.k.a Whiteknight

The RT System itself - Status changed from 'new' to 'open'

From @bbkr

tests added to t/spec/S05-substitution/subst.t
Committed revision 31445.

@bbkr - Status changed from 'open' to 'resolved'

From @kyleha

This is an automatically generated mail to inform you that tests are now available in t/spec/S05-substitution/subst.t

commit 14f573cb19cc83439e8b70117484b058fd03e2a3
Author​: bbkr <bbkr@​c213334d-75ef-0310-aa23-eaa082d1ae64>
Date​: Fri Jun 25 14​:47​:38 2010 +0000

  [t/spec/S05-substitution/subst.t] tests for RT #​66816 Zero-width substitution makes the GC recurse
 
  git-svn-id​: http://svn.pugscode.org/pugs@&#8203;31445 c213334d-75ef-0310-aa23-eaa082d1ae64

diff --git a/t/spec/S05-substitution/subst.t b/t/spec/S05-substitution/subst.t
index 6f4d1e1..46ab4d6 100644
--- a/t/spec/S05-substitution/subst.t
+++ b/t/spec/S05-substitution/subst.t
@@ -304,6 +304,13 @@ is '12'.subst(/(.)(.)/,{$()*2}),'24', '.. and do nifty things in closures';
     is 'The foo and the bar'.subst(/:i the/, {$str++}, :g, :samecase), 'Thau foo and thav bar', '.substr and :g and :samecase, worked with block replacement';
 }
 
+# RT #66816
+{
+    my $str = "a\nbc\nd";
+    is $str.subst(/^^/, '# ', :g), "# a\n# bc\n# d",
+        'Zero-width substitution does not make the GC recurse';
+}
+
 done_testing;
 
 # vim: ft=perl6