1.8.7 segfault

[Izit Izit]

Hi,

I have a Rails app that keeps on segfaulting if I request a very
intensive task. The task involves reading a lot of DB records and
parsing each record in hashes and arrays.

If I limit the DB record scope the segfault does not happen but with a
larger set it happens every time. Strange thing is it only happens on my
linux system not on my OS X dev system. Both run ruby 1.8.7 latest
version compiled from source. The process eats a lot of memory, keeps
growing and then the fault occurs.

I ran gdb on the faulty process, here is some output:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 48005413886848 (LWP 29060)]
ruby_re_compile_pattern (pattern=0x10f3070 ":", size=<value optimized
out>,
bufp=0x96f51f0) at regex.c:2730
2730 skip[c] = m;
(gdb) whe
#0 ruby_re_compile_pattern (pattern=0x10f3070 ":",
size=<value optimized out>, bufp=0x96f51f0) at regex.c:2730
#1 0x0000000000461cae in rb_reg_initialize (obj=48005969553640,
s=0x10f3070 ":", len=1, options=<value optimized out>) at re.c:643
#2 0x0000000000461f51 in rb_reg_new (s=0x10f3070 ":", len=1, options=0)
at re.c:1459
#3 0x0000000000461fba in rb_reg_regcomp (str=0) at re.c:1480
#4 0x0000000000476abd in rb_str_split_m (argc=<value optimized out>,
argv=<value optimized out>, str=48005969558200) at string.c:3560
#5 0x000000000041b84a in rb_call0 (klass=48005399435280,
recv=48005969558200,
id=4665, oid=4665, argc=1, argv=0x7fffff953900, body=0x2ba920c7c9b0,
flags=<value optimized out>) at eval.c:5904
#6 0x000000000041c598 in rb_call (klass=48005399435280,
recv=48005969558200,
mid=4665, argc=1, argv=0x7fffff953900, scope=0, self=48005506071920)
at eval.c:6151
#7 0x000000000041713b in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3492
#8 0x0000000000417047 in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3486
#9 0x0000000000416701 in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:2997
#10 0x0000000000417641 in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3047
#11 0x0000000000418b1e in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3332

#12 0x000000000041c0b3 in rb_call0 (klass=48005434552560,
recv=48005506071920,
id=51241, oid=<value optimized out>, argc=32767,
argv=0x7fffff955df0,
body=0x2ba922eb97a8, flags=<value optimized out>) at eval.c:6055
#13 0x000000000041c598 in rb_call (klass=48005434552560,
recv=48005506071920,
mid=51241, argc=2, argv=0x7fffff955de0, scope=1,
self=48005506071920)
at eval.c:6151
#14 0x0000000000417273 in rb_eval (self=<value optimized out>,
n=<value optimized out>) at eval.c:3507
#15 0x00000000004164ec in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3677
#16 0x000000000041c0b3 in rb_call0 (klass=48005434552560,
recv=48005506071920,
id=3769, oid=<value optimized out>, argc=32767, argv=0x7fffff957070,
body=0x2ba922e69050, flags=<value optimized out>) at eval.c:6055
#17 0x000000000041c598 in rb_call (klass=48005434552560,
recv=48005506071920,
mid=3769, argc=2, argv=0x7fffff957060, scope=1, self=48005506071920)
at eval.c:6151
#18 0x0000000000417273 in rb_eval (self=<value optimized out>,
n=<value optimized out>) at eval.c:3507
#19 0x000000000041c0b3 in rb_call0 (klass=48005457401280,
recv=48005506071920,
id=54129, oid=<value optimized out>, argc=11177,
argv=0x7fffff957c90,
body=0x2ba924411fb0, flags=<value optimized out>) at eval.c:6055


Does anyone have an idea ?

thanks

Jeroen

 

[Jim Morris]

Quite often I find this is caused by a bad memory SIMM, try running memtst86 or equivalent to
eliminate a H/W problem, then you can focus on potential S/W problems.

Hi,

I have a Rails app that keeps on segfaulting if I request a very
intensive task. The task involves reading a lot of DB records and
parsing each record in hashes and arrays.

If I limit the DB record scope the segfault does not happen but with a
larger set it happens every time. Strange thing is it only happens on my
linux system not on my OS X dev system. Both run ruby 1.8.7 latest
version compiled from source. The process eats a lot of memory, keeps
growing and then the fault occurs.

I ran gdb on the faulty process, here is some output:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 48005413886848 (LWP 29060)]
ruby_re_compile_pattern (pattern=0x10f3070 ":", size=<value optimized
out>,
bufp=0x96f51f0) at regex.c:2730
2730 skip[c] = m;
(gdb) whe
#0 ruby_re_compile_pattern (pattern=0x10f3070 ":",
size=<value optimized out>, bufp=0x96f51f0) at regex.c:2730
#1 0x0000000000461cae in rb_reg_initialize (obj=48005969553640,
s=0x10f3070 ":", len=1, options=<value optimized out>) at re.c:643
#2 0x0000000000461f51 in rb_reg_new (s=0x10f3070 ":", len=1, options=0)
at re.c:1459
#3 0x0000000000461fba in rb_reg_regcomp (str=0) at re.c:1480
#4 0x0000000000476abd in rb_str_split_m (argc=<value optimized out>,
argv=<value optimized out>, str=48005969558200) at string.c:3560
#5 0x000000000041b84a in rb_call0 (klass=48005399435280,
recv=48005969558200,
id=4665, oid=4665, argc=1, argv=0x7fffff953900, body=0x2ba920c7c9b0,
flags=<value optimized out>) at eval.c:5904
#6 0x000000000041c598 in rb_call (klass=48005399435280,
recv=48005969558200,
mid=4665, argc=1, argv=0x7fffff953900, scope=0, self=48005506071920)
at eval.c:6151
#7 0x000000000041713b in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3492
#8 0x0000000000417047 in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3486
#9 0x0000000000416701 in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:2997
#10 0x0000000000417641 in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3047
#11 0x0000000000418b1e in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3332

#12 0x000000000041c0b3 in rb_call0 (klass=48005434552560,
recv=48005506071920,
id=51241, oid=<value optimized out>, argc=32767,
argv=0x7fffff955df0,
body=0x2ba922eb97a8, flags=<value optimized out>) at eval.c:6055
#13 0x000000000041c598 in rb_call (klass=48005434552560,
recv=48005506071920,
mid=51241, argc=2, argv=0x7fffff955de0, scope=1,
self=48005506071920)
at eval.c:6151
#14 0x0000000000417273 in rb_eval (self=<value optimized out>,
n=<value optimized out>) at eval.c:3507
#15 0x00000000004164ec in rb_eval (self=48005506071920,
n=<value optimized out>) at eval.c:3677
#16 0x000000000041c0b3 in rb_call0 (klass=48005434552560,
recv=48005506071920,
id=3769, oid=<value optimized out>, argc=32767, argv=0x7fffff957070,
body=0x2ba922e69050, flags=<value optimized out>) at eval.c:6055
#17 0x000000000041c598 in rb_call (klass=48005434552560,
recv=48005506071920,
mid=3769, argc=2, argv=0x7fffff957060, scope=1, self=48005506071920)
at eval.c:6151
#18 0x0000000000417273 in rb_eval (self=<value optimized out>,
n=<value optimized out>) at eval.c:3507
#19 0x000000000041c0b3 in rb_call0 (klass=48005457401280,
recv=48005506071920,
id=54129, oid=<value optimized out>, argc=11177,
argv=0x7fffff957c90,
body=0x2ba924411fb0, flags=<value optimized out>) at eval.c:6055


Does anyone have an idea ?

thanks

Jeroen

 

[Izit Izit]

I am running on a VPS, so cannot check H/W. I am quite sure this is ok.

 

[Tim Hunter]

Hi,

I have a Rails app that keeps on segfaulting if I request a very
intensive task. The task involves reading a lot of DB records and
parsing each record in hashes and arrays.

If I limit the DB record scope the segfault does not happen but with a
larger set it happens every time. Strange thing is it only happens on my
linux system not on my OS X dev system. Both run ruby 1.8.7 latest
version compiled from source. The process eats a lot of memory, keeps
growing and then the fault occurs.

I ran gdb on the faulty process, here is some output:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 48005413886848 (LWP 29060)]
ruby_re_compile_pattern (pattern=0x10f3070 ":", size=<value optimized
out>,
bufp=0x96f51f0) at regex.c:2730
2730 skip[c] = m;

Have you tried recompiling with less optimization?

 

[Izit Izit]

Hi,

I ran gdb on the faulty process, here is some output:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 48005413886848 (LWP 29060)]
ruby_re_compile_pattern (pattern=0x10f3070 ":", size=<value optimized
out>,
bufp=0x96f51f0) at regex.c:2730
2730 skip[c] = m;

Have you tried recompiling with less optimization?

I have compiled with only --enable-pthreads, the rest was all default.

 

[Izit Izit]

Hi,

I ran gdb on the faulty process, here is some output:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 48005413886848 (LWP 29060)]
ruby_re_compile_pattern (pattern=0x10f3070 ":", size=<value optimized
out>,
bufp=0x96f51f0) at regex.c:2730
2730 skip[c] = m;

Have you tried recompiling with less optimization?

I have compiled with only --enable-pthreads, the rest was all default.

I have moved to ruby enterprise edition (latest version) and the problem
is gone.

 

[Bouquet]

I have a Rails app that keeps on segfaulting if I request a very
intensive task. The task involves reading a lot of DB records and
parsing each record in hashes and arrays.

If I limit the DB record scope the segfault does not happen but with a
larger set it happens every time. Strange thing is it only happens on my
linux system not on my OS X dev system. Both run ruby 1.8.7 latest
version compiled from source. The process eats a lot of memory, keeps
growing and then the fault occurs.

I ran gdb on the faulty process, here is some output:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 48005413886848 (LWP 29060)]
ruby_re_compile_pattern (pattern=0x10f3070 ":", size=<value optimized
out>,
bufp=0x96f51f0) at regex.c:2730
2730 skip[c] = m;

I just got the same crash in a very simple 1.8.7 program that
limits its memory using Process.setrlimit(Process::RLIMIT_AS, 10**6).

It looks the crash is due to a failure to check the success of
the the must_skip buffer allocation on line 2471 of regex.c