malloc error with String#<< / String#+

[Tj Holowaychuk]

The repo is here if anyone wants to take a look
http://github.com/visionmedia/dm-forms/tree/master

This is the first time I have had a memory error of any kind, heres the
output message:

forms ruby(7152,0xa041efa0) malloc: *** mmap(size=1073741824) failed
(error code=12)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
ruby(7152,0xa041efa0) malloc: *** mmap(size=1073741824) failed (error
code=12)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
/examples/../lib/dm-forms/base.rb:21:in `<<': failed to allocate memory
(NoMemoryError)
from ./examples/../lib/dm-forms/base.rb:21:in `form'
from ./examples/../lib/dm-forms/helpers.rb:19:in `form'
from examples/benchmarks.rb:11
...


I did a ruby-prof test on that same file and got

%self total self wait child calls name
79.15 4.50 4.50 0.00 0.00 174 String#+
13.24 0.75 0.75 0.00 0.00 84 String#<<
2.23 5.41 0.13 0.00 5.29 21
DataMapper::Form::Helpers#form


Looks REALLY odd to me ... when running more than 10-ish times it seems
to cause this memory error so something is really wrong that I have
done. Any suggestions? if you run 'rake benchmark' you will (probably)
see what i mean, I have tried without benchmarking as well this still
occurs so something stupid is going on.

Thanks in advance

 

[Brian Candler]

Ruby version? (ruby -v)
Operating system and version? (uname -a)
Source of Ruby package? (e.g. from distribution X, or compiled yourself)

 

[Tj Holowaychuk]

Ruby version? (ruby -v)
Operating system and version? (uname -a)
Source of Ruby package? (e.g. from distribution X, or compiled yourself)

right sorry, heres all that:

ruby 1.8.6 (2008-03-03 patchlevel 114) [universal-darwin9.0]

Darwin Macintosh.local 9.5.0 Darwin Kernel Version 9.5.0: Wed Sep 3
11:29:43 PDT 2008; root:xnu-1228.7.58~1/RELEASE_I386 i386

Its just 1.8.6 which was installed on the mac already

 

[Brian Candler]

OK, I'm not a Mac expert I'm afraid.

ulimit -a

would be the first thing I'd try, to see what process memory and stack
limits have been set, if any.

*** mmap(size=1073741824) failed (error code=12)

looks like something is asking to allocate exactly 1GB of RAM, which
seems fishy to me. Maybe it's been growing this space and has now hit
1GB.

If your Ruby process really is chewing up this much RAM (monitor it
using 'top' while it runs), you'll need to find out why.

- building a data structure which is too big?

- holding onto references longer than you need?

- however there are some strange failure modes very occasionally (esp.
with threads or continuations) where a phantom object reference remains
in the stack area scanned by the GC, but is not really live. There have
been some patches posted a while back which you could try.