What glad times! Mauricio and Jamis have come out from behind the curtain to show us their GDB hooks and whoa these are shiny hooks and handy hooks.

Mauricio gives us some ~/.gdbinit settings that add a stellar set of functions for taking apart Ruby processes. Most of the wizardry here is Nobu’s.

After attaching with gdb ruby [process-id]:

 (gdb) session-ruby
 (gdb) redirect_stdout
 [Switching to Thread -1210431808 (LWP 4584)]
 $1 = 2
 (gdb) rb_object_counts
 $2 = 3078718700

The session-ruby command loads Mauricio’s functions. Standard out gets redirected to the file /tmp/ruby-debug.[pid]. In the above, the rb_object_counts will make a list of all the classes with instantiated objects and a count of objects created of that class.

  32640  String
   2435  Array
    951  Regexp
    819  Class
    ...

Jamis’ library (gdb.rb) takes a totally different approach by offering a Ruby wrapper for GDB and preferring dissecting Ruby structs byte by byte. So you don’t have to redirect STDOUT.

 >> require 'gdb'
 => true
 >> gdb = GDB::Ruby.new([pid])
 => #<GDB::Ruby:0xb7d72550 @gdb=#<IO:0xb7d72438>, @names={}, @interns={}>
 >> gdb.object_space
 => {"File"=>24, "OptionParser"=>1, "Gem::Version::Requirement"=>38, ... }

I don’t know about you guys, but I’m kind of freaking out that I can do all this from irb. It’s been a thrilling couple of days watching this unfold. (Jamis’ original post and his explanation of the library.)