Hello!

I'm trying to use hijack as an inferior Ruby process in Emacs, but sending regions and definitions seems not to work. Code gets sent from my editing buffer to the repl, but the code doesn't actually seem to get eval'd.

Maybe an issue with how the hijack prompt is started?

this hijack attempt fails, either because the process is always stuck in GC, or because it's 1.9.1?

rdp       1880 32308 93 16:41 pts/0    01:46:04 ruby multiple_runs_same_setting_grapher.rb yanc_30mb_3_url_use_bittorrent_atfalse
rdp       7126 23734  0 13:26 pts/6    00:00:07 ruby /home/rdp/go/src/mongrel_dir_handler_current_dir.rb
rdp       9177  9053  0 18:35 pts/5    00:00:00 grep ruby
[27:1835][rdp@ilab1:~]$ hijack 1880
=> Hijacking...
=> 1880 doesn't appear to be a Ruby process!

Here is the output I get running hijack --gdb-debug command on Rails 3/Ruby 1.9.2 process:

...
> #50 0x08167afc in vm_exec (th=0x8430740) at vm.c:1132
> #51 0x08167e44 in rb_iseq_eval_main (iseqval=138667220) at vm.c:1373
> #52 0x0805dfa2 in ruby_exec_internal (n=0x843e4d4) at eval.c:204
> #53 0x0805f8e2 in ruby_run_node (n=0x843e4d4) at eval.c:251
> #54 0x0805d370 in main (argc=3, argv=0xbffc7c44) at main.c:35
(gdb) set variable (int)rb_trap_pending=1
> No symbol "rb_trap_pending" in current context.
(gdb) break rb_trap_exec
> Function "rb_trap_exec" not defined.
> Make breakpoint pending on future shared library load? (y or [n]) [answered N; input not from terminal]
(gdb) continue

Obviously nothing happens afterwards.

PS. It looks like rb_trap_exec was removed in this commit
http://redmine.ruby-lang.org/repositories/diff/ruby-19/signal.c?rev=19080

On exit or when the connection is lost.

It looks like if the gdb command executes before we select() hijacking will just hang.

I just installed the latest 0.1.9 and when i tried to hijack a rails app, running on mongrel, it would just hang after the first screen output, i.e. "=> Hijacking ... ".
I am running ruby 1.8.6 patch 287. Is there any workaround on this?

$ hijack 20554
=> Hijacking.../var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/gdb.rb:8: Use RbConfig instead of obsolete and deprecated Config.
/var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/gdb.rb:113:in `write': Broken pipe (Errno::EPIPE)
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/gdb.rb:113:in `puts'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/gdb.rb:113:in `exec'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/gdb.rb:100:in `backtrace'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/gdb.rb:92:in `ensure_attached_to_ruby_process'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/gdb.rb:34:in `attach_outside_gc'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/gdb.rb:9:in `initialize'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/payload.rb:9:in `new'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/payload.rb:9:in `inject'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack/console.rb:10:in `initialize'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack.rb:18:in `new'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/lib/hijack.rb:18:in `start'
    from /var/lib/gems/1.9.1/gems/hijack-0.2.1/bin/hijack:27:in `<top (required)>'
    from /usr/local/bin/hijack:19:in `load'
    from /usr/local/bin/hijack:19:in `<main>'

The issue was that gdb is not installed. After installing it, hijacking appears to be working as intended (I get a "20554 doesn't appear to be a Ruby process!" error, but that's obviously a different issue).

I suggest to preflight a check for gdb and if not found, print "GDB could not be found. Please ensure it is installed." or similar message.

"object allocation during garbage collection phase"

Hi,
I get an error about mirror process every time I start hjak. I have to either reinstall the gem or to run the update to make the gem work.

details of the error:

/Users/xxxxxxxx/.rvm/gems/ruby-1.9.2-p290/gems/hijack-0.2.0/lib/hijack/console.rb:52:in mirror_process': undefined method-' for nil:NilClass (NoMethodError)
from /Users/shaikhtabrez/.rvm/gems/ruby-1.9.2-p290/gems/hijack-0.2.0/lib/hijack/console.rb:15:in initialize' from /Users/shaikhtabrez/.rvm/gems/ruby-1.9.2-p290/gems/hijack-0.2.0/lib/hijack.rb:18:innew'
from /Users/shaikhtabrez/.rvm/gems/ruby-1.9.2-p290/gems/hijack-0.2.0/lib/hijack.rb:18:in start' from /Users/shaikhtabrez/.rvm/gems/ruby-1.9.2-p290/gems/hijack-0.2.0/bin/hijack:27:in<top (required)>'
from /Users/shaikhtabrez/.rvm/gems/ruby-1.9.2-p290/bin/hijack:19:in load' from /Users/shaikhtabrez/.rvm/gems/ruby-1.9.2-p290/bin/hijack:19:in

Actually, today it has stopped working with the update too. Las time I did a full uninstall and an update. Any suggestions?

When attaching with just plain gdb, a simple rb_eval_string call just hangs.
script/server and shotgun must be doing something special. Note that hijack works with 'mongrel_rails start' and 'thin start'.

~/hijack/test$ ruby test.rb
33890
...

~/hijack/test$ ps aux | grep test.rb
1000 33890 97.0 4.1 531808 506212 pts/3 Rl+ 14:45 0:11 ruby test.rb

~/hijack/test$ ../bin/hijack 33890
=> Hijacking...
=> 33890 doesn't appear to be a Ruby process!

Joker ?

I tried hijacking irb and get the "hijacking ..." status.

As mentioned elsewhere, i did connect to the irb process using gdb and then run bt.

Here is part of the output:
#35 0x00000001001610d5 in rb_catch_obj (tag=4050190, func=0x10017d7a0 <catch_i>, data=0) at vm_eval.c:1523
#36 0x0000000100161b00 in rb_f_catch (argc=<value temporarily unavailable, due to optimizations>, argv=<value temporarily unavailable, due to optimizations>) at vm_eval.c:1499
#37 0x0000000100165487 in vm_call_cfunc inlined at /Users/rahul/.rvm/src/ruby-1.9.2-head/vm_insnhelper.c:401
#38 0x0000000100165487 in vm_call_method (th=0x1003016a0, cfp=0x1004ffce8, num=1, blockptr=0x1004ffd11, flag=8, id=<value temporarily unavailable, due to optimizations>, me=0x10031d590, recv=4321140880) at vm_insnhelper.c:523
#39 0x0000000100168215 in vm_exec_core (th=0x1003016a0, initial=<value temporarily unavailable, due to optimizations>) at insns.def:1006
#40 0x000000010016fe53 in vm_exec (th=0x1003016a0) at vm.c:1145
#41 0x000000010017d9e1 in catch_i (tag=3214606, data=<value temporarily unavailable, due to optimizations>) at vm.c:557
#42 0x00000001001610d5 in rb_catch_obj (tag=3214606, func=0x10017d7a0 <catch_i>, data=0) at vm_eval.c:1523
#43 0x0000000100161b00 in rb_f_catch (argc=<value temporarily unavailable, due to optimizations>, argv=<value temporarily unavailable, due to optimizations>) at vm_eval.c:1499
#44 0x0000000100165487 in vm_call_cfunc inlined at /Users/rahul/.rvm/src/ruby-1.9.2-head/vm_insnhelper.c:401
#45 0x0000000100165487 in vm_call_method (th=0x1003016a0, cfp=0x1004ffea0, num=1, blockptr=0x1004ffec9, flag=8, id=<value temporarily unavailable, due to optimizations>, me=0x10031d590, recv=4303825320) at vm_insnhelper.c:523
#46 0x0000000100168215 in vm_exec_core (th=0x1003016a0, initial=<value temporarily unavailable, due to optimizations>) at insns.def:1006
#47 0x000000010016fe53 in vm_exec (th=0x1003016a0) at vm.c:1145
#48 0x0000000100170158 in rb_iseq_eval_main (iseqval=4303807640) at vm.c:1386
#49 0x000000010003f0d2 in ruby_exec_internal (n=0x10086e498) at eval.c:214
#50 0x0000000100041a5c in ruby_exec_node inlined at /Users/rahul/.rvm/src/ruby-1.9.2-head/eval.c:261
#51 0x0000000100041a5c in ruby_run_node (n=<value temporarily unavailable, due to optimizations>) at eval.c:254
#52 0x0000000100000ecf in main (argc=7, argv=0x7fff5fbfcfe0) at main.c:35

I'm not sure whether this is an issue or a feature, but I thought it would be handy to hijack a passenger rails process to debug it :) Is this theoretically possible? Anyway, I tried the following without success (I tried with and without sudo):

$ passenger-memory-stats
#75.6 MB 1.0 MB /Library/Ruby/Gems/1.8/gems/passenger-2.1.3/ext/apache2/ApplicationPoolServerExecutable
#2828 ? 177.7 MB 102.5 MB Rails:

Gives pid 2828

So I try to hijack it :)

$ hijack 2828 --gdb-debug

Ends with:

. done
0x903e16fa in select$DARWIN_EXTSN ()
(gdb) call (void)rb_eval_string("require 'stringio'\n require 'drb'\n\n unless defined?(Hijack)\n module Hijack\n class OutputCopier\n def self.remote\n @@Remote\n end\n\n def self.start(pid)\n @@Remote = DRbObject.new(nil, 'drbunix://tmp/hijack.' + pid + '.sock')\n\n class << $stdout\n def write_with_copying(str)\n write_without_copying(str)\n Hijack::OutputCopier.remote.write('stdout', str) rescue nil\n end\n alias_method :write_without_copying, :write\n alias_method :write, :write_with_copying\n end\n\n class << $stderr\n def write_with_copying(str)\n write_without_copying(str)\n Hijack::OutputCopier.remote.write('stderr', str) rescue nil\n end\n alias_method :write_without_copying, :write\n alias_method :write, :write_with_copying\n end\n end\n end\n\n class Evaluator\n def initialize(context)\n @context = context\n @file = FILE\n end\n\n def evaluate(rb)\n if rb =~ /_hijack_output_receiver_ready([\d]+)/\n OutputCopier.start($1)\n elsif rb =~ /__hijack_get_remote_file_name/\n @file\n else\n @context.instance_eval(rb)\n end\n end\n end\n\n def self.start(context)\n return if @service && @service.alive?\n evaluator = Hijack::Evaluator.new(context)\n @service = DRb.start_service('drbunix://tmp/hijack.2998.sock', evaluator)\n File.chmod(0600, '/tmp/hijack.2998.sock')\n end\n end\n end\n __hijack_context = self\n Signal.trap('USR2') { Hijack.start(__hijack_context) }")
Detaching from program: `/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/bin/ruby', process 2998 thread 0x203.