From: Kevin Smith <sent@...>
Date: 2000-12-27T11:49:10+09:00
Subject: [ruby-talk:8099] Converting a Ruby app into an executable

Well, today was an interesting day. I had some 
ideas about how to create a standalone executable 
out of a Ruby program. I almost made it, but hit 
a brick wall. I'll share, in case this sparks any 
ideas for other folks.

The app is named "rex", for Ruby-EXecutable.

My plan was:
1. Create a rexmake.rb utility to convert a 
collection of .rb files into one big .rex file:
   a. Start with the "main" .rb file
   b. Replace each require of another .rb file 
with the .rb file itself
   c. Remember all requires, and strip subsequent 
requires of the same file
   d. Now I have a HUGE .rex file that contains 
all the .rb code for the app
   e. Display the list of .so (or .dll) files 
that are still required
2. Create a rex.c app:
   a. Embed the Ruby interpreter
   b. Look at the end of the executable itself to 
find the HUGE source file
   c. Write that source to a tempfile
   d. Invoke the interpreter pointing to the 
tempfile
3. Glue it together:
   a. Append a .rex file to a copy of the rex 
app
   b. Include necessary signatures for error-
checking

I completed steps 1 and 2, and have been doing 
step 3 manually. For a simple app, it works 
great. You end up with a single executable that 
contains the interpreter and all the required 
Ruby code. You still need the shared libraries 
(if any) to be available at run-time. Is this 
worth packaging up and putting in the RAA?

But for complex apps (like mine) there are a few 
problems:
1. Tests for (__FILE__ == $0) are true, so 
debugging code wants to execute. I worked around 
this in my tests by manually deleting the two 
places it was happening. I can't think of an 
automated solution, but it would probably be an 
acceptable documented limitation.
2. $LOAD_PATH is not set in the embedded 
interpreter environment. I worked around this by 
appending the necessary search paths at run-time. 
This could be automated.
3. When I require 'gtk', I get a runtime error 
that rb_eRuntimeError is undefined. It would 
appear that still more of the normal Ruby runtime 
environment is missing from the embedded 
interpreter. I haven't looked to see if there's 
an easy solution to this.

By the way, I can run my HUGE .rex file through 
the normal Ruby interpreter, and it works just 
fine. It's about 8000 lines and 160k.

Thoughts?

Kevin