From: Jeff Putsch <putsch@...>
Date: 2003-04-17T03:33:31+09:00
Subject: Debugger not working in 1.8.0 snapshot


--Apple-Mail-5--803635688
Content-Transfer-Encoding: 7bit
Content-Type: text/plain;
	charset=US-ASCII;
	format=flowed

Howdy,

I've built ruby from the April 15, 2003 snapshot ("ruby -v" returns 
"ruby 1.8.0 (2003-04-15) [sparc-solaris2.6]"). I used gcc-3.2.1, built 
on Solaris 2.6 and am running on Solaris 2.6 and 8.

Basically the debugger does not work. It does not stop at breakpoints. 
I suspect the issue is the same as was discussed in [ruby-talk:65644].

I have determined the problem is related to 
"debug.rb:check_break_points()" getting called with a pathname that is 
fully qualified. There is a commented out line (#    file = 
File.basename(file)) at the beginning of that function. If I uncomment 
(enable) that line, then the debugger works.

The attached patch to debug.rb adds minimal instrumentation to the file 
to help illustrate the problem, it is NOT a solution. If I apply the 
patch then do the following

   % cat /tmp/foobar.rb
   #!/usr/sepp/bin/ruby

   a=1
   b=2
   c=3
   d=[a,b,c]
   p a
   p b
   p c
   p d
   d[1]=4
   p b
   p d

   % /usr/sepp/bin/ruby -r debug /tmp/foobar.rb
   Debug.rb
   Emacs support available.

   /tmp/foobar.rb:3:a=1
   (rdb:1) b 6
   Set breakpoint 1 at foobar.rb:6
   (rdb:1) c
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   1
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   2
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   3
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   [1, 2, 3]
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   2
   check_break_points: file = /tmp/foobar.rb
   Breakpoint 1 b = [true, 0, "foobar.rb", 6]
   [1, 4, 3]

Notice that, in this case, check_break_points() is always called with 
the full path to the source file, while the breakpoint information has 
just the basename. Uncommenting the mentioned previously corrects this 
problem.

Thanks,

Jeff.


--Apple-Mail-5--803635688
Content-Disposition: attachment;
	filename=debug.rb.patch
Content-Transfer-Encoding: 7bit
Content-Type: application/octet-stream;
	x-unix-mode=0644;
	name="debug.rb.patch"

--- debug.rb.orig	Wed Apr 16 10:52:05 2003
+++ debug.rb	Wed Apr 16 11:18:08 2003
@@ -655,8 +655,10 @@
   def check_break_points(file, pos, binding, id)
     return false if break_points.empty?
 #    file = File.basename(file)
+    stdout.printf "check_break_points: file = #{file}\n"
     n = 1
     for b in break_points
+      stdout.print "Breakpoint #{n} b = #{b.inspect}\n" 
       if b[0]
 	if b[1] == 0 and b[2] == file and b[3] == pos
 	  stdout.printf "Breakpoint %d, %s at %s:%s\n", n, debug_funcname(id), file, pos

--Apple-Mail-5--803635688--