Ruby/DL problem

[Aston]

Hello,

I have very little experience with dl. But I want to use it to
call SetEnvironmentVariable() from Win32. I could use FFI but cannot,
since this
script will be supplied to a customer where it is difficult to install.

SetEnvironmentVariable on
MSDN SetEnvironmentVariableA function (processenv.h) - Win32 apps | Microsoft Learn

reading sparse, old, and cryptic documentation I came up with this
snippet,
which of course does not work

require ‘dl/import’

module Win32
extend DL::Importer
extern “int SetEnvironmentVariableA( char* lpName, char* lpValue)”
extern “int GetEnvironmentVariableA( char* lpName, char* lpValue, int
size)”
end

buff = Array.new( 256, 0).pack(“c*”)
Win32.GetEnvironmentVariableA( “ENV_VAR”, buff, 255)
puts buff.unpack(“a*”)

this is the error i get
D:/ruby/usr/lib/ruby/1.9.1/dl/import.rb:194:in import_function': undefined method sym’ for nil:NilClass (NoMethodError)
from D:/ruby/usr/lib/ruby/1.9.1/dl/import.rb:118:in extern' from test.rb:5:in module:Win32’
from test.rb:3:in `’

machine is WinXP
ruby 1.9.1 (2008-10-28 revision 19983) [i386-mswin32_80]

can anybody suggest a better signature ?

Aston

[Aston]

On Fri, Jul 2, 2010 at 10:37 AM, Aston [email protected]
wrote:

which of course does not work
Win32.GetEnvironmentVariableA( “ENV_VAR”, buff, 255)
ruby 1.9.1 (2008-10-28 revision 19983) [i386-mswin32_80]

can anybody suggest a better signature ?

Aston

I can’t help you with your issue, but I’m curious why it’s necessary.
The MSDN page you link to says SetEnvironmentVariable is only for the
current process. Doesn’t ENV do what you need?

irb(main):001:0> ENV[‘FOO’]=‘bar’
=> “bar”
irb(main):002:0> echo %FOO%
=> “bar\n”
irb(main):003:0>

[Aston]

----- Original Message ----

module Win32
D:/ruby/usr/lib/ruby/1.9.1/dl/import.rb:194:in `import_function’: undefined
Aston

hmm, nobody here using Ruby.DL ?

[Aston]

On Jul 3, 12:20 pm, Aston [email protected] wrote:

this

this is the error i get

Aston

hmm, nobody here using Ruby.DL ?

Is not DL the problem, but the version you’re using. revision 19983 is
way older than latest released one:

Pathlevel 378 was revision 26273

Lurking in the NEWS and ChangeLog pointed several changes in DL.

Trying your example with latest RubyInstaller 1.9.1 release:

C:\Users\Luis\Desktop>ruby -v t.rb
ruby 1.9.1p378 (2010-01-10 revision 26273) [i386-mingw32]
C:/Users/Luis/Tools/Ruby/ruby-1.9.1-p378-i386-mingw32/lib/ruby/1.9.1/
dl/import.rb:116: warning: instance variable @type_alias not
initialized
C:/Users/Luis/Tools/Ruby/ruby-1.9.1-p378-i386-mingw32/lib/ruby/1.9.1/
dl/import.rb:194: warning: instance variable @handler not initialized
C:/Users/Luis/Tools/Ruby/ruby-1.9.1-p378-i386-mingw32/lib/ruby/1.9.1/
dl/import.rb:194:in import_function': undefined method sym’ for
nil:NilClass (NoMethodError)
from C:/Users/Luis/Tools/Ruby/ruby-1.9.1-p378-i386-mingw32/lib/
ruby/1.9.1/dl/import.rb:118:in extern' from t.rb:5:in module:Win32’
from t.rb:3:in `’

Which brings the question, why bother on doing this way when Ruby does
it out of the box?

[Aston]

Hi,

At Sat, 3 Jul 2010 00:37:00 +0900,
Aston wrote in [ruby-core:30996]:

module Win32
extend DL::Importer
dlload “kernel32” # is this correct DLL?
extern “int SetEnvironmentVariableA( char* lpName, char* lpValue)”
extern “int GetEnvironmentVariableA( char* lpName, char* lpValue, int size)”
end

The following patch refines the error messages.

diff --git a/ext/dl/lib/dl/import.rb b/ext/dl/lib/dl/import.rb
index fd23bc9…f9e8faf 100644
— a/ext/dl/lib/dl/import.rb
+++ b/ext/dl/lib/dl/import.rb
@@ -194,8 +194,12 @@ module DL
return ptr
end

• def handler

•  @handler or raise "call dlload before importing symbols and 
  

functions"

• end
• def import_symbol(name)

•  addr = @handler.sym(name)
  

•  addr = handler.sym(name)
   if( !addr )
     raise(DLError, "cannot find the symbol: #{name}")
   end
  

@@ -203,7 +207,7 @@ module DL
end

 def import_function(name, ctype, argtype, call_type = nil)

•  addr = @handler.sym(name)
  

•  addr = handler.sym(name)
   if( !addr )
     raise(DLError, "cannot find the function: #{name}()")
   end

[Aston]

On Sun, Jul 04, 2010 at 10:02:28AM +0900, Nobuyoshi N. wrote:

The following patch refines the error messages.

I like this, so I’ve added it. Thanks Nobu!