[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

ocaml 3.04 packaging issues ...



Hello, ...

I am just doing the ocaml 3.04 package, there is still some issues with the
install, particluarly with the camlp4 one which don't seem to support the
PREFIX stuff :((, and i need to look a bit more into the multi-package stuff.

That said, i have planed to :

  o ocaml will provide and replace camlp4
  o ocaml will provide and replace ocaml-base
  o ocaml will recommend ledit

  o i will split off a ocaml-base package
  o i am not sure about splitting off a camlp4 package ?

  o i will expressely use tcl/tk 8.3 in the build process. Sorry Georges, this
    will break your potato builds, but the previous solution caused a problem 
    when multiple tcl/tk -dev packages were installed, as is the case on some
    autobuilders.

Beside that we have to decide what is going to happen to the libraries and
executable packages, which regard to native/bytecode.

My current idea is to do :

Libraries :
    
  We will split the packages into a bytecode one (normal name) and a
  nativecode one (adding -native to the name).

  o the bytecode libraries will depend on ocaml-base, and be arch: all
  o the nativecode libraries will be arch: i386, powerpc, ... (list of
    supported arches only)

  Not sure if it is trully necessary to split them, what do you all think
  about it ?

Executables :

  We will split the packages into a bytecode one (normal name) and a
  nativecode one (adding -native to the name).

  o the bytecode libraries will depend on ocaml-base, and be arch: all
  o the nativecode libraries will be arch: i386, powerpc, ... (list of
    supported arches only)
  
  o bytecode executables will be postfixed with -byte, and nativecode
    executable will be postfixed with -native (or -opt or whatever is
    best)
  => We will use a diversion (or whatever is the correct mechanism) to
     have foo linked to foo_native if it is present, and if not to foo_byte.
  => We could use a debconf question to ask the user if he prefers to use 
     rather native code or rather bytecode, as a global switch to the above
     behavior
  => We could also use a wrapper script instead of a diversion.

What do you think of it ?

Friendly,

Sven Luther



Reply to: