Re: pr-msvc-support merge

[Peter Rosin]

Den 2010-06-22 05:38 skrev Charles Wilson:
> What if $CC is specified as 'cl.exe'? IIUC, this means that the compile
> script will be invoked as "compile cl.exe ...." but then this case:
>
> +  cl | *[/\\]cl)
> +    func_cl_wrapper "$@"      # Doesn't return...
> +    ;;
>
> won't fire.

(also folding in the "cl\ * | */cl\ *" cases suggested by Ralf in
another threadlet)

Since the following is way over the top IMHO:

  [cC][lL] | *[/\\][cC][lL] | [cC][lL]\ * | *[/\\][cC][lL]\ * \
  | [cC][lL].[eE][xX][eE]    | *[/\\][cC][lL].[eE][xX][eE]    \
  | [cC][lL].[eE][xX][eE]\ * | *[/\\][cC][lL].[eE][xX][eE]\ * )
    func_cl_wrapper "$@"      # Doesn't return...
    ;;

I'm going with this instead:

  cl | *[/\\]cl | cl\ * | *[/\\]cl\ * \
  | cl.exe | *[/\\]cl.exe | cl.exe\ * | *[/\\]cl.exe\ * )
    func_cl_wrapper "$@"      # Doesn't return...
    ;;

> Is it possible to test the path translation logic? E.g. a test that is
> skipped unless $build+$host is one of the supported combinations?

I added compile5.test. I don't think $build is part of the equation
for the purpose of this test, and the only supported $host at this
point is mingw. Is it what you had in mind?

I'm should add that the test suite works as expected with
CC="/path/to/compile cl -MD" and CFLAGS empty (i.e. check8
and depcomp3 now passes, everything else the same as before.)

Cheers,
Peter