Re: Allow SHELL to include variables and functions.

[Eli Zaretskii]

> From: Paul Smith <address@hidden>
> Cc: address@hidden
> Date: Wed, 10 Oct 2007 09:17:11 -0400
> 
> Hi Eli; I was reviewing this patch:
> 
>       http://www.mail-archive.com/make-w32%40gnu.org/msg01387.html
> 
> and I had a few questions: first, I tried this with UNIX make (using the
> warning in the SHELL variable) and it already works there; I'm not sure
> why the windows port is different in this respect?

The Windows port is different because it does not blindly obey the
value of SHELL: it knows about cmd.exe, the stock Windows shell and
about Unixy shells whose names include the substring "sh".  Any other
value of SHELL is required to be an existing file, specified either as
an absolute file name, or as a relative file name that can be found
along PATH.  See the function find_and_set_default_shell, which is
specific to Windows, for the gory details.  If the value of SHELL
fails the tests of that function, that value will be rejected.

> Second, I don't quite get how this works; it seems to me that you're
> expanding the value when the variable is defined (where things like $@
> etc. are not set yet) and then storing the expanded variable...?

Yes, that's true.  But I think that things like $@ will get evaluated
when the command

     $(SHELL) -c whatever

is run by Make, because at that time all the variables present in the
command line are expanded, right?

> Third, I think you have a memory leak; you're not freeing the
> (potential) contents of alloc_value before you reassign it.

Yep, my bad.  Sorry.