[Developers] Test windows install procedure

[Johnoel Ancheta]

We should all vote on this.

I choose that the build files should set windows PATH.

Anyhow, I can fix the sh.exe issue tomorrow.  Can you test?

On 4/18/13, Arni Magnusson <arnima at hafro.is> wrote:
> Altogether, setting up a temporary PATH is a small price to pay for
> building ADMB in Windows. That said, here are some further findings and
> thoughts.
>
> ---
>
> The current makefile tree is sensitive to whether sh.exe is in the PATH or
> not.
>
> If sh.exe is in the PATH (there are 3 of them in my default PATH), then
> make.exe follows some Linux-like branch in the makefile tree. To study
> this in detail, see the attached 1.txt (stdout) and 2.txt (stderr) log
> files from make. As you can see in 1.txt, the make process is using
> Linux-like commands such as cp.exe and even ln.exe. As you can see in
> 2.txt, it's running into problems with the 'find' command (which works
> differently in Windows) and crashes in the middle of the build because of
> some unsupported command options.
>
> If I change my three sh.exe files to shh.exe, everything builds fine.
>
> In the GNUmakefile branches, there are clauses like "ifeq
> ($(SHELL),sh.exe)" whose purpose is to distinguish the platform.
>
> ---
>
> 1. We can just recommend using a truncated PATH for building ADMB, as is
> done in the current INSTALL.txt. It is a clear way to explain to advanced
> users what programs are required to build ADMB.
>
> 2. We might be able to improve the makefile tree so it does the right
> thing, even if some sh.exe are in the PATH. It would be great if users
> (like me) wouldn't have to modify their path before building ADMB. After
> all, GCC, make, flex, and sed are already in my default PATH, and this is
> true for many ADMB users interested in building it from scratch.
>
> 3. We can do both. Keep the truncated recommendation for now, but start
> looking into possible improvements in the makefiles, so ADMB can later be
> built with a full-fledged PATH.
>
> Arni