Re: [DotGNU]Platform testers wanted

[Glenn Chambers]

On Sunday, January 5, 2003, at 08:57 PM, I wrote:

> Tests so far:
>
> NetBSD 1.5.5 on x86:
>
> pnetC fails to configure; /usr/bin/cpp is a shell script that
> runs the real 'cpp', and it's choking on something it's being fed and 
> thinks it's
> preprocessing STDIN, so it just hangs until you hit EOF during the C 
> compiler check.
>
> I haven't tried to debug this beyond this point.  I'll try to get back 
> to it tomorrow.

Ok, I've continued to fiddle with it, with the following results.

1)  The failing process tree looks essentially like this:

Configure
  -> cscc ...
    -> cscc-c-s ...
      ->  /bin/sh /usr/bin/cpp ...
        -> /usr/libexec/cpp -traditional ...

As nearly as I can tell by inserting ECHO commands into the /usr/bin/cpp 
script, by
the time the script starts looping over its arguments, there is no 
quoting left, so
it treats '-D __SIZE_TYPE__=unsigned int' (note embedded spaces) as 
three seperate
arguments, which means it tries to treat the '__SIZE_TYPE__' as an input 
file to
parse.  This is notably unsuccessful, and things go rapidly downhill 
from there.

Perusing the code, I noticed the "CPP" environment variable, so I tried
"setenv CPP /usr/libexec/cpp" and reconfigured.

This time, I got a successful configure, and a successful build.  All 
samples where
run, and all succeeded.

I was unable to figure out how to induce the C compiler to echo the 
commands it was
running, so I can't tell if there was any quoting that got lost along 
the way.  Hints
are welcomed.

I'm out of hacking time for tonight, so off to bed.  Tomorrow, I'll see 
what I can
find out about debugging the FreeBSD crashes, and trying a MacOS X 
build.  (I'll
have to fiddle with that a bit, since I don't have AUTOMAKE on this 
Mac.  I'll
probably just run autogen.sh in each directory on a different box, then 
tar that
up and move it to the iMac.)