[Zlib-devel] configure - test execution method

[Mark Adler]

On Apr 20, 2010, at 9:21 PM, Mike Frysinger wrote:
> any attempt to build with distcc when any one of the nodes happens to be down 
> will trivially lead to random failures.  or running the shell script with 
> debugging turned on (bash -x ./configure).  or with ccache enabled and random 
> warnings originating from that.
> 
> relying on warnings in general is a buggy process.  the only time ive seen it 
> be required is attribute detection in gcc.  your experience on solaris isnt 
> specific to that OS ... you'll see the same behavior with gcc on any system.

So if I understand what you're saying, it is in fact necessary to test warnings in some cases (e.g. for attributes).  You are also saying that if you try to test warnings with distcc you will get spurious warnings depending on the state of the distributed machines.

Why would I not conclude from that that distcc is broken?  Why would I not conclude that ccache is broken for the same reason?

Mark