FreeBSD ccache port is wonderfiul!

[tridge at samba.org]

Andrew,

 > I am pretty sure that most of failures are caused by
 > bad makefiles/configure scripts.

The only think ccache should affect (assuming no bugs!) is how fast
individual compiler runs are. While its certainly possible to write
makefiles and configure scripts that are affected by timing, I would
expect those would be found pretty quickly even without ccache.

 > Still, it seems like sometimes ccache is the cause. Makeworld
 > script has been tested quite thouroughly - and it almost surely
 > fails on -STABLE after running it about 5 times with ccache
 > (provided there are some changes in src). Or is it just me?

I haven't heard of this before. Are you sure you aren't just running
out of disk space? If you run with a small ccache size can you
reproduce this? (set the maximum cache size with 'ccache -M size').

 > A couple of weeks ago I just grew tired and stopped using
 > ccache. Maybe it's time for some investigation.  Lack of any
 > relevant experience doesn't help, but let me try. I'll start with
 > the next build.

thanks

One other thing you might try to narrow things down is to set the
environemnt variable CCACHE_DISABLE when you hit a failure. That way
you will still be launching ccache, but it will immediately call out
to the real compiler, avoiding the cache completely. If it still fails
then we know that only the its either the find_compiler() logic in
ccache or its not a ccache problem. If it doesn't fail then we'd need
to try to narrow down what file is causing a problem.

Cheers, Tridge