1Make sure build process dumps config errors in separate file from build
2errors (and warnings), from build output
3
4Fix lmbench's config stage stuff
5
6Redirect stdout and stderr for each test's setup and execute to 4 sep files
7  (APW: ok, not clear i really like them (out/err) being separated.
8   the loss of chronology is bad IMHO)
9  (MJB: I was thinking of kernel compiles I guess. Maybe we should just
10	do it for that, and make the rest more generic)
11
12Caching version of kernelexpand to grab new kernels
13
14Errors from profilers are not being caught (shove a syntax error in a profiler
15python script to see what I mean) ... generates a file in results/debug/debug,
16but doesn't cascade up
17