darcs and ghc..

[Claus Reinke]

> A number of git commands are implemented as crufty shell scripts that
> don't check the results of running its plumbing commands. I've been
> bitten by silent command failure on a number of occasions. Just
> saying.

The earlier we start collecting _concrete_ instances of such silent
command failures, the better. Quite a bit of darcs trouble is at
least indicated by warnings. Many of those warnings should be
errors, since overlooking them gives you a not-quite-right repo,
with surprising follow-on failures, but at least one can search for 
the warnings in logs, if you keep them.

If there are odd corner cases where git commands fail silently
(without even a warning?), then ghc development will be quite
efficient at finding those corners, I'd guess!-). So it'd be good
to collect and report any and all such instances.

Part of the rationale for git was increased likelyhood of getting
bugs fixed quickly, after all.

Claus