After working this some more, I've come to the conclusion that the cruft directory is caused by my interrupting the program.
I thought that I had cleaned it up each time, but apparently not. Second, I'm reducing the 'hang' time claim to two minutes, not seven, when everything is properly cleaned up. But that's still a long time with no console or noticeable disk activity. Top shows 93% for the first 100 seconds or so, while darcs is around 6% and nothing else is getting much 'user' time. Then darcs jumps up to 90% for the finish. If I reduce the top update to 0.1 sec and sort on command, I can see that during that first 100 seconds there is one darcs_load_dirs process, but many hundreds of darcs processes are spawned and they get high cpu for a short time. It seems that much of the time is spent spawning and destroying these processes. I hacked darcs_load_dirs to switch on verbose output. Now there's plenty :-) and it doesn't seem to slow things up by more than 10%. Now I get the warm feeling that something's making progress. :0) -- KBK -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
