On Thu, 31 Aug 2006, Junio C Hamano wrote: > > Actually it was me who suggested the use of fd directly, along > with trying to do one message in a single write (i.e. not doing > stdio as much as possible). I definitely agree with the "one message in a single write" thing. That's in fact required to make sure that O_APPEND works properly (ie that you don't have potentially different processes mixing output) > I somehow felt there would be less interference between writes > issued by process A and process B, but probably I am mistaken. The atomicity guarantees are the same regardless of whether you do a separate open(), the only thing that matters is that you do want to try to do messages with a single "write()" call, and that you use O_APPEND. That should guarantee that you get good behaviour. Without O_APPEND, you would obviously have to share one single file descriptor, because otherwise people will just be writing all over each other. Linus - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html