On Tue, Dec 20, 2011 at 02:34:22PM +0000, YAMAMOTO Takashi wrote: > i haven't explored either.
Ok, I will give it a closer look (but that will take a few days). > well, i confess that i don't understand why in-kernel implementation is > desirable in the first place. I don't know what alternatives you consider better - IMHO the in kernel version is way smaller, minimal slightly more efficient, and a lot more elegant than any vfork based hack I could think of. Besides, I wouldn't know how to do all the dirty libpthread changes to make that thread safe. > i don't like having the lwp argument because they generally don't work for > non-curlwp. l_dupfd, "single threaded" optimization, ... I see the first point (but it is ok in this usage, maybe needs a few warning comments here and there). I'm not confinced it falls into the same categories as the other points ;-) Anyway, I will check if we can avoid it - this would make the overall change far less intrusive, which is always a plus. Martin