From: Junio C Hamano Date: Thu, 14 Jan 2010 22:40:27 GMT Subject: Re: [PATCH 3/4] start_command: detect execvp failures early Message-ID: <7vska88g4k.fsf@alter.siamese.dyndns.org> In-Reply-To: <201001142253.19595.j6t@kdbg.org> Johannes Sixt writes: > On Donnerstag, 14. Januar 2010, Junio C Hamano wrote: >> Johannes Sixt writes: >> > Previously, failures during execvp could be detected only by >> > finish_command. However, in some situations it is beneficial for the >> > parent process to know earlier that the child process will not run. >> > >> > The idea to use a pipe to signal failures to the parent process and >> > the test case were lifted from patches by Ilari Liusvaara. >> >> I wonder if we can do this without pipe, perhaps using "vfork, exec, then >> update a variable".... > > Except that some systems implement vfork(2) in terms of fork(2), I heard. As long as they implement it correctly, we don't care if it is implemented in terms of fork or knife or chopstick. The only thing we care about in this crazy-idea was the shared address space the child can write into until it exec or _exit while the parent is in a frozen state. > Moreover, I think that we do way too many things before the exec; isn't that > dangerous? Yeah, we do seem to do quite a lot more than I thought. Scratch that crazy idea, then.