Hi Iain, Am 2012-03-19 10:33, schrieb Iain Hibbert: > On Sun, 18 Mar 2012, Hendrik Sattler wrote: >> Am Mittwoch, 14. März 2012, 15:44:01 schrieb Iain Hibbert: >> > To be honest, any new release (1.6?) with the current changes >> would be >> > welcomed >> > >> > As to the API, then yes I do agree that there are some problems. I >> would >> > not choose to use openobex if I were to write an OBEX program, not >> least >> > because the documentation is minimal. I have worked on software >> using >> > openobex and found that it was really difficult to work with. >> However, I >> > don't really have any better ideas :) >> >> For a start, I submitted some additional patches to my master >> branch. >> Can you check them out and do some compile and runtime tests for >> your BSDs? >> If everything goes well, the current state will be version 1.6. >> >> I compile tested it on Linux, Windows cross-compiled on Linux and >> Windows. > > No time to investigate further just now can look at it later, but > NetBSD > has SOCK_CLOEXEC but not accept4() .. we have a paccept(2) instead > which > also takes a sigmask, like so: > > int paccept(int s, struct sockaddr * restrict addr, socklen t * > restrict addrlen, const sigset t * restrict sigmask, int flags); > > You could do some autoconfig detection for this but its probably just > simpler to use fcntl all the time.. I mean, the overhead of the > extra > syscall is not likely to be significant in this environment, is it?
In a multi-threaded environment, it is useful to not leak file descriptors. fcntl() helps but is not race-free without a lot of hassle. The non-blocking-io flag (in a non-included later patch) doesn't have this problem. > I don't think any of the other BSDs have SOCK_CLOEXEC or accept4() or > paccept() so that shouldn't be a problem for them I will add paccept() support. NetBSD is easily detected using http://sourceforge.net/apps/mediawiki/predef/index.php?title=Operating_Systems#NetBSD and I add an accept4() function wrapper: #if defined(SOCK_CLOEXEC) && defined(__NetBSD__) static __inline int accept4(int s, struct sockaddr * addr, socklen t *addrlen, int flags) { return paccept(s, addr, addrlen, NULL, flags); } #endif It seems that SOCK_CLOEXEC and paccept() appeared at the same time (NetBSD 6.0), so this should be sufficiently safe. Thanks for the hint :-) Regards, HS ------------------------------------------------------------------------------ This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure _______________________________________________ Openobex-users mailing list Openobex-users@lists.sourceforge.net http://lists.sourceforge.net/lists/listinfo/openobex-users
