On 2004.03.01 15:32 J. Ali Harlow wrote:
On 2004.03.01 13:35 J. Ali Harlow wrote:Owen, I've now implemented this. I've tested with child-test and spawn-test under Linux, and child-test under WIN32. Unfortunately, spawn-testdoesn't work for me under WIN32 (this is before my changes), so I'm not able to test this bit. Something to worry about later, I think.Of course, it would help if I actually attached the patch.
Tor/Hans, To put you in the picture: Owen and I discussed this on IRC and agreed that the fact that my patch caused the GPid passed to the callback to differ from the caller's pid mean that this was not such a good idea. Instead we've reverted to the original implementation with some notes in the documentation to explain that the callback function would be the normal place for the caller to close the handle. We've left my changes to g_spawn pretty much as is so applications can (and should) use g_spawn_close_pid() to close pids returned by g_spawn. I dearly hope I haven't broken anything. Cheers, Ali.