On Tuesday 30 June 2009 12:48:31 Theodore Tso wrote: > On Tue, Jun 30, 2009 at 05:44:02PM +0200, Kay Sievers wrote: > > Well, the "nice" part of it is usually not nice enough to justify to > > have libraries to magically fork() processes and let the possible > > non-aware user of the library receive signals and all that stuff. It > > gets even worse with multithreaded apps using many libs. > > So you are basically worried about broken applications which actually > pay attention to SIGCHILD and uses wait(2) instead of waitpid(2)? are those applications really broken ? if an app is only forking one or two children, then assuming they're the only ones doing fork() isnt really a bad assumption. as long as the library (interface/documentation) is explicit about when a function may fork() the daemon, i think it's fine. -mike
Attachment:
signature.asc
Description: This is a digitally signed message part.
