On 06.11.2016 12:54, Robert Elz wrote: > Date: Sun, 6 Nov 2016 11:53:06 +0100 > From: Kamil Rytarowski <n54%gmx.com@localhost> > Message-ID: <a89bd3c2-7607-4ff3-58c4-866f782f5c0d%gmx.com@localhost> > > | I will extend our ATF tests to validate it. > > That's good. But test for getting ECHILD from waitpid() - it is fairly > clear that is what should happen. > > I am currently testing a fix, but I won't commit it (if it works) > until I hear from the austin group people (posix spec maintainers) > on what is intended for waitid() so if needed, it can be adjusted too. > > From reading the spec, it may be that 0 is correct from waitid(), it may > be that ECHILD is correct, and if an incorrect pid is passed (which is > also ECHILD from waitpid() - test that one as well .. just specify any > 0 > pid arg to waitpid() when the process has no children) it may be that > waitid() is supposed to return EINVAL (just to be different from waitpid() ...) > > kre > Please file a PR for it. I will reference it in ATF tests - I'm preparing a combination of all possible cases. Please make sure that all t_ptrace tests still pass. I will extend t_ptrace as well, as there is an important case (in attach1 use-case) to hold parent from seeing its child to exit, before an attached debugger.
Description: OpenPGP digital signature