Re: odd ATF failure for sh: ulimit_redirection_interaction failed

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Fri, 12 Mar 2021 21:38:54 -0800
    From:        "Greg A. Woods" <woods%planix.ca@localhost>
    Message-ID:  <m1lKwzG-0036x9C@more.local>

Gred quoting Mouse quoting Greg:
  | > > The attempted FD is #18, but the error just says "1", not "18":

I will look into the content/format of the
error message, but that's not the main issue.

Greg quoting Mouse:
  | > Does >&N _ever_ work for multi-digit N?

Yes.   It is a NetBSD sh extension (an old one)
neither required nor forbidden by POSIX.  If
the test is written properly (I forget) it
will check fir support in the shell being
tested before running thise tests.

Just Greg:
  | Well, the test does appear to execute without
  | error IFF the condition the test is meant
  | to exercise is not enforced

Something is odd in your environment somehow,
but I am not sure what at the minute.

  | (i.e. if the ulimit for opee FDs is not kept
  | lower than the number of currently open FDs);

That limit should only apply to things which
create fds (open/creat/dup/...) not to uses of
the fd once opened.  That is what this particular
test (or that part of it) is testing.  Avoiding
that by not lowering the limit just makes the
test a waste of time.

  | though I have not done any other kind of
  | test to be sure the data sent to es
  | multi-digit FD is actually received
  | from the given FD.

That works, or should,  The fd_redirections sub-test
of t_redir should verify that

kre