POSIX?

["Nicholas S Castellano" <entropy@terminator.rs.itd.umich.edu>]

>Does anyone have any idea how far away from Posix compliance we are ? All
>I seem to see in the ChangeLog is "Posix says do it this way". I don't
>suppose we're close yet, but an idea of how close would be nice. It'd
>also solve an argument :-)

It's difficult to exactly quantify this, so I'll say "somewhat close
but still a long way to go."

I really doubt the library/MiNT will ever be 100% POSIX but we should
eventually get to a point where almost any POSIX application will
compile and run without modification.

One of the difficulties in quantifying just how compatible we are is
that some of it depends on what external device drivers etc. are
loaded.  For example, with minixfs link() will work, but on the
default TOS filesystem it just fails.  Also rename("foo", "foo"); will
result in foo being deleted, which violates POSIX.  (A fix for this
should be in the next release of the library, but the fix will only
help on filesystems that support the struct stat st_ino field
properly, which means it won't work on TOSfs...).

Things I can think of offhand that need more POSIXification are the
read() and write() system calls, the locale conversion stuff (I think
we only support the "C" locale at the moment), mkfifo() (MiNT doesn't
seem to support fifos without processes attached to them, they are
deleted on the last close), a few minor job-control related things,
and we don't yet completely support termios.

Cheers,
entropy

--
entropy -- it's not just a good idea, it's the second law.
Personal mail:      entropy@gnu.ai.mit.edu
MiNT library mail:  entropy@terminator.rs.itd.umich.edu