Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!cmcl2!rutgers!princeton!udel!mmdf
From: ESC1332%ESOC.BITNET@wiscvm.wisc.edu (K.Keyte)
Newsgroups: comp.os.minix
Subject: A (general) word about getting new patches with the SHAR files
Message-ID: <609@louie.udel.EDU>
Date: Mon, 19-Oct-87 08:28:17 EDT
Article-I.D.: louie.609
Posted: Mon Oct 19 08:28:17 1987
Date-Received: Tue, 20-Oct-87 02:08:32 EDT
Sender: mmdf@udel.EDU
Lines: 46


There are a few problems which I've noticed concerning SHAR difference
archives/files being sent around the system over the past months.

 i.    The files are mainly created using PC-IX tools, UNIX, XENIX, but
       rarely MINIX. This obviously affects the compatibility of the
       transfers. Isn't it possible for development to be done under
       MINIX or at least _tested_ under MINIX before being distributed
       to the world (or the MINIX world in any case)?

 ii.   I've compiled my version of 'sed' (eventually after increasing
       the stack space of 'sh' and performing a manual 'asld' to get
       around the RAM disc overflow (by using an 'asld' -T. option).
       However... the 'sed' call doesn't seem to work too well. It
       truncates some of the files at strange places and leaves most
       of them completely empty. My guess is that most people using
       these SHAR files are either using a DOS utility they've written
       themselves to break them up, or are going through a long-winded
       session with some editor. It seems a pity because the SHAR files
       are designed specifically to allow easy unpacking.

 iii.  The checksums on the files are completely useless at the moment.
       This is probably due to the fact that lines over 80 characters
       are split up by most mailers, tab/space combinations go through
       numerous comversions, and heaven knows what else. The checksum
       test ALWAYS fails when I try to unSHAR my files.

 iv.   The FIX program simply calls 'strcmp', but due to the tab/space
       (etc) problem(s) mentioned in (iii) above, it nearly always fails
       the test even when it SHOULD pass. The solution is to write a
       replacement 'strcmp' which ignores differing white-space characters,
       but the solution is not ideal.


Is anyone out there finding similar problems (and I'm pretty sure you are!),
and has anyone got any solutions, ideas, etc. as to how the problem may be
solved?

Also - has anyone actually got a completely working, tested version of the
Turbo-C MINIX 1.2 for which the differences were recently distributed? If
so, I'd be grateful to hear of any problems you encountered, and how you
got around them. I'm in the very long and tedious process of applying the
differences - the problems mainly caused by the points mentioned above.


Karl Keyte