[comp.bugs.sys5] NCR's symbolic links

ccdn@levels.sait.edu.au (DAVID NEWALL) (08/01/89)

In article <737@toro.UUCP>, nick@toro.UUCP (Nicholas Jacobs) writes:
> In article <485@umigw.MIAMI.EDU> aem@Mthvax.CS.Miami.Edu writes:
>>greg@tcnz2.tcnz.co.nz writes:
>>System V only has hard links. If you modify a file, the link is destroyed.
>>(I'm probably explaining this badly).
>
> Actually, NCR has implemented symbolic links for System V.

NCR's implementation of symbolic links is almost unusable.  There is no
way of telling if a link is a symbolic link or a hard link; and most
programs break when they come across a symbolic link that points to a
non-existant file (can't stat xxxx).

A related problem: A symbolic link that points to a non-existant file
stops ftw(3) from walking the file tree.  I don't think that should
happen.  It seems to me that ftw(3) is supposed to continue walking,
even after a file that it can't stat(2).

Grump.  (Hello NCR, Columbia.  Are you listening?)


David Newall                     Phone:  +61 8 343 3160
Unix Systems Programmer          Fax:    +61 8 349 6939
Academic Computing Service       E-mail: ccdn@levels.sait.oz.au
SA Institute of Technology       Post:   The Levels, South Australia, 5095

greg@tcnz2.tcnz.co.nz (super) (08/05/89)

In article <966@levels.sait.edu.au> ccdn@levels.sait.edu.au (DAVID NEWALL) writes:
>In article <737@toro.UUCP>, nick@toro.UUCP (Nicholas Jacobs) writes:
>> Actually, NCR has implemented symbolic links for System V.
>
>NCR's implementation of symbolic links is almost unusable.  There is no
>way of telling if a link is a symbolic link or a hard link; and most
>programs break when they come across a symbolic link that points to a
>non-existant file (can't stat xxxx).

A friend of mine (wdr@csoft.co.nz) wrote a small number of utilities
to get around these problems, including one to cruise down the file
system and generate a script which could rebuild the current symbolic
links.

The major headache we had with them is that our dbms appeared to do
file locking on the filessystem:inode basis, but did not handle
symbolic links, so that two programs could lock the same data if they
were running in the DBMS on different file systems. We have not had 
the spate of DB corruptions we used to have since the last upgrade of
the software, but it is something to watch for.

greg
Disclaimer : no-one tells me what to write
Quote :   Have I aligned
            with a blown mind
              Wasted my time
                On a drawn blind -- Zappa

Greg Calkin                                   Thomas Cook N.Z. Limited,
...!uunet!vuwcomp!dsiramd!marcamd!tcnz2!greg  PO Box 24, Auckland CPO,
or greg@tcnz.co.nz                            New Zealand.    Phone (09)-793920

jba@harald.ruc.dk (Jan B. Andersen) (08/05/89)

greg@tcnz2.tcnz.co.nz (super) writes:

>In article <966@levels.sait.edu.au> ccdn@levels.sait.edu.au (DAVID NEWALL) writes:
>>In article <737@toro.UUCP>, nick@toro.UUCP (Nicholas Jacobs) writes:
>>> Actually, NCR has implemented symbolic links for System V.
>>
>>NCR's implementation of symbolic links is almost unusable.  There is no
>>way of telling if a link is a symbolic link or a hard link;...

Even then, the latest man page for /usr/ucb/ls (Page 1, Feb. 1989) still
insists on, that -F will mark symbolic links with '@'. Does it work? NO
Does the assembler command swbeg generate correct code? NO
Can you put a comment line after swbeg? NO
Can you change a register value with adb? NO
Does the C modifier work? NO
Is there a man page for the alternative assembler? NO
Can you write assembler on a TOWER? ????
-- 
Jan B. Andersen                              ("SIMULA does it with CLASS")