POSIX for OpenVMS: Anyone using it?
Frederick W. Verwijs
This post is out of curiosity. Is anybody building/maintaining a system
on Posix for OpenVMS? Particularly in C. Have you had DEC support?
Do you have manuals for POSIX and C under POSIX? Is this platform
effective?
I'm not asking for help from you. I'd just like to know if anybody is
using this platform.
regards,
Eric Verwijs
ver...@esltd.com or
ah...@freenet.carleton.ca
--
"Black flies, the little black flies.
Always the black flies no matter where you go.
I'll die with a black fly picking my bones
In North Ontario-io in North Ontario."
Black Fly Song by Wade Hemsworth
Nigel Campbell
: Hello out there,
: This post is out of curiosity. Is anybody building/maintaining a system
: on Posix for OpenVMS? Particularly in C. Have you had DEC support?
Did the POSIX support extend to adding a korn shell? Is it bundled or
an option on OpenVMS?
--
Nigel Campbell Voice: (613) 738-1338 ext 3016 P.O. Box 9707,Stn.T
Business Intelligence FAX: (613) 738-0002 3755 Riverside Dr.
Cognos Incorporated Internet: nig...@cognos.com Ottawa ON K1G 4K9
CANADA
Stephen Hoffman
In article <3ret1q$c...@gatekeeper.cognos.com>, nig...@Cognos.COM (Nigel Campbell) writes:
:Frederick W. Verwijs (verwijs@$TIN_DOMAIN) wrote:
<snip>
:: This post is out of curiosity. Is anybody building/maintaining a system
:: on Posix for OpenVMS? Particularly in C. Have you had DEC support?
(I make regular use of various POSIX utilities. But given my employer,
I probably don't count in your survey. :-)
:Did the POSIX support extend to adding a korn shell? Is it bundled or
:an option on OpenVMS?
POSIX includes the POSIX shell. (There may be a ksh around -- I have
a recollection of seeing references to a POSIX ksh.) OpenVMS includes
a license to load and use POSIX; the POSIX kit is installed seperately.
(The POSIX kit is on the consolidated software distribution.)
------------------------------ Opinionative -------------------------------
Stephen Hoffman OpenVMS Engineering hof...@xdelta.enet.dec.com
---------------------------------------------------------------------------
David Woodruff, MIT Lab for Nuclear Science
>This post is out of curiosity. Is anybody building/maintaining a system
>on Posix for OpenVMS? Particularly in C. Have you had DEC support?
>Do you have manuals for POSIX and C under POSIX? Is this platform
>effective?
>
One of our user groups is using posix as a way of using rcs. Another
individual is using posix to compile and link C programs since he is much
more familiar with unix.
We don't have manuals for posix or C, but I have made the bookreader
documentation available online. We haven't had DEC support for this,
but it is available if we want it.
The posix platform is quite effective... at least I haven't gotten any
complaints!
*******************************************************************************
"Call me Ishmael. Some years ago-- never | David S. Woodruff 26-402d
mind how long precisely-- having little | MIT Lab for Nuclear Science
or no money in my purse, and nothing | 77 Mass. Ave, Cambridge MA, 02139
particular to interest me on shore, I | (617) 253-6943
thought I would sail about a little and |
see the watery part of the world." | d...@mitlns.mit.edu MITLNS::DSW
Tony Scandora 708-252-7541
>This post is out of curiosity. Is anybody building/maintaining a system
>on Posix for OpenVMS? Particularly in C. Have you had DEC support?
>Do you have manuals for POSIX and C under POSIX? Is this platform
>effective?
A VMS license includes the right to run POSIX, which must be installed
separately. Most of it is there, but we have found it useless.
Performance is a joke. Emulation of POSIX on VMS, though substantially
correct, is slow beyond belief. A $ POSIX command takes between five and
ten seconds on a VAX 6400 to deliver its first psx> prompt. During that
time, you can not type ahead, as the terminal input buffer is flushed.
Once in the POSIX environment, there are widely varying degrees of slowness
depending on the application.
There is an impenetrable wall between the VMS and the POSIX environments.
A program linked for VMS can not be run from the POSIX shell, and a program
linked for POSIX can not be run from DCL. Technically, it is possible, but
only by spawning a subprocess in the other environment, which can be
unbearably slow, and can't do wildcards:
$ posix "grep kurtosis *.*"
%POSIX-F-PARSEFAIL, error parsing grep kurtosis *.*
-LIB-E-NOWILD, no wildcard permitted
The first job I tried to do on VMS POSIX was extract the files from a tar
archive of a GNU software distribution. That job is impossible. Like much
of the UNIX environment, most GNU directory names have dots in them, and
many file names have two dots. Since these are illegal in VMS directory
and file names, VMS POSIX simply refuses to deal with them and dies. The
late, unlamented DEC/Shell's tar changed excess dots and other special
characters to underscores and did a fine job of extracting files from GNU
tar archives. When I pointed out in an SPR that the old DEC/Shell's tar
did something reasonable but VMS POSIX's tar chokes on excess dots, the
reply was VMS doesn't support excess dots, so VMS POSIX did the right
thing. Fine. I can't extract files from the overwhelming majority of
freeware software distributions, but VMS POSIX is correct, so I have
nothing to complain about.
The m4 macro processor is part of many UNIX systems, but not VMS POSIX. I
got the GNU distribution, unpacked it into a genuine UNIX filesystem in a
container file, and tried to build it. The configure script failed, as did
every other GNU configure script I tried. I didn't bother complaining
about it, as I didn't want to waste my time getting another we don't care
about compatibility with the world, we only care about being correct
answer.
VMS POSIX has no man pages on our system. I have not checked to see if
they are an optional part of the distribution.
The only real work we have done with VMS POSIX is compiling yacc grammar.
We could use GNU's BISON, but its license terms are more restrictive than
the standard GNU terms, so we stay away from it. We compile yacc grammars
in VMS POSIX, save the .C files, and then compile them in the native VMS
environment.
If you want to wait five to ten seconds before getting a psx> prompt, you
can actually do some work, however, if you only need grep and a few of the
best UNIX utilities, you can get native VMS versions of GNU grep and many
other standard UNIX utilities, which can be run in a VMS process.
In short, it's there, it's included with the VMS license, but it's not
close to ready for prime time.
Tony Scandora, Argonne National Lab, 708-252-7541
scan...@cmt.anl.gov or scan...@anlcmt.bitnet
Michael Hjorth
<snip>
: There is an impenetrable wall between the VMS and the POSIX environments.
: A program linked for VMS can not be run from the POSIX shell, and a program
: linked for POSIX can not be run from DCL. Technically, it is possible, but
Which version of POSIX are you using? The newest version 2.0 has no
problems with this, I can run most executables from POSIX as well as DCL.
The only problem I've got is if I call the TPU editor it won't work
in POSIX. (It would increase the usability if the TPU editor was
available from POSIX, but anyway if porting is the general concern
it should of course not be used).
<snip>
: VMS POSIX has no man pages on our system. I have not checked to see if
: they are an optional part of the distribution.
Again, we have all the man pages in POSIX 2.0, no extra installation.
: In short, it's there, it's included with the VMS license, but it's not
: close to ready for prime time.
I think it works fine, we can use a number UNIX commands. But we will
of course not use a VMS system to do UNIX work.
(PS. the system is DEC 2000/500 with OpenVMS/AXP 6.1)
Michael.
--
\|/
o o
+---------------------------------+--------------------------+----m--~--m----+
| Michael Hjorth | mic...@skemi.klb.dtu.dk |---------------|
| Chemistry Department B | m...@tkemi.klb.dtu.dk |---------------|
| Technical University of Denmark | |---------------|
| DK-2800 Lyngby | +45 45881777 ext. 2026 |---------------|
| Denmark | +45 45252026 direct |---------------|
+---------------------------------+--------------------------+---------------+
Menneskene synes ikke at have modtaget sproget for at kunne skjule tankerne,
men for at kunne skjule, at de ingen har. Søren Kirkegaard.
Joe Pollizzi
I'd have to admit that my own personal experience with the POSIX interface
is much like what Tony mentioned previously. (I did check, and we are
using version 2.)
Perhaps my real problem is that POSIX simply does not do the "same" things as
my OSF and SunOS environments. We don't have the inclination or time
for system managers to jump into setting up the shells and tools under POSIX
that we typically have under UNIX.
But the real thrust of my entry into this dialog is "Does anyone really
care anymore?"
With Digital's announcement of its intention to fully support NT
server applications under VMS, as well as their already healthy
investment in a premier UNIX implementation, do we care if VMS
implements spec 1170? If we do, then we better gather together a
reasonable voice to argue for it and help Digital understand what our
expectations are for this facility.
--
****************************************************************************
*Joe Pollizzi poll...@stsci.edu Caution - ex Techie who still *
*Space Telescope Science Institute thinks he knows what he's talking *
*Baltimore, MD. about. *
ever...@star.enet.dec.com
I've used VMS POSIX to compile gnu tar (gtar); also various other
things. I've had no problems running posix symbols from normal DCL
(the syntax was a bit weird, but not hard.) Most off-putting thing I
found was that "$" is a shell metacharacter and has to be escaped,
a giant pain in the ***...but one that appears in unix boxes also.
I had some grief getting gtar linked and had to do that "by hand"
linking the posix runtime bits in (this was with Posix 1.1; 2.0 may
have fixed it) where the "cc" command wasn't getting all the way to
linking without a few undefineds. However, the resulting image worked
very nicely and the bulk of the build was the unix makefile.
On the whole the level of work I experienced was no different from
that I've had porting random unix code to sunos. Nor did I find
execution particularly slow, though again I had no applications that
forked a lot.
I suspect that I could have gained considerable speed by replacing the
RMS usage by something as low-level as unix uses (i.e., a thin layer
over virtual $qio to manipulate sequential files), and the Posix 1.1
implementation of pipes passed thru disk. However, the new file system
will automatically make many small files go thru memory only, whether
they're programmed to write scratch disk files or not, if they are
transient, and Posix 2.0 does seem more responsive. The man pages are
almost identical to what you see in unix, and roughly as fast to get to
as far as I can tell (by the "seat of the pants").
On large intermediate files I have noticed posix tar, for example,
seemed slower than vmstar, and the container file system is a crock
(which I believe is going away [but mind I'm not developing Posix...only
using it some]). Still, speed is improving and the system is useful
enough that I use some Posix symbols fairly frequently. (For example
"ls -lrt" displays files in reverse chronological order; you can arrange
to pass that from a DCL symbol pretty simply.)
MY recent new product development has not used posix, but that has more
to do with the nature of those developments (HSM, security systems,
undelete systems, etc.) in VMS than with Posix. The choice not to use
posix for other stuff would tend to have more to do with the erroneous
perception that it's an extra-cost item which leads to it not being
widely installed than with the merits of the package.
glenn