Solved: Jenkins Build agent for VMS 8.3 Alpha via RSH.

[John E. Malmberg]

It turns out that the environment variable parameters for Linux Jenkins
build agents are one of the parameters that do not take effect with out
disconnecting the build agent and reconnecting it.

I use an environment variable in the Linux Jenkins build agent to tell
it how to connect to VMS. SSH for the VMS 8.4 build agents, and rsh for
VMS 8.3 and VMS 7.3 build agents.

Which means that I never tested to see if the build worked from an SSH
session from the Jenkins build agent to OpenVMS Alpha 8.3. It was still
using RSH after the upgrade.

I had to downgrade from TCP/IP 5.7-13 ECO 4 to the TCP/IP V5.6-9
supplied with VMS 8.3 Alpha though.

* SSH from a Jenkins build agent does work to VMS/Alpha 8.3 TCP/IP
5.7-13 ECO 4 though. I only tested that combination after I implemented
the solution for RSH. As each complete test build is just under 3
hours, I did not take the time to test it, since I could not use that
TCP version on VMS 8.3.

* NFS Client of TCP/IP 5.7-13 ECO 4 does not work correctly on VMS/Alpha
8.3. During my build writes to the the NFS share were failing.

%COPY-E-WRITEERR, error writing
JSTAGE_ROOT:[kit]GNV-AXPVMS-GZIP-V0107--1.PCSI;1
-RMS-F-WER, file write error
-SYSTEM-F-MEDOFL, medium is offline
%COPY-W-NOTCMPLT, STAGE_ROOT:[KIT]GNV-AXPVMS-GZIP-V0107--1.PCSI;1 not
completely copied

After downgrading to TCP/IP 5.6-9 on VMS/Alpha 8.3:

(02:16:03.84) $ copy stage_root:[kit]*.* jstage_root:[kit]/log
%COPY-S-COPIED, STAGE_ROOT:[KIT]gnv-axpvms-83-gzip-v01^.07--1.zip;1
copied to JSTAGE_ROOT:[kit]gnv-axpvms-83-gzip-v01^.07--1.zip;3 (3934 blocks)
%COPY-S-COPIED, STAGE_ROOT:[KIT]GNV-AXPVMS-GZIP-V0107--1.PCSI;2 copied
to JSTAGE_ROOT:[kit]gnv-axpvms-gzip-v0107--1.pcsi;3 (13328 blocks)
%COPY-S-NEWFILES, 2 files created

The fix to get the builds working on VMS 8.3 Alpha from Jenkins control
is to have a wrapper to the build script submit the build script to
batch and then wait for the batch job to complete.

I still do not know what the environment difference is between the RSH
environment and the interative ssh environment that GNV is not happy
with. But that no longer matters.

Regards,
-John
***@qsl.net_work

[Stephen Hoffman]

Post by John E. Malmberg
It turns out that the environment variable parameters for Linux Jenkins
build agents are one of the parameters that do not take effect with out
disconnecting the build agent and reconnecting it.
...During my build writes to the the NFS share were failing.

Seems rather more like an NFS problem than one of rsh and ssh?

[John E. Malmberg]

Post by Stephen Hoffman

Post by John E. Malmberg
It turns out that the environment variable parameters for Linux
Jenkins build agents are one of the parameters that do not take effect
with out disconnecting the build agent and reconnecting it.
...During my build writes to the the NFS share were failing.

Seems rather more like an NFS problem than one of rsh and ssh?

There was an issue with rsh with VMS 8.3 originally and only for that
specific build. Other products build just fine.

I never tested if using ssh fixed the original build issue do to not
actually activating the config change.

Upgrading to TCP/IP 5.7 on VMS 8.3 Alpha introduced an NFS problem that
was not there before. That was a second independent problem from the first.

Downgrading TCP/IP on VMS 8.3 Alpha to the version that was shipped with
it made the second problem go away.

Regards,
-John
***@qsl.net_work

[Peter 'EPLAN' LANGSTOeGER]

Post by John E. Malmberg
I had to downgrade from TCP/IP 5.7-13 ECO 4 to the TCP/IP V5.6-9
supplied with VMS 8.3 Alpha though.

Why not upgrade to ECO5 instead?

[John E. Malmberg]

Post by Peter 'EPLAN' LANGSTOeGER

Post by John E. Malmberg
I had to downgrade from TCP/IP 5.7-13 ECO 4 to the TCP/IP V5.6-9
supplied with VMS 8.3 Alpha though.

Why not upgrade to ECO5 instead?

From the last time I downloaded Hobbyist kits, about a year ago, TCP/IP
5.17-13 ECO 4 was the latest available.

Are there release notes for ECO-5 that indicate this issue was fixed for
VMS 8.3 Alpha?

Regards,
-John
***@qsl.net_work

[IanD]

Post by John E. Malmberg

Post by Peter 'EPLAN' LANGSTOeGER

Post by John E. Malmberg
I had to downgrade from TCP/IP 5.7-13 ECO 4 to the TCP/IP V5.6-9
supplied with VMS 8.3 Alpha though.

Why not upgrade to ECO5 instead?

From the last time I downloaded Hobbyist kits, about a year ago, TCP/IP
5.17-13 ECO 4 was the latest available.
Are there release notes for ECO-5 that indicate this issue was fixed for
VMS 8.3 Alpha?
Regards,
-John

And this is why the Hobbyist kits and the removal of patches is in part VMS shooting itself in the foot

I can understand there are some slack bas*ards out there who will run Hobbyist licenses as production but this situation is akin to trying to stop piracy - it's never going to work and your only harming your own interests in the long run as your seen as a stumbling block by your customers (look at how the likes of the MPAA etc are hated and seen as alienating their customer base)

For someone like yourself that puts effort into making open source available to the greater VMS community, one would think that the likes of HP/VSI would be busting down your door to assist *sigh*

On a side note, my only gripe with the current round of open source 'stuff' is that it targets people who have access to environments where one can build the targets of the kits.
I was looking at GAWK and I would love to try the latest version on IA64 but the release is not in EXE format and I don't have access to a C compiler :-(
I know there is ramifications around people just picking up and running an EXE but not everyone has the ability to build the environments or EXE's for themselves - just saying - it's not meant to be a whinge :-)

I wouldn't care if VSI added something into a hobbyist kit that collected data for verifying the hobbyist system isn't being used in a production environment (how you would measure that specifically I don't know). It ultimately comes down to trust and forging a relationship with your user base. Keep it impersonal and punitive and restrictive and your customers will view you the same way as well...

I do however think that the current no patch policy is actually hurting VMS more than it is helping and your case is a classic example as to why

If you want people to take up paid support, then add the value. The value isn't in the product purchase, that model died and went to IT hell a long time ago. I see they have options for things like regular service checks, regular tune up's etc - this is more like it - give something for that service contract.

Customers stopped being a one size fits all representation long ago. In previous places I worked in, they were moving to a model where testing / development ran totally unsupported (not VMS!).
They really didn't care or want to pay for support in these environments. Production was another matter, they paid top dollar in that arena. Linux cleans up here yet again, basically free to acquire and you pay just for the service if you want it

[Craig A. Berry]

Post by IanD
I was looking at GAWK and I would love to try the latest version on IA64 but
the release is not in EXE format and I don't have access to a C compiler :-(

As far as I can see the latest release is 4.1.4 and that's what you get at:

<https://sourceforge.net/projects/gnv/files/gawk/>

[John Reagan]

Post by IanD
For someone like yourself that puts effort into making open source available to the greater VMS community, one would think that the likes of HP/VSI would be busting down your door to assist *sigh*

I routinely trade email with John. I actively use GNV. I keep close watch on their Mercurial checkins. I have all of John's bug reports and suggestions in my active email folders for future work.

Post by IanD
On a side note, my only gripe with the current round of open source 'stuff' is that it targets people who have access to environments where one can build the targets of the kits.
I was looking at GAWK and I would love to try the latest version on IA64 but the release is not in EXE format and I don't have access to a C compiler :-(
I know there is ramifications around people just picking up and running an EXE but not everyone has the ability to build the environments or EXE's for themselves - just saying - it's not meant to be a whinge :-)

So you want to build open source code on your non-hobbyist system but don't have a C compiler, did that I get that right? Does the C compiler have to be free or just low cost? (yes, I know they are free on UNIX platforms but Visual Studio isn't free) What about a compiler with limited functionality? No listings (you don't get those on UNIX :) )? Certain pragmas ignored? /NOOPT only? Only certain /STANDARD values accepted? Anything like this make sense?

[Chris Scheers]

Post by John Reagan

Post by IanD
For someone like yourself that puts effort into making open source available to the greater VMS community, one would think that the likes of HP/VSI would be busting down your door to assist *sigh*

I routinely trade email with John. I actively use GNV. I keep close watch on their Mercurial checkins. I have all of John's bug reports and suggestions in my active email folders for future work.

Post by IanD
On a side note, my only gripe with the current round of open source 'stuff' is that it targets people who have access to environments where one can build the targets of the kits.
I was looking at GAWK and I would love to try the latest version on IA64 but the release is not in EXE format and I don't have access to a C compiler :-(
I know there is ramifications around people just picking up and running an EXE but not everyone has the ability to build the environments or EXE's for themselves - just saying - it's not meant to be a whinge :-)

So you want to build open source code on your non-hobbyist system but don't have a C compiler, did that I get that right? Does the C compiler have to be free or just low cost? (yes, I know they are free on UNIX platforms but Visual Studio isn't free) What about a compiler with limited functionality? No listings (you don't get those on UNIX :) )? Certain pragmas ignored? /NOOPT only? Only certain /STANDARD values accepted? Anything like this make sense?

RE: Visual Studio

That's not entirely true. I believe that the Visual Studio Community
edition is free. One of the documented targets for this version is open
source developers.

Also, if you don't need the IDE, Microsoft provides the C compiler and
linker in various versions of the DDK and PDK.

[Steven Schweda]

Post by John Reagan
Does the C compiler have to be free or just low cost?

Define "low". How much should I pay for the privilege of
enhancing your OS?

Post by John Reagan
[...] but Visual Studio isn't free)

In my experience, it's free enough. I have Visual Studio
Express (or at least Visual C++ Express) in versions from
2005 through 2015 on my Windows (10) system, and it does all
I need to build the Info-ZIP programs on Windows. No doubt
the non-free editions provide some highly valuable features,
but basic C[++] build capability seems free to me. I gather
that there's some new non-"Express"-labeled thing, too, now.

Post by John Reagan
[...] What about a compiler with limited functionality? No
listings (you don't get those on UNIX :) )? Certain pragmas
ignored? /NOOPT only? Only certain /STANDARD values
accepted? Anything like this make sense?

Perhaps. HP-UX has a tradition of hobbled "bundled" C
compilers. The one on IA64 even accepts function prototypes.
It's not immediately clear that it'd be worth the bother,
however.

[David Froble]

Post by Steven Schweda

Post by John Reagan
Does the C compiler have to be free or just low cost?

Define "low". How much should I pay for the privilege of
enhancing your OS?

Very valid point.

If seeking developers, trying to charge them for their efforts won't end well.

[Stephen Hoffman]

Post by John Reagan
Does the C compiler have to be free or just low cost? (yes, I know
they are free on UNIX platforms....

The tools are free on the platforms that VSI is primarily competing
with for wholly new deployments, and for wholly new customers.

VSI does subsidize application development tools for various folks, as
HPE once also offered. Just not for the commercial end-users of the
platform.

Post by John Reagan
but Visual Studio isn't free)

As other replies have mentioned, Microsoft Visual Studio Community
edition (on Windows) and Microsoft Xamarin (on macOS) are both free:
https://www.visualstudio.com/free-developer-offers/

Clang for Windows has been around for a while, too:
https://blogs.msdn.microsoft.com/vcblog/2015/05/01/bringing-clang-to-windows/

Post by John Reagan
What about a compiler with limited functionality? No listings (you
don't get those on UNIX :) )?

You're right! Hadn't particularly noticed the omission of listing
output though, given both the IDE I'm commonly using (Xcode for macOS,
based on clang) and — if I really need it — preprocessor output and
generated assembler output is available, and diagnostic output where
that's really needed — though that diagnostic output feeds back into
the IDE without having to deal with the format. I find the
as-the-code-is-input "continuous" compilation capability of Xcode is
really quite nice, too.

Post by John Reagan
Certain pragmas ignored? /NOOPT only? Only certain /STANDARD values
accepted? Anything like this make sense?

Maybe we could fill the extra slots in the compiler "backplane" with
epoxy, as DEC once did? Or embed busy-work — no-ops or otherwise —
sufficient to slow down application performance, as DEC also once did.
Is this whole pricing and packaging and performance and features
adventure going to be measured in units of bullets-per-foot? But
seriously... Tough spot. The development tools have to be built and
tested and supported, certainly. How the associated costs are
accounted for does vary. With the size of the OpenVMS installed
base, there are fewer opportunities to amortize and to distribute the
costs across the customer base, too.

[Arne Vajhøj]

Post by John Reagan
So you want to build open source code on your non-hobbyist system but
don't have a C compiler, did that I get that right? Does the C
compiler have to be free or just low cost? (yes, I know they are
free on UNIX platforms but Visual Studio isn't free) What about a
compiler with limited functionality? No listings (you don't get
those on UNIX :) )? Certain pragmas ignored? /NOOPT only? Only
certain /STANDARD values accepted? Anything like this make sense?

I think the business plan of charging for compilers is practically dead
and charging for more elaborate development tools is going the same way
soon.

MS C/C++ command line compiler is free

Visual Studio Community Edition IDE is free

the higher versions of Visual Studio cost money, but MS will have
problems with that going forward because the majority of IDE's for other
platforms are free

I know that put VSI in a tough spot.

IBM, Oracle, Redhat etc. make so much money that giving
development tools away is no big deal.

VSI is a smaller company and needs revenue.

But the market situation is as it is.

Arne

[Craig A. Berry]

Post by John Reagan
So you want to build open source code on your non-hobbyist system but
don't have a C compiler, did that I get that right?

I think his problem is that he doesn't have an IA64 hobbyist system, so
what he can build on his hobbyist system isn't something he can turn
around and use at the day job. The long-term solution to this is to get
everybody on "hardware" that anyone can easily get and maintain, i.e.
x86_64 virtual or otherwise.

Post by John Reagan
Does the C
compiler have to be free or just low cost? (yes, I know they are
free on UNIX platforms but Visual Studio isn't free) What about a
compiler with limited functionality? No listings (you don't get
those on UNIX :) )? Certain pragmas ignored? /NOOPT only? Only
certain /STANDARD values accepted? Anything like this make sense?

This sounds like an extra administrative cost for small benefit. Give
Eisner an Itanium sibling for now and let the x86_64 port take care of
the rest.

[David Froble]

Post by IanD

Post by John E. Malmberg

Post by Peter 'EPLAN' LANGSTOeGER

Post by John E. Malmberg
I had to downgrade from TCP/IP 5.7-13 ECO 4 to the TCP/IP V5.6-9
supplied with VMS 8.3 Alpha though.

Why not upgrade to ECO5 instead?

From the last time I downloaded Hobbyist kits, about a year ago, TCP/IP
5.17-13 ECO 4 was the latest available.
Are there release notes for ECO-5 that indicate this issue was fixed for
VMS 8.3 Alpha?
Regards,
-John

And this is why the Hobbyist kits and the removal of patches is in part VMS shooting itself in the foot
I can understand there are some slack bas*ards out there who will run Hobbyist licenses as production but this situation is akin to trying to stop piracy - it's never going to work and your only harming your own interests in the long run as your seen as a stumbling block by your customers (look at how the likes of the MPAA etc are hated and seen as alienating their customer base)

Now, I am not all knowing, (almost but not quite), but I'm going to say that few
people using VMS in a commercial environment will try to avoid support payments.
Why? It's simple, you need to support your vendor, if you're found out there
can be substantial penalties, enforceable in court, and would you really want
your company to get the bad publicity of being consider a thief?

I've worked with companies who demanded periodic audits of all software to
insure they were legal.

While there may be some looking to go cheap, most times there will be
responsible people in the company who will point out one or more problems with
doing so, and responsible management will demand to be "legal".

And even if there is a few cheats, their staff will be working with VMS, and
cheats usually don't retain people very well.

[John E. Malmberg]

Post by IanD

Post by John E. Malmberg

Post by Peter 'EPLAN' LANGSTOeGER

Post by John E. Malmberg
I had to downgrade from TCP/IP 5.7-13 ECO 4 to the TCP/IP V5.6-9
supplied with VMS 8.3 Alpha though.

Why not upgrade to ECO5 instead?

From the last time I downloaded Hobbyist kits, about a year ago, TCP/IP
5.17-13 ECO 4 was the latest available.
Are there release notes for ECO-5 that indicate this issue was fixed for
VMS 8.3 Alpha?

I have discovered that I actually had the TCP/IP ECO 5 kit in the
hobbyists downloads of 1 year ago. The file is a .ZIP not the same
extensions as the other kits, so I missed it when I did the install.

But the systems appear stable now, so I am not going to change anything
on the VMS 8.3 system.

Post by IanD
On a side note, my only gripe with the current round of open source
'stuff' is that it targets people who have access to environments
where one can build the targets of the kits.

What is the current round that you write of?

Post by IanD
I was looking at GAWK and I would love to try the latest version on
IA64 but the release is not in EXE format and I don't have access to
a C compiler I know there is ramifications around people just picking
up and running an EXE but not everyone has the ability to build the
environments or EXE's for themselves - just saying - it's not meant
to be a whinge

The IA64 hobbyist kits include a C and other compilers.

* The updated GNV packages are all PCSI kits that include binaries.

* They work for both DCL and GNV mode.

* Some of them like GAWK have a DCL mode that is auto detected.

* Some of them like GAWK have a DCL .CLD file so that they can be
installed as a DCL command.

* The older GNV package is no longer required to use the new GNV
packages to run most configure and make scripts to build new packages.
If you do install the older GNV package, the newer packages must be
reinstalled.

Gawk on VMS is being built with the self-tests run automatically by my
Jenkins system with in 24 hours of an update to the gawk repository.
I do not make those kits available, because I have not taken the time to
set up them up as "pre-release" builds with a different kit name and
destination directory. Just one of the side projects that I have not
had time for.

Last development Gawk build was 3.5 days ago.
* VMS 8.3 Alpha on FreeAXP takes 1 hour 24 minutes to build/test
* VMS 8.4 Alpha on 600 Mhz DS-10 IDE disks takes 10 minutes
* VMS 8.4 Alpha on 2 CPU 1.6 Ghz RX 2600 takes 13 minutes
* VMS 7.3 VAX on SimH LXD Container takes 48 minutes.

Regards,
-John
***@qsl.net_work

[hb]

Post by John E. Malmberg
I have discovered that I actually had the TCP/IP ECO 5 kit in the
hobbyists downloads of 1 year ago. The file is a .ZIP not the same
extensions as the other kits, so I missed it when I did the install.

The ssh_575g_i64.zip, containing TCPIP-SSH-IA64_V57-ECO5G_2015-11-27.BCK
containing Alpha images as well, as mentioned here
https://groups.google.com/forum/#!msg/comp.os.vms/a_oux_0E6zc/Fw5u6AOJBgAJ
?

Post by John E. Malmberg
* VMS 8.4 Alpha on 2 CPU 1.6 Ghz RX 2600 takes 13 minutes

^^^^^ ^^^^^^^ ?

[John E. Malmberg]

Post by hb

Post by John E. Malmberg
I have discovered that I actually had the TCP/IP ECO 5 kit in the
hobbyists downloads of 1 year ago. The file is a .ZIP not the same
extensions as the other kits, so I missed it when I did the install.

The ssh_575g_i64.zip, containing TCPIP-SSH-IA64_V57-ECO5G_2015-11-27.BCK
containing Alpha images as well, as mentioned here
https://groups.google.com/forum/#!msg/comp.os.vms/a_oux_0E6zc/Fw5u6AOJBgAJ
?

Cut/paste error.

Post by hb

Post by John E. Malmberg
* VMS 8.4 I64 on 2 CPU 1.6 Ghz RX 2600 takes 13 minutes

Regards,
-John