[Info-vax] VMS survivability

[Craig A. Berry]

On 2/20/23 6:11 AM, Dan Cross wrote:

> There is an ongoing effort to port LLVM to VMS to get native
> compiler support, but with the existing language front-ends in
> addition to clang etc, correct?  Where is that work happening?

At VSI, obviously.  There have been numerous public presentations about
what is being done and how it is being done.  The head of the compiler
group chimes in here pretty frequently with updates and explanations.  I
believe the clang++ port is in field test.  The customer portal now has
a download of "VSI C X7.4-725" (a different compiler from the clang++
compiler).  I forget exactly what the "X" stands for, but if it's not a
"V" in that position it's not a production-quality release.  As far as I
know, this is the first release of a native compiler based on the
GEM-to-LLVM conversion.  I don't think they could open source this one
if they wanted to because they don't own the intellectual property to
the front end.

> Is there an open source repository where an outside contributer
> can work _on that port_?  If this is happening in the open, it
> does not appear to be in the LLVM repository.
> (for ref: https://groups.google.com/g/llvm-dev/c/MYfZW2DOU2I/m/q8oDU0UTAAAJ
> and https://llvm.org/devmtg/2017-10/slides/Reagan-Porting%20OpenVMS%20Using%20LLVM.pdf)

As has been said many times, there will be some changes to LLVM
submitted upstream eventually.  Anyone who expected all the compiler
work to be done completely in the open hasn't been paying attention.
You can't contribute to the next version of Apple's XCode just because
it's based on LLVM and the VSI compilers likely won't be any different.
Would it be nice if there were one compiler example available in source
form, even just for a toy language, that people could use as a template
for producing there own LLVM-based compilers?  Sure.  I'd love to see
it. That's obviously not going to be VSI's highest priority, but it
could happen.

> Similarly with contributions to the operating system itself.
> ARM is gaining ground in the server space, and it seems that a
> VMS port to ARM is likely at some point, if VMS survives.  How
> does an independent third-party contribute to that, beyond just
> asking VSI?  As Arne said earlier, VMS needs things to be done,
> but what if those things require, or would be significantly
> aided, by open sourcing the OS or large parts of the
> infrastructure?  

Many, many open source projects do not have enough contributors.  This
usually only makes the news when some critical piece of infrastructure
such as ntp or OpenSSL has a vulnerability that gets noticed, but the
basic problem is ubiquitous in open source.  Having the doors wide open
was completely ineffective at keeping those projects adequately
resourced.  So it's hard to see how open sourcing VMS would magically
allow more OS development to get done.

Consider the case of Ada, which a very few people really, really want,
but no one can make a business case to VSI that it's worth their time to
do it.  Most of the important pieces are already there in the GCC
toolchain, so interested parties could retrace the steps of ACT (who
also didn't think there was a business case to continue supporting Ada
on VMS) and produce an open source Ada compiler for VMS.  There is no
waiting on VSI or anybody else.  Anyone with the interest, the time, and
the skill can do it right now.  Why haven't they?

> Without that native clang++ port, which they don't
> seem to be able to contribute to as prerequisite, the upside for
> the OSS folks to port to VMS just doesn't seem to be there.

The availability of clang++ looks like it's happening, if much later
than anyone wanted it to be.  It remains to be seen whether there will
be sufficient pieces included that someone could port Rust or Swift or
GNAT-LLVM on top of it.

There are other things VSI can and should do (and in some cases is
already doing) to enable open source ports.  I believe there is a cmake
port expected at some point, and there has been mention of updates to
GNV.  There is ongoing work on the CRTL that needs to continue (someone
from VSI mentioned adding posix_spawn() not long ago).

> If open-sourcing VMS has been tried, then can anyone share the
> story?  And note the context now: just because it didn't work
> before doesn't mean it can't be done now.

Search for "Clair Grant" and "open source" in the archives of this
newsgroup.  I think he said "We can't open source it because we don't
own it."  I believe he also said he'd tried to open source it more than
once and it wasn't going to happen.

So if you have sufficient millions burning a hole in your pocket, feel
free to approach HPE to obtain the intellectual property so you can then
give it away for free. I don't see the context now being much different
from what it was five or ten or twenty years ago.  I'd be happy to see
it happen, but it doesn't seem very realistic and it's not a panacea for
the dwindling VMS ecosystem.