Hello Rainer,

On Fri, 2024-05-10 at 10:20 +0200, Rainer Orth wrote:

Support for Solaris 11.3 had already been obsoleted in GCC 13. However, since the only Solaris system in the cfarm was running 11.3, I've kept
it in tree until now when both Solaris 11.4/SPARC and x86 systems have been added.

This patch actually removes the Solaris 11.3 support.

I'm not sure I like this change since Solaris 11.3 is the last version of Solaris supported by a large number of SPARC systems.

Oracle unfortunately raised the hardware baseline with Solaris 11.4 such that every system older than the SPARC T4 is no longer supported by 11.4 while 11.3 still runs perfectly fine on these machines.

I wonder why you didn't raise your concerns 1 1/2 years ago when I
announced the obsoletion of Solaris 11.3 support?

Because I wasn't subscribed to gcc-patches and I'm also only subscribed now without receiving messages due to the large message volume on this list.

The problem with announcements on developer mailing lists is usually that they usually don't reach any users. I was made aware of this change only when I checked about the recent changes to GCC Git.

While Oracle does no longer provide feature updates to Solaris 11.3, there is still LTSS security support so that users still receive security updates so that their systems are continued to be protected against vulnerabilities.

The Solaris 11.3 ESUs (Extended Support Updates) are available at a
premium only, and just contain the bare minimum of security updates,
often 6 to 9 month in between.

That's not an argument for throwing away hardware that still works perfectly fine and that still has some users.

I think Solaris 11.3 support should be kept since the resulting code removal
is not that large that it would justify dropping support for such a large userbase.

Do you have any indication on the size of the userbase? I seriously
doubt it's large beyond some hobbyists that keep the old hardware
running.

I don't have the exact numbers, no. But I know there are many users out there with pre-11.4 hardware that they still use. As you may know, there are no
11.4 SPARC desktop systems and most 11.4-capable hardware is usually very expensive.

You also seem to forget that my GCC (and LLVM) Solaris support work is
purely voluntary, done in my spare time.

Not sure what makes you think so. I'm perfectly aware of the fact that lots of people do this work in their spare time as this applies to me as well.

I'm not getting paid for my Debian work, my kernel maintenance and all the other
stuff that I'm doing either. That doesn't mean users are not allowed to ask me questions or send me comments about my work.

Keeping Solaris 11.3 support working would be much more than restoring
the removal patch:

* For each and every of my Solaris patches, I'd have to investigate if
it works on 11.3 or needs adjustments and workarounds.

* I'd also need to regularly test the result to keep things working.

I honestly don't have the time or the energy to do this, nor the
hardware required for testing Besides, I have too much on my plate
already, and rather spend it on more beneficial work.

Does Solaris support in GCC really change that often that the necessary tests cannot be run by volunteers? I'd be happy to test changes for Solaris 11.3 which can be installed inside an LDOM.

Above all, I always wonder why people insist on running ancient hardware
with an almost-unsupported OS, but require a bleeding edge version of
GCC. What's wrong with continuing to use GCC 13 (or even 14, although I haven't tested that on Solaris 11.3) instead?

You could also ask why people use operating systems other than Linux and architectures other than x86_64. I don't think you will get a satisfactory answer to that question.

Removing Solaris 11.3 support might make sense in the future when SPARC support in Illumos has matured enough that people can switch over their machines.

As has been noted, SPARC is on its way out for Illumos.

Which makes my point to keep Solaris 11.3 support even more valid.

Adrian

--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer
`. `' Physicist
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi John,

Support for Solaris 11.3 had already been obsoleted in GCC 13. However,
since the only Solaris system in the cfarm was running 11.3, I've kept
it in tree until now when both Solaris 11.4/SPARC and x86 systems have
been added.

This patch actually removes the Solaris 11.3 support.

I'm not sure I like this change since Solaris 11.3 is the last version of Solaris supported by a large number of SPARC systems.

Oracle unfortunately raised the hardware baseline with Solaris 11.4 such
that every system older than the SPARC T4 is no longer supported by 11.4 while 11.3 still runs perfectly fine on these machines.

I wonder why you didn't raise your concerns 1 1/2 years ago when I
announced the obsoletion of Solaris 11.3 support?

While Oracle does no longer provide feature updates to Solaris 11.3, there
is still LTSS security support so that users still receive security updates so that their systems are continued to be protected against vulnerabilities.

The Solaris 11.3 ESUs (Extended Support Updates) are available at a
premium only, and just contain the bare minimum of security updates,
often 6 to 9 month in between.

I think Solaris 11.3 support should be kept since the resulting code removal is not that large that it would justify dropping support for such a large userbase.

Do you have any indication on the size of the userbase? I seriously
doubt it's large beyond some hobbyists that keep the old hardware
running.

You also seem to forget that my GCC (and LLVM) Solaris support work is
purely voluntary, done in my spare time.

Keeping Solaris 11.3 support working would be much more than restoring
the removal patch:

* For each and every of my Solaris patches, I'd have to investigate if
it works on 11.3 or needs adjustments and workarounds.

* I'd also need to regularly test the result to keep things working.

I honestly don't have the time or the energy to do this, nor the
hardware required for testing Besides, I have too much on my plate
already, and rather spend it on more beneficial work.

Above all, I always wonder why people insist on running ancient hardware
with an almost-unsupported OS, but require a bleeding edge version of
GCC. What's wrong with continuing to use GCC 13 (or even 14, although I haven't tested that on Solaris 11.3) instead?

Removing Solaris 11.3 support might make sense in the future when SPARC support in Illumos has matured enough that people can switch over their machines.

As has been noted, SPARC is on its way out for Illumos.

Rainer

-- ----------------------------------------------------------------------------- Rainer Orth, Center for Biotechnology, Bielefeld University

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Fri, 2024-05-10 at 12:14 +0200, Richard Biener wrote:

Because I wasn't subscribed to gcc-patches and I'm also only subscribed now without receiving messages due to the large message volume on this list.

https://gcc.gnu.org/gcc-13/changes.html

The problem with announcements on developer mailing lists is usually that they
usually don't reach any users. I was made aware of this change only when I checked about the recent changes to GCC Git.

Where do you expect such announcement then?

That's a difficult question, to be honest. From a user perspective, it's hard to
track these upstream announcements. I would argue that most users don't follow all such changes in the upstream projects they are using and, to be honest, I really
wouldn't have expected that Solaris 11.3 would be considered obsolete.

If it had been for Solaris 7, 8 or 9, I would totally understand. But even Solaris 10
is something that Oracle still supports [1].

Adrian

[1] https://blogs.oracle.com/support/post/extended-support-for-oracle-solaris-10-operating-system

--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer
`. `' Physicist
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi John,

On Fri, 2024-05-10 at 12:14 +0200, Richard Biener wrote:

Because I wasn't subscribed to gcc-patches and I'm also only subscribed now
without receiving messages due to the large message volume on this list. >>

https://gcc.gnu.org/gcc-13/changes.html

The problem with announcements on developer mailing lists is usually
that they
usually don't reach any users. I was made aware of this change only when I >> > checked about the recent changes to GCC Git.

Where do you expect such announcement then?

That's a difficult question, to be honest. From a user perspective, it's hard to
track these upstream announcements. I would argue that most users don't follow
all such changes in the upstream projects they are using and, to be honest,
I really
wouldn't have expected that Solaris 11.3 would be considered obsolete.

users don't have to follow such announcements at all unless they build
GCC from source. And those that do better should. As I mentioned, if
they try, they'll get the info from configure anyway.

If it had been for Solaris 7, 8 or 9, I would totally understand. But even Solaris 10
is something that Oracle still supports [1].

Right, and you pay an arm and a leg for that. Yet, Solaris 10 support
was removed from GCC 10 back in 2019 without much fuss.

Rainer

-- ----------------------------------------------------------------------------- Rainer Orth, Center for Biotechnology, Bielefeld University

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi Adrian,

While Oracle does no longer provide feature updates to Solaris 11.3, there >> > is still LTSS security support so that users still receive security updates
so that their systems are continued to be protected against vulnerabilities.

The Solaris 11.3 ESUs (Extended Support Updates) are available at a
premium only, and just contain the bare minimum of security updates,
often 6 to 9 month in between.

That's not an argument for throwing away hardware that still works perfectly fine and that still has some users.

who talks about throwing away the hardware: just keep it running with
the latest available OS and GCC and be done with it. You still refused
to explain what's the *need* for GCC > 14 on Solaris 11.3. Nice to have
maybe, but crucial?

I think Solaris 11.3 support should be kept since the resulting code removal
is not that large that it would justify dropping support for such a large >> > userbase.

Do you have any indication on the size of the userbase? I seriously
doubt it's large beyond some hobbyists that keep the old hardware
running.

I don't have the exact numbers, no. But I know there are many users out there with pre-11.4 hardware that they still use. As you may know, there are no 11.4 SPARC desktop systems and most 11.4-capable hardware is usually very expensive.

Sure, and they are free to to so. I've been running Solaris/SPARC
desktops for a long time myself, like an Ultra 1 or a Blade 1500.
However, once OS support was gone, I stopped and moved on.

You also seem to forget that my GCC (and LLVM) Solaris support work is
purely voluntary, done in my spare time.

Not sure what makes you think so. I'm perfectly aware of the fact that lots of
people do this work in their spare time as this applies to me as well.

I'm not getting paid for my Debian work, my kernel maintenance and all the other
stuff that I'm doing either. That doesn't mean users are not allowed to ask me
questions or send me comments about my work.

Who talks about being allowed? But you spend your time on your
Debian/sparc64 work as you see fit, and so do I. There's only so much
time in a day, and sometime it's nice to have a private life, too.
There's simply no time left with me for keeping S11.3 support going, so
it stops. It's that simple.

Keeping Solaris 11.3 support working would be much more than restoring
the removal patch:

* For each and every of my Solaris patches, I'd have to investigate if
it works on 11.3 or needs adjustments and workarounds.

* I'd also need to regularly test the result to keep things working.

I honestly don't have the time or the energy to do this, nor the
hardware required for testing Besides, I have too much on my plate
already, and rather spend it on more beneficial work.

Does Solaris support in GCC really change that often that the necessary tests cannot be run by volunteers? I'd be happy to test changes for Solaris 11.3 which can be installed inside an LDOM.

Hardware is just one concern here, but the limiting factor is my time.
No argument from you can increase that.

Above all, I always wonder why people insist on running ancient hardware
with an almost-unsupported OS, but require a bleeding edge version of
GCC. What's wrong with continuing to use GCC 13 (or even 14, although I
haven't tested that on Solaris 11.3) instead?

You could also ask why people use operating systems other than Linux and architectures other than x86_64. I don't think you will get a satisfactory answer to that question.

The question is who is doing the work. As long as there's someone who
does, fine, if not then support for those OSes dies (I've had that for
both IRIX and Tru64 Unix in the past, btw.).

Removing Solaris 11.3 support might make sense in the future when SPARC
support in Illumos has matured enough that people can switch over their
machines.

As has been noted, SPARC is on its way out for Illumos.

Which makes my point to keep Solaris 11.3 support even more valid.

You can keep wishing, but I won't be doing the work. It's that simple.

Rainer

-- ----------------------------------------------------------------------------- Rainer Orth, Center for Biotechnology, Bielefeld University

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Greetings, Rainer.

On 2024-05-10 07:44, Rainer Orth wrote (in part):

Besides, if John had ever tried to build either GCC 13 or 14 on Solaris
11.3, gcc/configure would have told him about the obsoletion in no
uncertain terms.

No, the option --enable-obsolete has allowed me to build on my T2000
running Solaris 11.3 until recently. (I just built GCC 14.1.0 on said machine.)

S.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Stuff Received <stuff@riddermarkfarm.ca> writes:

On 2024-05-10 07:44, Rainer Orth wrote (in part):

Besides, if John had ever tried to build either GCC 13 or 14 on Solaris
11.3, gcc/configure would have told him about the obsoletion in no
uncertain terms.

No, the option --enable-obsolete has allowed me to build on my T2000
running Solaris 11.3 until recently. (I just built GCC 14.1.0 on said machine.)

of course, but with default options you get a message indicating the obsoletion:

echo "*** Configuration ${target}${target_min} is obsolete." >&2
echo "*** Specify --enable-obsolete to build it anyway." >&2
echo "*** Support will be REMOVED in the next major release of GCC," >&2
echo "*** unless a maintainer comes forward." >&2

So nobody can say they didn't know about the obsoletion.

Rainer

-- ----------------------------------------------------------------------------- Rainer Orth, Center for Biotechnology, Bielefeld University

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Friday 2024-05-10 15:59, Rainer Orth wrote:

Stuff Received <stuff@riddermarkfarm.ca> writes:

On 2024-05-10 07:44, Rainer Orth wrote (in part):

Besides, if John had ever tried to build either GCC 13 or 14 on Solaris
11.3, gcc/configure would have told him about the obsoletion in no
uncertain terms.

No, the option --enable-obsolete has allowed me to build on my T2000
running Solaris 11.3 until recently. (I just built GCC 14.1.0 on said
machine.)

of course, but with default options you get a message indicating the >obsoletion:

echo "*** Configuration ${target}${target_min} is obsolete." >&2
echo "*** Specify --enable-obsolete to build it anyway." >&2
echo "*** Support will be REMOVED in the next major release of GCC," >&2
echo "*** unless a maintainer comes forward." >&2

So nobody can say they didn't know about the obsoletion.

I can. With my distro hat on, I can tell you that, when rpmbuild/dpkg-build* completes with exit status 0, there certainly is much reduced incentive to go looking at the build log.

This is also why people have started cutting dependency autodetection
from configure scripts (and the like) and making packagers having to use explicit --disable-thisandthat flags since about the start of the decade.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi Jan,

On Friday 2024-05-10 15:59, Rainer Orth wrote:

Stuff Received <stuff@riddermarkfarm.ca> writes:

On 2024-05-10 07:44, Rainer Orth wrote (in part):

Besides, if John had ever tried to build either GCC 13 or 14 on Solaris >>>> 11.3, gcc/configure would have told him about the obsoletion in no
uncertain terms.

No, the option --enable-obsolete has allowed me to build on my T2000
running Solaris 11.3 until recently. (I just built GCC 14.1.0 on said
machine.)

of course, but with default options you get a message indicating the >>obsoletion:

echo "*** Configuration ${target}${target_min} is obsolete." >&2
echo "*** Specify --enable-obsolete to build it anyway." >&2
echo "*** Support will be REMOVED in the next major release of GCC," >&2
echo "*** unless a maintainer comes forward." >&2

So nobody can say they didn't know about the obsoletion.

I can. With my distro hat on, I can tell you that, when rpmbuild/dpkg-build* completes with exit status 0, there certainly is much reduced incentive to go looking at the build log.

but without --enable-obsolete (which is off by default), the build will
abort with exit 1 on an obsolete target. This simply cannot be
overlooked.

Rainer

-- ----------------------------------------------------------------------------- Rainer Orth, Center for Biotechnology, Bielefeld University

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi Peter,

On Fri, 2024-05-10 at 12:07 +0100, Peter Tribble wrote:

Tribblix is built from the last commit that worked (November 2021), with any relevant changes
since cherry-picked on top. So in terms of timeline Tribblix is contemporary with 11.4, with
hardware support matching the original Solaris 11 release.

Thanks, good to know! And thanks a lot for your efforts!

But we're well into the second decade since the fork, so there's enough divergence that illumos
and Solaris are really different, even if some of what you see looks very similar.

(And illumos on SPARC uses gcc4 to build the kernel [!], although applications on Tribblix use gcc7.
Given the target market, having the latest and greatest toolchains isn't the highest priority.)

Well, at some point you will run into code that won't build with that old toolchain anymore.

Adrian

--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer
`. `' Physicist
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)