Sparc64 and FreeBSD 13.0

classic Classic list List threaded Threaded
11 messages Options
Reply | Threaded
Open this post in threaded view
|

Sparc64 and FreeBSD 13.0

Ed Maste-2
Work is in progress to remove the obsolete GCC 4.2.1 from the FreeBSD
tree before FreeBSD 13.0[1], and a tracking PR[2] is open with a list
of dependencies[3]. In the absence of any action to address these
(known, and perhaps additional unknown) dependencies this means
Sparc64 will fail to build as of 2020-03-31.

[1] https://lists.freebsd.org/pipermail/freebsd-arch/2019-August/019674.html
[2] https://bugs.freebsd.org/228919
[3] https://bugs.freebsd.org/bugzilla/showdependencytree.cgi?id=228919&hide_resolved=1
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Ed Maste-2
On Tue, 13 Aug 2019 at 19:57, Ed Maste <[hidden email]> wrote:
>
> Work is in progress to remove the obsolete GCC 4.2.1 from the FreeBSD
> tree before FreeBSD 13.0[1], and a tracking PR[2] is open with a list
> of dependencies[3]. In the absence of any action to address these
> (known, and perhaps additional unknown) dependencies this means
> Sparc64 will fail to build as of 2020-03-31.

Note that the next step in the GCC 4.2.1 deprecation process is four weeks away:
| 2019-12-31: Turn off gcc 4.2.1 build by default (can be turned on)

I have received some private mail about addressing the compiler
situation on sparc64, but not the other issues tracked as dependencies
in PR228919.
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Mark Linimon-2
On Tue, Dec 03, 2019 at 05:53:28AM -0500, Ed Maste wrote:
> I have received some private mail about addressing the compiler
> situation on sparc64, but not the other issues tracked as dependencies
> in PR228919.

I personally don't intend to spend cycles working on sparc64 in the
future.

As one of the last people who was supporting it, I think it's fair to
say that sparc64 has dropped below critical mass.

mcl
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Chris Ross


> On Dec 3, 2019, at 13:11, Mark Linimon <[hidden email]> wrote:
>
> On Tue, Dec 03, 2019 at 05:53:28AM -0500, Ed Maste wrote:
>> I have received some private mail about addressing the compiler
>> situation on sparc64, but not the other issues tracked as dependencies
>> in PR228919.
>
> I personally don't intend to spend cycles working on sparc64 in the
> future.
>
> As one of the last people who was supporting it, I think it's fair to
> say that sparc64 has dropped below critical mass.

I was an active user, but lack of support for more modern spare processors and the lack of active use have caused me to step away too.  If I had the time to look into porting T1/T2/etc support, and try to get the compiler working, I’d love to have the results.  But I do not.

        - Chris

_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Michelle Sullivan
Chris Ross wrote:

>
>> On Dec 3, 2019, at 13:11, Mark Linimon <[hidden email]> wrote:
>>
>> On Tue, Dec 03, 2019 at 05:53:28AM -0500, Ed Maste wrote:
>>> I have received some private mail about addressing the compiler
>>> situation on sparc64, but not the other issues tracked as dependencies
>>> in PR228919.
>> I personally don't intend to spend cycles working on sparc64 in the
>> future.
>>
>> As one of the last people who was supporting it, I think it's fair to
>> say that sparc64 has dropped below critical mass.
> I was an active user, but lack of support for more modern spare processors and the lack of active use have caused me to step away too.  If I had the time to look into porting T1/T2/etc support, and try to get the compiler working, I’d love to have the results.  But I do not.
>

I don't have the skills or I'd do it.

--
Michelle Sullivan
http://www.mhix.org/

_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Gordon Zaft
I am still interested in using this platform but doubt I have the skills
and time to deal with something like a compiler upgrade.

G

On Wed, Dec 4, 2019 at 5:08 AM Michelle Sullivan <[hidden email]> wrote:

> Chris Ross wrote:
> >
> >> On Dec 3, 2019, at 13:11, Mark Linimon <[hidden email]> wrote:
> >>
> >> On Tue, Dec 03, 2019 at 05:53:28AM -0500, Ed Maste wrote:
> >>> I have received some private mail about addressing the compiler
> >>> situation on sparc64, but not the other issues tracked as dependencies
> >>> in PR228919.
> >> I personally don't intend to spend cycles working on sparc64 in the
> >> future.
> >>
> >> As one of the last people who was supporting it, I think it's fair to
> >> say that sparc64 has dropped below critical mass.
> > I was an active user, but lack of support for more modern spare
> processors and the lack of active use have caused me to step away too.  If
> I had the time to look into porting T1/T2/etc support, and try to get the
> compiler working, I’d love to have the results.  But I do not.
> >
>
> I don't have the skills or I'd do it.
>
> --
> Michelle Sullivan
> http://www.mhix.org/
>
> _______________________________________________
> [hidden email] mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
> To unsubscribe, send any mail to "[hidden email]"
>


--
Gordon Zaft
[hidden email]
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Joe Kozak
Ive got two t-3 and one t4 id love to do more with.   Linux support is
ancient, and id love to learn much more  about freebsd as i have very
little experience,  but alas i have to drywall and mud my house... so time
time time.   I put my vote of support in for someone doing some updating if
possible.

On Wed, Dec 4, 2019 at 5:00 PM Gordon Zaft <[hidden email]> wrote:

> I am still interested in using this platform but doubt I have the skills
> and time to deal with something like a compiler upgrade.
>
> G
>
> On Wed, Dec 4, 2019 at 5:08 AM Michelle Sullivan <[hidden email]>
> wrote:
>
> > Chris Ross wrote:
> > >
> > >> On Dec 3, 2019, at 13:11, Mark Linimon <[hidden email]> wrote:
> > >>
> > >> On Tue, Dec 03, 2019 at 05:53:28AM -0500, Ed Maste wrote:
> > >>> I have received some private mail about addressing the compiler
> > >>> situation on sparc64, but not the other issues tracked as
> dependencies
> > >>> in PR228919.
> > >> I personally don't intend to spend cycles working on sparc64 in the
> > >> future.
> > >>
> > >> As one of the last people who was supporting it, I think it's fair to
> > >> say that sparc64 has dropped below critical mass.
> > > I was an active user, but lack of support for more modern spare
> > processors and the lack of active use have caused me to step away too.
> If
> > I had the time to look into porting T1/T2/etc support, and try to get the
> > compiler working, I’d love to have the results.  But I do not.
> > >
> >
> > I don't have the skills or I'd do it.
> >
> > --
> > Michelle Sullivan
> > http://www.mhix.org/
> >
> > _______________________________________________
> > [hidden email] mailing list
> > https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
> > To unsubscribe, send any mail to "
> [hidden email]"
> >
>
>
> --
> Gordon Zaft
> [hidden email]
> _______________________________________________
> [hidden email] mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
> To unsubscribe, send any mail to "[hidden email]"
>
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Mark Linimon-2
On Wed, Dec 04, 2019 at 05:34:46PM -0600, Joe Kozak wrote:
> I put my vote of support in for someone doing some updating if
> possible.

Unfortunately it doesn't work like that in volunteer projects.

I estimate it would take 1-3 months to fix the toolchain -- which
would only save us the existing systems (manufactured up to 2005).
When I last looked at sun4v (the newer systems) I estimated 3-5
months.

That's full-time activity from someone who *used* to do systems
work, but has not for many years, and would have to come up to
speed on 18 zillion things.

mcl

(fwiw I am a software engineer, ergo, my estimates suck.)
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Joe Kozak
Is that for basic support in the compiler linker bits?

On Wed, Dec 4, 2019 at 9:58 PM Mark Linimon <[hidden email]> wrote:

> On Wed, Dec 04, 2019 at 05:34:46PM -0600, Joe Kozak wrote:
> > I put my vote of support in for someone doing some updating if
> > possible.
>
> Unfortunately it doesn't work like that in volunteer projects.
>
> I estimate it would take 1-3 months to fix the toolchain -- which
> would only save us the existing systems (manufactured up to 2005).
> When I last looked at sun4v (the newer systems) I estimated 3-5
> months.
>
> That's full-time activity from someone who *used* to do systems
> work, but has not for many years, and would have to come up to
> speed on 18 zillion things.
>
> mcl
>
> (fwiw I am a software engineer, ergo, my estimates suck.)
>
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

Mark Linimon-2
On Wed, Dec 04, 2019 at 10:00:35PM -0600, Joe Kozak wrote:
> Is that for basic support in the compiler linker bits?

The latter.  It hasn't bitrotted *that* much.

I *think*.

mcl
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: Sparc64 and FreeBSD 13.0

John Baldwin
In reply to this post by Mark Linimon-2
On 12/4/19 7:58 PM, Mark Linimon wrote:

> On Wed, Dec 04, 2019 at 05:34:46PM -0600, Joe Kozak wrote:
>> I put my vote of support in for someone doing some updating if
>> possible.
>
> Unfortunately it doesn't work like that in volunteer projects.
>
> I estimate it would take 1-3 months to fix the toolchain -- which
> would only save us the existing systems (manufactured up to 2005).
> When I last looked at sun4v (the newer systems) I estimated 3-5
> months.
>
> That's full-time activity from someone who *used* to do systems
> work, but has not for many years, and would have to come up to
> speed on 18 zillion things.
>
> mcl
>
> (fwiw I am a software engineer, ergo, my estimates suck.)

Fixing the toolchain is probably less than 1-3 months.

I have a freebsd-gcc9 port that I haven't posted for review yet, but
it hits the same snag (assertion failure in ld.bfd) that freebsd-gcc6
hits.  I think the assertion failure is wrong and needs to use >=
instead of >, but I just haven't sat down and looked at it in a
debugger to see the exact values at the time of the assertion failure.

However, I think even once you get it building there might be a few
more bumps with getting things like PCPU_* working in the kernel
correctly.  I probably won't get to it anytime soon, and sparc64
really needs multiple active folks to keep it viable.

--
John Baldwin
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-sparc64
To unsubscribe, send any mail to "[hidden email]"