Updating Primary and Secondary platform list for gcc-4.5 ???

[Kaveh R. Ghazi]

I was looking through the gcc-4.5 primary and secondary platform list
to ensure we have coverage for MPC testing.  It occurs to me that some
of the OS versions are outdated.

I've included the list from the page
http://gcc.gnu.org/gcc-4.5/criteria.html

Should we update:

1.  solaris2.10 -> 2.11
2.  hpux11.11 -> ???
3.  aix5.2 -> 5.3

For the purposes of MPC testing I also want to know:

1.  Is mipsisa64-elf cross-config only?  I.e. MPC portability testing
is only relevant to the host platform where GCC is run.  So we don't
have to worry about this one?

2.  i386-unknown-freebsd and i686-apple-darwin are generic, but
config.guess will supply specific version numbers.  What version
should MPC be shown to work on?  Any one of them would do?

3.  For freebsd and darwin, do we want to include specific version
numbers for our platform list, or leave them generic?

                Thanks,
                --Kaveh


Primary Platform List

arm-eabi
i386-unknown-freebsd
i686-pc-linux-gnu
mipsisa64-elf
powerpc64-unknown-linux-gnu
sparc-sun-solaris2.10
x86_64-unknown-linux-gnu

Secondary Platform List

hppa2.0w-hp-hpux11.11
powerpc-ibm-aix5.2.0.0
i686-apple-darwin
i686-pc-cygwin
i686-mingw32
ia64-unknown-linux-gnu
s390-linux-gnu

[Eric Botcazou-3]

> I was looking through the gcc-4.5 primary and secondary platform list
> to ensure we have coverage for MPC testing.  It occurs to me that some
> of the OS versions are outdated.
>
> I've included the list from the page
> http://gcc.gnu.org/gcc-4.5/criteria.html
>
> Should we update:
>
> 1.  solaris2.10 -> 2.11

Why move to a not-yet-released version?

--
Eric Botcazou

On Sat, 7 Nov 2009, Kaveh R. GHAZI wrote:
> 2.  i386-unknown-freebsd and i686-apple-darwin are generic, but
> config.guess will supply specific version numbers.  What version
> should MPC be shown to work on?  Any one of them would do?

For FreeBSD, I'd specify versions 6.x and above.  Older versions should
work down to 3.x or 4.x, but it is not worth bothering unless someone is
into retro computing.  On the contrary, FreeBSD 6.x to some extent, and
now 7.x primarily and now also 8.x are tested and used regularily.

> 3.  For freebsd and darwin, do we want to include specific version
> numbers for our platform list, or leave them generic?

I think it's fine keeping things as are.  If you prefer, we can also
make it "whatever Gerald or Loren are currently testing" for FreeBSD,
of course. ;-)

Gerald

[Rainer Orth]

Eric Botcazou <ebotcazou@...> writes:

> > I was looking through the gcc-4.5 primary and secondary platform list
> > to ensure we have coverage for MPC testing.  It occurs to me that some
> > of the OS versions are outdated.
> >
> > I've included the list from the page
> > http://gcc.gnu.org/gcc-4.5/criteria.html
> >
> > Should we update:
> >
> > 1.  solaris2.10 -> 2.11
>
> Why move to a not-yet-released version?

Indeed: while I regularly test on Solaris 11/SPARC at the moment, it's
still so much of a moving target that this doesn't make any sense.

        Rainer

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

[Dennis Clarke-2]

> Eric Botcazou <ebotcazou@...> writes:
>
>> > I was looking through the gcc-4.5 primary and secondary platform list
>> > to ensure we have coverage for MPC testing.  It occurs to me that some
>> > of the OS versions are outdated.
>> >
>> > I've included the list from the page
>> > http://gcc.gnu.org/gcc-4.5/criteria.html
>> >
>> > Should we update:
>> >
>> > 1.  solaris2.10 -> 2.11
>>
>> Why move to a not-yet-released version?
>
> Indeed: while I regularly test on Solaris 11/SPARC at the moment, it's
> still so much of a moving target that this doesn't make any sense.

The issue may be one of "de facto" vs "defined as being" released.

There is no such thing as a released Solaris revision that responds to
uname with SunOS5.11 yet. When Sun/Oracle actually releases something AND
you can buy a support contract for it then you have a valid platform in
commercial use.

Having said that .. I see roughly 30% of all my traffic from SunOS5.11
users on either Solaris Nevada or OpenSolaris beta releases.

The question should be ... do we in the community end user world see
SunOS5.11 as being a de facto release? I would say yes.

Solaris 10 is the enterprise class commercial grade Solaris release and it
is staying put for a long long long time yet.

--
Dennis Clarke
dclarke@...  <- Email related to the open source Solaris
dclarke@...   <- Email related to open source for Solaris

[Rainer Orth]

Dennis Clarke writes:

> > Eric Botcazou <ebotcazou@...> writes:
> >
> >> > I was looking through the gcc-4.5 primary and secondary platform list
> >> > to ensure we have coverage for MPC testing.  It occurs to me that some
> >> > of the OS versions are outdated.
> >> >
> >> > I've included the list from the page
> >> > http://gcc.gnu.org/gcc-4.5/criteria.html
> >> >
> >> > Should we update:
> >> >
> >> > 1.  solaris2.10 -> 2.11
> >>
> >> Why move to a not-yet-released version?
> >
> > Indeed: while I regularly test on Solaris 11/SPARC at the moment, it's
> > still so much of a moving target that this doesn't make any sense.
>
> The issue may be one of "de facto" vs "defined as being" released.
>
> There is no such thing as a released Solaris revision that responds to
> uname with SunOS5.11 yet. When Sun/Oracle actually releases something AND
> you can buy a support contract for it then you have a valid platform in
> commercial use.

You can get support for the OpenSolaris distribution if you like, yet this
is still very much work in progress, not a stable platform we can rely on.

> Having said that .. I see roughly 30% of all my traffic from SunOS5.11
> users on either Solaris Nevada or OpenSolaris beta releases.
>
> The question should be ... do we in the community end user world see
> SunOS5.11 as being a de facto release? I would say yes.

Certainly not, even if it is widely used (primarily on laptops, I suppose).

> Solaris 10 is the enterprise class commercial grade Solaris release and it
> is staying put for a long long long time yet.

Indeed, and even if we chose sparc-sun-solaris2.10 as the primary platform
doesn't mean that *-*-solaris2.11 doesn't work, quite the contrary: I
regularly test both and try to keep them working.

        Rainer

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

[Dennis Clarke-2]

>> you can buy a support contract for it then you have a valid platform in
>> commercial use.
>
> You can get support for the OpenSolaris distribution if you like

I just went and looked ... you are correct, they have three levels in
fact. It looks like $1080 for premium, $720 is standard business hours
$324 is patches and updates with email tech support I think.

So that makes it a real commercial platform in my mind.

> is still very much work in progress, not a stable platform we can rely on.

However, Solaris 10 was also a moving platoform in its first few releases
but no one would debate it as a commercial grade release or not. I think
Opensolaris must be looked at as viable and commercial grade. I am not at
all biased in this regardless of the fact that I have been involved one
way or another in the OpenSolaris project since day one. I'm very much an
outside guy that just loves to experiment and perhaps even attempt to help
where I can.

>> Having said that .. I see roughly 30% of all my traffic from SunOS5.11
>> users on either Solaris Nevada or OpenSolaris beta releases.
>>
>> The question should be ... do we in the community end user world see
>> SunOS5.11 as being a de facto release? I would say yes.
>
> Certainly not, even if it is widely used (primarily on laptops, I
> suppose).

Well, would Fedora Core on PowerPC or Ubuntu or Debian ( any release ) be
considered a platform or does that just fall under a wide umbrella of
"Linux" ? Some of those are barely used at all anymore. Consider running
Linux on a DEC Alpha. Who does that anymore? Is this a popularity
measurement or is this based on something more tangible and quantitative
like "commercially supported"?

>> Solaris 10 is the enterprise class commercial grade Solaris release and
>> it is staying put for a long long long time yet.
>
> Indeed, and even if we chose sparc-sun-solaris2.10 as the primary platform
> doesn't mean that *-*-solaris2.11 doesn't work, quite the contrary: I
> regularly test both and try to keep them working.

I test everything on *-*-solaris2.8 which by way of the ABI golden rule
instantly qualifies as tested on anything up to SunOS2.10. It does not
imply SunOS2.11 however.

--
Dennis

[Richard Guenther-2]

On Mon, Nov 9, 2009 at 5:50 PM, Dennis Clarke <dclarke@...> wrote:

>
>>> you can buy a support contract for it then you have a valid platform in
>>> commercial use.
>>
>> You can get support for the OpenSolaris distribution if you like
>
> I just went and looked ... you are correct, they have three levels in
> fact. It looks like $1080 for premium, $720 is standard business hours
> $324 is patches and updates with email tech support I think.
>
> So that makes it a real commercial platform in my mind.
>
>> is still very much work in progress, not a stable platform we can rely on.
>
> However, Solaris 10 was also a moving platoform in its first few releases
> but no one would debate it as a commercial grade release or not. I think
> Opensolaris must be looked at as viable and commercial grade. I am not at
> all biased in this regardless of the fact that I have been involved one
> way or another in the OpenSolaris project since day one. I'm very much an
> outside guy that just loves to experiment and perhaps even attempt to help
> where I can.
>
>>> Having said that .. I see roughly 30% of all my traffic from SunOS5.11
>>> users on either Solaris Nevada or OpenSolaris beta releases.
>>>
>>> The question should be ... do we in the community end user world see
>>> SunOS5.11 as being a de facto release? I would say yes.
>>
>> Certainly not, even if it is widely used (primarily on laptops, I
>> suppose).
>
> Well, would Fedora Core on PowerPC or Ubuntu or Debian ( any release ) be
> considered a platform or does that just fall under a wide umbrella of
> "Linux" ? Some of those are barely used at all anymore. Consider running
> Linux on a DEC Alpha. Who does that anymore? Is this a popularity
> measurement or is this based on something more tangible and quantitative
> like "commercially supported"?
>
>>> Solaris 10 is the enterprise class commercial grade Solaris release and
>>> it is staying put for a long long long time yet.
>>
>> Indeed, and even if we chose sparc-sun-solaris2.10 as the primary platform
>> doesn't mean that *-*-solaris2.11 doesn't work, quite the contrary: I
>> regularly test both and try to keep them working.
>
> I test everything on *-*-solaris2.8 which by way of the ABI golden rule
> instantly qualifies as tested on anything up to SunOS2.10. It does not
> imply SunOS2.11 however.

If config.gcc handles both triples the same (*-*-solaris2.10 and
*-*-solaris2.11) then we can consider both at the same level.
Just as we don't make a distinction for glibc or kernel releases
for the *-*-linux targets.  That some targets specify certain versions
is because in the past they handled each OS version slightly
different during GCC build and some even in use (for example
fixincludes have to be adjusted, etc.).

Richard.

[Mark Mitchell]

Richard Guenther wrote:

> If config.gcc handles both triples the same (*-*-solaris2.10 and
> *-*-solaris2.11) then we can consider both at the same level.

Indeed.  Furthermore, we certainly wouldn't want to break support for
Solaris 2.10 at this point, so having 2.10 listed seems to make sense to
me.

--
Mark Mitchell
CodeSourcery
mark@...
[phone] x713

[Kaveh R. Ghazi]

From: "Mark Mitchell" <mark@...>

> Richard Guenther wrote:
>
>> If config.gcc handles both triples the same (*-*-solaris2.10 and
>> *-*-solaris2.11) then we can consider both at the same level.
>
> Indeed.  Furthermore, we certainly wouldn't want to break support for
> Solaris 2.10 at this point, so having 2.10 listed seems to make sense to
> me.

Agreed.  I guess my remaining questions are for AIX and mipsisa64-elf.

Can someone please confirm that mipsisa64-elf is a cross-compile-only target
and therefore not relevant for host-based MPC portability testing?

And do we want to update aix5.2 to aix5.3 in our platforms list?

        Thanks,
        --Kaveh

[David Edelsohn-3]

On Thu, Nov 12, 2009 at 10:00 AM, Kaveh R. Ghazi <ghazi@...> wrote:

> And do we want to update aix5.2 to aix5.3 in our platforms list?

AIX should be updated to 5.3 or 6.1.

David

[Kaveh R. Ghazi]

From: "David Edelsohn" <dje.gcc@...>

> On Thu, Nov 12, 2009 at 10:00 AM, Kaveh R. Ghazi <ghazi@...>
> wrote:
>
>> And do we want to update aix5.2 to aix5.3 in our platforms list?
>
> AIX should be updated to 5.3 or 6.1.
> David

For the last two months or so, the AIX reports I see are mostly (all?) for
5.3, so I suggest we use that version.

        --Kaveh

[Richard Sandiford-5]

"Kaveh R. Ghazi" <ghazi@...> writes:
> Agreed.  I guess my remaining questions are for AIX and mipsisa64-elf.
>
> Can someone please confirm that mipsisa64-elf is a cross-compile-only target
> and therefore not relevant for host-based MPC portability testing?

Yes, that's right.

I'm sure it's technically possible to compile GCC with a mipsisa64-elf
compiler, but it would be hard, and it isn't a well-defined operation.
I don't think you could do it with the usual libgloss runtime environment;
you'd need to use a "real" OS.

Richard