Discussion:
VAX Ultrix? (Re: GDB dropping support for mips-irix and alpha-tru64)
(too old to reply)
Pedro Alves
2014-10-11 14:36:29 UTC
Permalink
Raw Message
On 09/11/2014 07:52 PM, Joel Brobecker wrote:
> Dear all,
>
> The GDB project is planning on removing support for the following
> platforms:
...
> - Alpha Tru64 (alpha*-*-osf*)

Going over the supported hosts in configure.host, I noticed we still
"support" VAX Ultrix / 4.2BSD:

vax-*-bsd*) gdb_host=vax ;;
vax-*-ultrix*) gdb_host=vax ;;

Does it make sense to keep support for old Ultrix given we're
dropping OSF/1 / Tru64?

Below's the table I was building, listing the full set of
supported hosts, according to configure.host, mapping OS to triplet.

| 4.2BSD | vax-*-bsd* |
| AIX | powerpc-*-aix* |
| AIX | powerpc64-*-aix* |
| AIX | rs6000-*-* |
| Cygwin | i[34567]86-*-cygwin* |
| Cygwin | x86_64-*-cygwin* |
| DJGPP/go32 | i[34567]86-*-go32* |
| DJGPP/go32 | i[34567]86-*-msdosdjgpp* |
| Darwin / Mac OS X | *-*-darwin* |
| FreeBSD | alpha*-*-freebsd* |
| FreeBSD | i[34567]86-*-freebsd* |
| FreeBSD | powerpc*-*-freebsd* |
| FreeBSD | sparc64-*-freebsd* |
| FreeBSD | x86_64-*-freebsd* |
| GNU/Hurd | i[34567]86-*-gnu* |
| GNU/Linux | aarch64*-*-linux* |
| GNU/Linux | alpha*-*-linux* |
| GNU/Linux | arm*-*-linux* |
| GNU/Linux | hppa*-*-linux* |
| GNU/Linux | i[34567]86-*-linux* |
| GNU/Linux | ia64-*-linux* |
| GNU/Linux | m32r*-*-linux* |
| GNU/Linux | m68*-*-linux* |
| GNU/Linux | mips*-*-linux* |
| GNU/Linux | powerpc*-*-linux* |
| GNU/Linux | powerpc64*-*-linux* |
| GNU/Linux | s390*-*-linux* |
| GNU/Linux | sparc-*-linux* |
| GNU/Linux | sparc64-*-linux* |
| GNU/Linux | tilegx-*-linux* |
| GNU/Linux | x86_64-*-linux* |
| GNU/Linux | xtensa*-*-linux* |
| GNU/kFreeBSD | alpha*-*-kfreebsd*-gnu |
| GNU/kFreeBSD | i[34567]86-*-kfreebsd*-gnu |
| GNU/kFreeBSD | sparc64-*-kfreebsd*-gnu |
| GNU/kFreeBSD | x86_64-*-kfreebsd*-gnu |
| GNU/kNetBSD | alpha*-*-knetbsd*-gnu |
| GNU/kNetBSD | arm*-*-knetbsd*-gnu |
| GNU/kNetBSD | i[34567]86-*-knetbsd*-gnu |
| GNU/kNetBSD | m68*-*-knetbsd*-gnu |
| GNU/kNetBSD | mips*-*-knetbsd*-gnu |
| GNU/kNetBSD | powerpc-*-knetbsd*-gnu |
| GNU/kNetBSD | sh*-*-knetbsd*-gnu |
| GNU/kNetBSD | sparc-*-knetbsd*-gnu |
| GNU/kNetBSD | sparc64-*-knetbsd*-gnu |
| GNU/kNetBSD | vax-*-knetbsd*-gnu |
| GNU/kNetBSD | x86_64-*-knetbsd*-gnu |
| HP-UX | hppa*-*-hpux* |
| HP-UX | ia64-*-hpux* |
| MinGW / MinGW-w64 | i[34567]86-*-mingw32* |
| MinGW / MinGW-w64 | x86_64-*-mingw* |
| NetBSD | alpha*-*-netbsd* |
| NetBSD | arm*-*-netbsdelf* |
| NetBSD | hppa*-*-netbsd* |
| NetBSD | i[34567]86-*-netbsdelf* |
| NetBSD | m68*-*-netbsdelf* |
| NetBSD | mips*-*-netbsd* |
| NetBSD | powerpc-*-netbsd* |
| NetBSD | sh*-*-netbsdelf* |
| NetBSD | sparc-*-netbsdelf* |
| NetBSD | sparc64-*-netbsd* |
| NetBSD | vax-*-netbsdelf* |
| NetBSD | x86_64-*-netbsd* |
| OpenBSD | alpha*-*-openbsd* |
| OpenBSD | arm*-*-openbsd* |
| OpenBSD | hppa*-*-openbsd* |
| OpenBSD | i[34567]86-*-openbsd* |
| OpenBSD | m68*-*-openbsd* |
| OpenBSD | m88*-*-openbsd* |
| OpenBSD | mips64*-*-openbsd* |
| OpenBSD | powerpc-*-openbsd* |
| OpenBSD | sh*-*-openbsd* |
| OpenBSD | sparc-*-openbsd* |
| OpenBSD | sparc64-*-openbsd* |
| OpenBSD | vax-*-openbsd* |
| OpenBSD | x86_64-*-openbsd* |
| QNX/NTO | i[3456]86-*-nto* |
| Solaris | i[34567]86-*-solaris* |
| Solaris | i[34567]86-*-solaris2.1[0-9]* |
| Solaris | sparc-*-solaris2* |
| Solaris | sparc64-*-solaris2* |
| Solaris | sparcv9-*-solaris2* |
| Solaris | x86_64-*-solaris2.1[0-9]* |
| Ultrix | vax-*-ultrix* |

Thanks,
Pedro Alves
Joel Brobecker
2014-10-13 13:38:09 UTC
Permalink
Raw Message
> Going over the supported hosts in configure.host, I noticed we still
> "support" VAX Ultrix / 4.2BSD:
>
> vax-*-bsd*) gdb_host=vax ;;
> vax-*-ultrix*) gdb_host=vax ;;
>
> Does it make sense to keep support for old Ultrix given we're
> dropping OSF/1 / Tru64?

Wikipedia says that VAX production ceased in 2005. The last VAX-specific
patch I can see being submitted to gdb-patches is us mentioning support
for VAX floats in the Ada mode (which we removed in 2010). VAX VMS was
removed in 2010 from BFD. There seems to be some regular activity around
VAX on the GCC side, though, but Ultrix itself seems to be no longer be
supported.

So this seems to indicate that we will be able to remove support
for Ultrix.

> Below's the table I was building, listing the full set of
> supported hosts, according to configure.host, mapping OS to triplet.

> | HP-UX | hppa*-*-hpux* |
> | HP-UX | ia64-*-hpux* |

I suspect that HP-UX will no longer find any takers. AdaCore had to
step-up many many years go to keep those alive. The situation is
different today, and we are stepping down.

--
Joel
Maciej W. Rozycki
2014-10-13 13:57:46 UTC
Permalink
Raw Message
On Mon, 13 Oct 2014, Joel Brobecker wrote:

> Wikipedia says that VAX production ceased in 2005. The last VAX-specific
> patch I can see being submitted to gdb-patches is us mentioning support
> for VAX floats in the Ada mode (which we removed in 2010). VAX VMS was
> removed in 2010 from BFD. There seems to be some regular activity around
> VAX on the GCC side, though, but Ultrix itself seems to be no longer be
> supported.

Oh, I have some VAX/Linux stuff waiting too; although whether it makes
sense to submit or not will depend on what happens to support for that
platform in the Linux kernel and glibc. I've got a port mostly working,
but both the kernel and glibc are rather ancient and glibc can't proceed
without defining a TLS ABI. I have no interest in Ultrix itself though.

Maciej
Pedro Alves
2014-10-13 15:02:04 UTC
Permalink
Raw Message
On 10/13/2014 02:57 PM, Maciej W. Rozycki wrote:
> On Mon, 13 Oct 2014, Joel Brobecker wrote:
>
>> Wikipedia says that VAX production ceased in 2005. The last VAX-specific
>> patch I can see being submitted to gdb-patches is us mentioning support
>> for VAX floats in the Ada mode (which we removed in 2010). VAX VMS was
>> removed in 2010 from BFD. There seems to be some regular activity around
>> VAX on the GCC side, though, but Ultrix itself seems to be no longer be
>> supported.
>
> Oh, I have some VAX/Linux stuff waiting too; although whether it makes
> sense to submit or not will depend on what happens to support for that
> platform in the Linux kernel and glibc. I've got a port mostly working,
> but both the kernel and glibc are rather ancient and glibc can't proceed
> without defining a TLS ABI. I have no interest in Ultrix itself though.

OK, thanks.

TBC, I'm looking at this from the "DEC replaced Ultrix with OSF/1"
angle, not really paying attention to VAX vs Alpha.

I'm more interested in getting rid of old operating systems
than old architectures, in the sense that those could potentially
hold back newer tool/compiler requirements. Modern tools running on
VAX is fine with me.

Thanks,
Pedro Alves
Mark Kettenis
2014-10-13 16:03:54 UTC
Permalink
Raw Message
> Date: Mon, 13 Oct 2014 06:38:09 -0700
> From: Joel Brobecker <***@adacore.com>
>
> > Going over the supported hosts in configure.host, I noticed we still
> > "support" VAX Ultrix / 4.2BSD:
> >
> > vax-*-bsd*) gdb_host=vax ;;
> > vax-*-ultrix*) gdb_host=vax ;;
> >
> > Does it make sense to keep support for old Ultrix given we're
> > dropping OSF/1 / Tru64?
>
> Wikipedia says that VAX production ceased in 2005. The last VAX-specific
> patch I can see being submitted to gdb-patches is us mentioning support
> for VAX floats in the Ada mode (which we removed in 2010). VAX VMS was
> removed in 2010 from BFD. There seems to be some regular activity around
> VAX on the GCC side, though, but Ultrix itself seems to be no longer be
> supported.

There are probably quite a few VAXen still running. Just learned last
week there are still some radiotelescopes around running a VAX to
control the telescope. But they're probably running VMS on those
though. If you want, there's the SIMH simulator.

> So this seems to indicate that we will be able to remove support
> for Ultrix.

FWIW, I kept the VAX Ultrix and VAX 4.2BSD code around as an example
of how a classic ptrace(2) implementation works. Helped me a great
bit when refactoring inf-ptrace.c back in the days. Linux really
turned ptrace(2) into a mess...

Other than the educational value, there is no point in keeping Ultrix
and BSD4.2 support alive. I'm pretty sure that GDB has become too
bloated to compile and/or run on these systems.

> > Below's the table I was building, listing the full set of
> > supported hosts, according to configure.host, mapping OS to triplet.
>
> > | HP-UX | hppa*-*-hpux* |
> > | HP-UX | ia64-*-hpux* |
>
> I suspect that HP-UX will no longer find any takers. AdaCore had to
> step-up many many years go to keep those alive. The situation is
> different today, and we are stepping down.

It'd be somewhat sad to see inf-ttrace.c go. IMHO ttrace(2) is by far
the best example of how to do a proper threads-aware debugger interface.
Pedro Alves
2014-10-15 13:22:40 UTC
Permalink
Raw Message
On 10/13/2014 05:03 PM, Mark Kettenis wrote:
>> Date: Mon, 13 Oct 2014 06:38:09 -0700
>> From: Joel Brobecker <***@adacore.com>
>>
>>> Going over the supported hosts in configure.host, I noticed we still
>>> "support" VAX Ultrix / 4.2BSD:
>>>
>>> vax-*-bsd*) gdb_host=vax ;;
>>> vax-*-ultrix*) gdb_host=vax ;;
>>>
>>> Does it make sense to keep support for old Ultrix given we're
>>> dropping OSF/1 / Tru64?
>>
>> Wikipedia says that VAX production ceased in 2005. The last VAX-specific
>> patch I can see being submitted to gdb-patches is us mentioning support
>> for VAX floats in the Ada mode (which we removed in 2010). VAX VMS was
>> removed in 2010 from BFD. There seems to be some regular activity around
>> VAX on the GCC side, though, but Ultrix itself seems to be no longer be
>> supported.
>
> There are probably quite a few VAXen still running. Just learned last
> week there are still some radiotelescopes around running a VAX to
> control the telescope. But they're probably running VMS on those
> though. If you want, there's the SIMH simulator.
>
>> So this seems to indicate that we will be able to remove support
>> for Ultrix.
>
> FWIW, I kept the VAX Ultrix and VAX 4.2BSD code around as an example
> of how a classic ptrace(2) implementation works. Helped me a great
> bit when refactoring inf-ptrace.c back in the days. Linux really
> turned ptrace(2) into a mess...

For my own education, which code are you referring to though?
The PT_READ_U / PT_WRITE_U bits in inf-ptrace.c?

>
> Other than the educational value, there is no point in keeping Ultrix
> and BSD4.2 support alive. I'm pretty sure that GDB has become too
> bloated to compile and/or run on these systems.
>
>>> Below's the table I was building, listing the full set of
>>> supported hosts, according to configure.host, mapping OS to triplet.
>>
>>> | HP-UX | hppa*-*-hpux* |
>>> | HP-UX | ia64-*-hpux* |
>>
>> I suspect that HP-UX will no longer find any takers. AdaCore had to
>> step-up many many years go to keep those alive. The situation is
>> different today, and we are stepping down.
>
> It'd be somewhat sad to see inf-ttrace.c go. IMHO ttrace(2) is by far
> the best example of how to do a proper threads-aware debugger interface.

Yeah, I agree here.

Thanks,
Pedro Alves
Pedro Alves
2014-10-17 13:47:26 UTC
Permalink
Raw Message
On 10/13/2014 05:03 PM, Mark Kettenis wrote:
> Other than the educational value, there is no point in keeping Ultrix
> and BSD4.2 support alive. I'm pretty sure that GDB has become too
> bloated to compile and/or run on these systems.

OK, I'll send a patch to remove support for these soon.

Thanks,
Pedro Alves
Loading...