Re: porterboxes vs. qemu
Please excuse if my tone came across incorrectly. Far from intending
to be antagonistic, I'm expending considerable attention on how to
preserve gcl, maxima, acl2, and axiom on mips. In particular, I've
written native mips object relocation code for this purpose.
These packages are large, and inevitably experience bit rot from time
to time needing per platform adjustment. I've been trying to test in
advance to the extent possible on porterboxes, but on mips this is
impossible. Even with this effort, its taken me 6 tries to get an
axiom that will likely pass on all arches after a gap of 4 years.
Alas, these are too many attempts for our current buildd
infrastructure, and the admins have decided to quasi-permanently
exclude axiom from the buildd queue, at least on slower machines, with
some justification to preserve the port as a whole.
However, when the turn around time for an auto build is two weeks to a
month, and when by hand builds are not possible without locally
installed emulation, this makes the management of the package very
difficult. All the other more popular platforms are blocked pending
some possibly trivial error that might arise a month hence, mandating
a restart from the beginning.
I do believe that our most precious resource is human and not machine
time. If there is interest in keeping this port alive, which I
certainly share, we need a hefty multiple of available hardware in
comparison to the other targets. The m68k people, before their final
phase out, provided a mighty example in this regard.
But of course we do need to ask -- are there people who really want
these programs on mips?
Tom McAvaney <email@example.com> writes:
>> Greetings! Are there any non-official porterboxes available for mips
>> only debugging? Or is the policy henceforward that packagers need to
>> setup qemu or exclude mips?
> I am a mere user and no authority at all, but my understanding is that
> now mahler is down the MIPS port is dependent on individual
> developers/packagers having access to/maintaining their own machines.
> It would be a shame if packagers abandoned MIPS for lack of hardware
> for debugging ... Does anyone know if replacement/repair of mahler is
> on anyones to do list?
> In the meantime, depending on how much hand-holding you're willing to
> do, I (or others on the list) might be able to help with testing on
> one of my SGI Indys.
Camm Maguire firstname.lastname@example.org
"The earth is but one country, and mankind its citizens." -- Baha'u'llah