status

[Neal Norwitz]

The snake farm is slowly getting into better shape.
There are still several problems, though.

1) The latest results page doesn't always show the most recent
build for each architecture. For example, right now the most
recent build for AIX is on 4/4, then most recent builds for
other boxes are the 10th. Could we change the script which
produces the page to be more current? The oldest build goes
back to Jan 20th. That doesn't seem too useful.

2) Solaris 8 test_time fails. This is a problem I've been working
on which deals with tzset. <http://python.org/sf/706707>

3) AIX fails to link. This works for me when built manually.
So there must be some difference in the environment. The
problem is that -lpthread is not present when linking.
I'll try to look into this. If I have write privileges
to the build area, I should be able to fix the problem.

4) On one of the alpha's (aristocat), all the DB tests fail:
architecture lacks fast mutexes:
applications cannot be threaded
I'm guessing this is a kernel/glibc/configure issue.
I haven't looked into this, but will eventually get around
to it. The other alpha passes all the tests I believe.

5) Many roxen.com builds failing. These all seem to be due
to compiler issues.

6) HP-UX--unknown, there isn't a current entry on the page.

I'll keep working on 2, 3, and 4.

Neal

[Neal Norwitz]

Post by Neal Norwitz
3) AIX fails to link. This works for me when built manually.
So there must be some difference in the environment. The
problem is that -lpthread is not present when linking.
I'll try to look into this. If I have write privileges
to the build area, I should be able to fix the problem.

This appears to be an environment issue. configure is told to use
CC='ccache gcc', but the linker it finds is cc_r. This mismatch
causes the threads linking problem. I believe passing --with-gcc to
configure will fix the problem. Another possible fix would be to set
LD='gcc', but I'm not sure if that works.

Anders, can you make the change to add --with-gcc for AIX?

Neal

[Quest]

Post by Neal Norwitz

Post by Neal Norwitz
3) AIX fails to link. This works for me when built manually.
So there must be some difference in the environment. The
problem is that -lpthread is not present when linking.
I'll try to look into this. If I have write privileges
to the build area, I should be able to fix the problem.

This appears to be an environment issue. configure is told to use
CC='ccache gcc', but the linker it finds is cc_r. This mismatch
causes the threads linking problem. I believe passing --with-gcc to
configure will fix the problem. Another possible fix would be to set
LD='gcc', but I'm not sure if that works.
Anders, can you make the change to add --with-gcc for AIX?

It seems that hal has gone belly up, so we'll have to find another AIX
machine to do our dirty work. I'll use dupond for manual builds for
the time being.

[Quest]

110 700Mbytes CD worth of backup later ...

Post by Neal Norwitz
The snake farm is slowly getting into better shape.
There are still several problems, though.
1) The latest results page doesn't always show the most recent
build for each architecture. For example, right now the most
recent build for AIX is on 4/4, then most recent builds for
other boxes are the 10th. Could we change the script which
produces the page to be more current? The oldest build goes
back to Jan 20th. That doesn't seem too useful.

Are you sure there are more recent builds? It's not just that there
aren't any more recent builds for that machine? The code that builds
the page can be found in CVS at
http://cvs.lysator.liu.se/viewcvs/viewcvs.cgi/. Look in the xenofarm
project, xenofarm/projects/python/. The page is built by webmaker.py.

Post by Neal Norwitz
2) Solaris 8 test_time fails. This is a problem I've been working
on which deals with tzset. <http://python.org/sf/706707>

*reading* Brr, I don't envy you that one.

[snip]

Post by Neal Norwitz
applications cannot be threaded
I'm guessing this is a kernel/glibc/configure issue.
I haven't looked into this, but will eventually get around
to it. The other alpha passes all the tests I believe.

Not right now, it doesn't. There is a SEGV in the tests. I'm looking
into that. See separate message for more info.

Post by Neal Norwitz
5) Many roxen.com builds failing. These all seem to be due
to compiler issues.

I don't have accounts to these boxen, but if you want to work on them,
talk to Henrik "Grubba" Grubbström (see the machineid.txt in one of
the result folders for a mail address). He's not directly involved,
but if you ask nice, he'll help you out.

Post by Neal Norwitz
6) HP-UX--unknown, there isn't a current entry on the page.

I stopped it a while ago because it hanged the tests. I clean forgot
to start it again. Let's see ...
--
Anders "Quest" Qvist
http://www.lysator.liu.se/~sfarmer

[Quest]

Post by Quest

Post by Neal Norwitz
6) HP-UX--unknown, there isn't a current entry on the page.

I stopped it a while ago because it hanged the tests. I clean forgot
to start it again. Let's see ...

Hm, there was a hang in the tests, but it seems to be gone now. I'll
reactivate the builds on tylor, and we'll see what happens.

[Neal Norwitz]

Post by Quest

Post by Quest

Post by Neal Norwitz
6) HP-UX--unknown, there isn't a current entry on the page.

I stopped it a while ago because it hanged the tests. I clean forgot
to start it again. Let's see ...

Hm, there was a hang in the tests, but it seems to be gone now. I'll
reactivate the builds on tylor, and we'll see what happens.

I don't think I fixed the actual problem/bug. But I think I fixed
the hang at least for HP-UX.

Neal