speaking of continuous integration...

[C. Titus Brown]

Two notes from pony-build -- http://lyorn.idyll.org/ctb/pb-dev/pygr/

First, pygr is working fine for me under Windows/python2.6:

http://lyorn.idyll.org/ctb/pb-dev/pygr/detail?result_key=1555

Second, there's some sort of weird problem happening for me with
python2.6 under Mac OS X:

http://lyorn.idyll.org/ctb/pb-dev/pygr/detail?result_key=1609

Even though it looks like it might be some sort of database error, it
doesn't occur under python2.5, oddly enough:

http://lyorn.idyll.org/ctb/pb-dev/pygr/detail?result_key=1533

It may be a problem caused by running pygr/python2.5 tests first,
followed by pygr/python2.6 tests... but I don't know why that would be,
either. I haven't had a chance to duplicate it on my laptop yet, but
I'm throwing it out there as something we should look at.

Mysteries wrapped in a conundrum wrapped in bacon. Sigh.

cheers,
--titus
--
C. Titus Brown, c...@msu.edu

[C. Titus Brown]

On Wed, Sep 09, 2009 at 06:02:36AM -0700, C. Titus Brown wrote:
-> Second, there's some sort of weird problem happening for me with
-> python2.6 under Mac OS X:
->
-> http://lyorn.idyll.org/ctb/pb-dev/pygr/detail?result_key=1609
->
-> Even though it looks like it might be some sort of database error, it
-> doesn't occur under python2.5, oddly enough:
->
-> http://lyorn.idyll.org/ctb/pb-dev/pygr/detail?result_key=1533
->
-> It may be a problem caused by running pygr/python2.5 tests first,
-> followed by pygr/python2.6 tests... but I don't know why that would be,
-> either. I haven't had a chance to duplicate it on my laptop yet, but
-> I'm throwing it out there as something we should look at.

I can duplicate the error from a shell on the same machine (i.e. it's
not a CI environment bug) but I can't run the tests on my own laptop
yet; I have to install bsddb for python2.6 on my laptop. I'll see if I
can test Marek's bsddb patch at the same time.

Does anyone else have python2.6+BSDDB running on their Mac OS X machine?

tnx,

[Christopher Lee]

On Sep 9, 2009, at 7:11 AM, C. Titus Brown wrote:

> I can duplicate the error from a shell on the same machine (i.e. it's
> not a CI environment bug) but I can't run the tests on my own laptop
> yet; I have to install bsddb for python2.6 on my laptop. I'll see
> if I
> can test Marek's bsddb patch at the same time.
>
> Does anyone else have python2.6+BSDDB running on their Mac OS X
> machine?

I have Mac OS 10.5.8 and Python 2.6 installed via fink; "import bsddb"
works on this install without separately installing BSDDB as an
external package. I am not able to reproduce this bug; all tests pass
both on master and my latest sscursor branch.

Do you have reason to believe this test failure is specific to Python
installs where bsddb doesn't import successfully by default, and with
the external BSDDB package installed?

-- Chris

[C. Titus Brown]

On Wed, Sep 09, 2009 at 11:44:47AM -0700, Christopher Lee wrote:
->
->
-> On Sep 9, 2009, at 7:11 AM, C. Titus Brown wrote:
->
-> > I can duplicate the error from a shell on the same machine (i.e. it's
-> > not a CI environment bug) but I can't run the tests on my own laptop
-> > yet; I have to install bsddb for python2.6 on my laptop. I'll see
-> > if I
-> > can test Marek's bsddb patch at the same time.
-> >
-> > Does anyone else have python2.6+BSDDB running on their Mac OS X
-> > machine?
->
-> I have Mac OS 10.5.8 and Python 2.6 installed via fink; "import bsddb"
-> works on this install without separately installing BSDDB as an
-> external package. I am not able to reproduce this bug; all tests pass
-> both on master and my latest sscursor branch.
->
-> Do you have reason to believe this test failure is specific to Python
-> installs where bsddb doesn't import successfully by default, and with
-> the external BSDDB package installed?

The Python 2.6/bsddb install on the lab iMac (the CI machine that is
failing) was done "properly" in the sense that 'import bsddb' works
directly.

OK, that suggests that it's something weird about the machine itself.
I'll look into it.

thanks,

[Christopher Lee]

By the way, the test failures reported in the logfile you posted were
specifically in the sqlite tests, so I guess the bsddb issue may not
be relevant to this:

ERROR: test_mapview_inverse (sqltable_test.SQLiteTable_Test)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/private/tmp/tmpar7a3e/pygr/tests/sqltable_test.py", line 172,
in test_mapview_inverse
assert self.sourceDB[3] == r[self.targetDB[99]]
File "/private/tmp/tmpar7a3e/pygr/pygr/sqlgraph.py", line 1847, in
__getitem__
% str(k))
KeyError: ' not found in MapView, or not unique'

[C. Titus Brown]

On Wed, Sep 09, 2009 at 11:53:16AM -0700, Christopher Lee wrote:
-> By the way, the test failures reported in the logfile you posted were
-> specifically in the sqlite tests, so I guess the bsddb issue may not
-> be relevant to this:
->
-> ERROR: test_mapview_inverse (sqltable_test.SQLiteTable_Test)
-> ----------------------------------------------------------------------
-> Traceback (most recent call last):
-> File "/private/tmp/tmpar7a3e/pygr/tests/sqltable_test.py", line 172,
-> in test_mapview_inverse
-> assert self.sourceDB[3] == r[self.targetDB[99]]
-> File "/private/tmp/tmpar7a3e/pygr/pygr/sqlgraph.py", line 1847, in
-> __getitem__
-> % str(k))
-> KeyError: ' not found in MapView, or not unique'

Right, I guess I managed to confuse you, too ;) -- I don't run the test
suite under Python2.6 on my laptop at all, because the bsddb errors
dominate. So I'm going to install bsddb, then run the tests to see if I
can duplicate the errors on the iMac (which may or may not have anything
to do with bsddb).

[C. Titus Brown]

On Wed, Sep 09, 2009 at 07:11:14AM -0700, C. Titus Brown wrote:
->

-> On Wed, Sep 09, 2009 at 06:02:36AM -0700, C. Titus Brown wrote:
-> -> Second, there's some sort of weird problem happening for me with
-> -> python2.6 under Mac OS X:

-> ->
-> -> http://lyorn.idyll.org/ctb/pb-dev/pygr/detail?result_key=1609
-> ->
-> -> Even though it looks like it might be some sort of database error, it

-> -> doesn't occur under python2.5, oddly enough:

-> ->
-> -> http://lyorn.idyll.org/ctb/pb-dev/pygr/detail?result_key=1533
-> ->
-> -> It may be a problem caused by running pygr/python2.5 tests first,

-> -> followed by pygr/python2.6 tests... but I don't know why that would be,
-> -> either. I haven't had a chance to duplicate it on my laptop yet, but
-> -> I'm throwing it out there as something we should look at.
->
-> I can duplicate the error from a shell on the same machine (i.e. it's
-> not a CI environment bug) but I can't run the tests on my own laptop
-> yet; I have to install bsddb for python2.6 on my laptop. I'll see if I
-> can test Marek's bsddb patch at the same time.
->
-> Does anyone else have python2.6+BSDDB running on their Mac OS X machine?

OK, all I could find was an error caused by failure to properly delete
some tables. Next time it crops up I will hunt it down and terminate
the problem with extreme prejudice.

I don't know why it was 2.6 specific though, which is a bit worrisome.

cheers,

[C. Titus Brown]

This turned out to be a problem of some sort caused by my sqlite module.
Upgrading from Python 2.6.1 to Python 2.6.4 fixed it.

It was very odd -- the attached Python script returned no results for
that sql query, which played merry havoc with the graphview stuff.

Anyway, the pony-build CI machines should now be all green.