svn revision 4485 and mod_python broken

[Michael]

Hi All,


After svn up to revision 4485 mod_python stop working:

giving an error:
PythonHandler django.core.handlers.modpython: EnvironmentError: Environment variable DJANGO_SETTINGS_MODULE is undefined.


Rollback to revision 4484 fixes this problem.

Is anyone having the same problem?


System Debian Sarge, apache 2.0.54, python 2.3


--
--
Michael

[Malcolm Tredinnick]

On Mon, 2007-02-12 at 12:19 +1300, Michael wrote:
> Hi All,
>
>
> After svn up to revision 4485 mod_python stop working:
>
> giving an error:
> PythonHandler django.core.handlers.modpython: EnvironmentError:
> Environment variable DJANGO_SETTINGS_MODULE is undefined.
>
>
> Rollback to revision 4484 fixes this problem.
>
> Is anyone having the same problem?

Oh, dear. I can see how this might be happening: When we import gettext,
we first check to see if we really want to i18n infrastructure (it slows
things down a little, so is worth omitting if you are not going to use
this). In order to work this out -- and hence load the right gettext()
function -- we need to check the USE_I18N setting.

I'll back out that changeset momentarily until we can think it through a
bit more.

[Russell: if that change was necessary to unbreak the tests, can you
drop in a reply about how to see the problem. If you were just doing it
to get a head-start, temporarily reverting should not hurt.]

Regards,
Malcolm

[Malcolm Tredinnick]

On Mon, 2007-02-12 at 10:42 +1100, Malcolm Tredinnick wrote:
[...]

> I'll back out that changeset momentarily until we can think it through a
> bit more.

This has now been done. Updating to [4486] (current tip of trunk) should
work. If not, please let us know.

Regards,
Malcolm

[Russell Keith-Magee]

On 2/12/07, Malcolm Tredinnick <mal...@pointy-stick.com> wrote:
>
> On Mon, 2007-02-12 at 12:19 +1300, Michael wrote:
> > Rollback to revision 4484 fixes this problem.
> >
> > Is anyone having the same problem?
>
> Oh, dear. I can see how this might be happening: When we import gettext,
> we first check to see if we really want to i18n infrastructure (it slows
> things down a little, so is worth omitting if you are not going to use
> this). In order to work this out -- and hence load the right gettext()
> function -- we need to check the USE_I18N setting.

Bother. Well that puts marzipan in my pie plate bingo... Apologies to all.

I'm surprised this one wasn't picked up by the system tests, though.

> I'll back out that changeset momentarily until we can think it through a
> bit more.
>
> [Russell: if that change was necessary to unbreak the tests, can you
> drop in a reply about how to see the problem. If you were just doing it
> to get a head-start, temporarily reverting should not hurt.]

[4482] had the unit test and a fix for the immediate _() problem. The
revert of [4485] shouldn't break the test (says he who didn't
anticipate the problem in the first place).

Yours,
Russ Magee %-)

[Michael]

Thank you,

it is working again [4490]


--
--
Michael