Remove PYTHONPATH requirement to execute run_slave.py.

Remove PYTHONPATH requirement to execute run_slave.py.
Make the error message in buildbot.tac more useful when the master cannot be determined automatically.

By putting the python paths at one place, it's easier to switch the values like
upgrading to a newer buildbot version.

TEST=none
BUG=none

[nsylvain, 2010-11-05 01:24:00]

http://codereview.chromium.org/4558002/diff/1/5
File slave/run_slave.py (right):

http://codereview.chromium.org/4558002/diff/1/5#newcode167
slave/run_slave.py:167: sys.path.append(os.path.join('..', 'third_party',
'buildbot_7_12'))
The reason why it was in makefile is that we could have created a "make
start_bbot8.2" or "run_slave_8_2.bat".

Now you are pretty much forcing us to fork run_slave.py instead, which is going
to be messy once we start having half the masters using one version, and half
the masters using another version.

[M-A Ruel, 2010-11-05 01:51:58]

http://codereview.chromium.org/4558002/diff/1/5
File slave/run_slave.py (right):

http://codereview.chromium.org/4558002/diff/1/5#newcode167
slave/run_slave.py:167: sys.path.append(os.path.join('..', 'third_party',
'buildbot_7_12'))
On 2010/11/05 01:24:00, nsylvain wrote:
> The reason why it was in makefile is that we could have created a "make
> start_bbot8.2" or "run_slave_8_2.bat".
> 
> Now you are pretty much forcing us to fork run_slave.py instead, which is
going
> to be messy once we start having half the masters using one version, and half
> the masters using another version.

My plan is to switch all the slaves to 0.8.2 but keep the masters at 0.7.12 and
slowly switch the masters to 0.8.2.

Disagree?

[nsylvain, 2010-11-05 01:55:39]

On Thu, Nov 4, 2010 at 6:51 PM, <maruel@chromium.org> wrote:

>
> http://codereview.chromium.org/4558002/diff/1/5
> File slave/run_slave.py (right):
>
> http://codereview.chromium.org/4558002/diff/1/5#newcode167
> slave/run_slave.py:167: sys.path.append(os.path.join('..',
> 'third_party', 'buildbot_7_12'))
> On 2010/11/05 01:24:00, nsylvain wrote:
>
>> The reason why it was in makefile is that we could have created a
>>
> "make
>
>> start_bbot8.2" or "run_slave_8_2.bat".
>>
>
>  Now you are pretty much forcing us to fork run_slave.py instead, which
>>
> is going
>
>> to be messy once we start having half the masters using one version,
>>
> and half
>
>> the masters using another version.
>>
>
> My plan is to switch all the slaves to 0.8.2 but keep the masters at
> 0.7.12 and slowly switch the masters to 0.8.2.
>
I was not under the impression that the slave code was backward compatible
with the old master.

I still think that we need a way to tell what version we use, and to make it
easy to switch. You may want to switch
chromium, but nacl will lag behind, and I assume chromeos will lag even
more... and this is without even talking
about master.chrome, which will take ages before it get switched.


>
> Disagree?
>
>
> http://codereview.chromium.org/4558002/show
>