Merge lp://staging/~jml/pkgme/build-source-package into lp://staging/pkgme

6 -test_command=python -m subunit.run $IDLIST
7 +test_command=PYTHONPATH=`pwd` python -m subunit.run $IDLIST

Is that not implicit?

41 + stdout, stderr = proc.communicate()

Perhaps we want to do something with the returncode? Also reporting the
stdout, stderr if it fails?

Also, I think it should use that new logging framework that some bright spark
wrote to log a message about what it is doing.

We're going to want to do something with GPG keys as well, but I'm fine deferring
that for now.

Thanks,

James

On Wed, Jul 27, 2011 at 5:42 PM, James Westby <email address hidden> wrote:
> Review: Needs Fixing
> 6       -test_command=python -m subunit.run $IDLIST
> 7       +test_command=PYTHONPATH=`pwd` python -m subunit.run $IDLIST
>
> Is that not implicit?
>

Well, sort of. It's there so that the spawned backend scripts will
have the top-level of the tree in their PYTHONPATH, so that they will
use the development version of pkgme libs instead of the system
installed one.

> 41      + stdout, stderr = proc.communicate()
>
> Perhaps we want to do something with the returncode? Also reporting the
> stdout, stderr if it fails?
>
> Also, I think it should use that new logging framework that some bright spark
> wrote to log a message about what it is doing.
>

Good calls both. Will address.

> We're going to want to do something with GPG keys as well, but I'm fine deferring
> that for now.
>

Agreed. Was thinking that. Should have said it.

jml

> On Wed, Jul 27, 2011 at 5:42 PM, James Westby <email address hidden>
> wrote:
> > Review: Needs Fixing
> > 6       -test_command=python -m subunit.run $IDLIST
> > 7       +test_command=PYTHONPATH=`pwd` python -m subunit.run $IDLIST
> >
> > Is that not implicit?
> >
>
> Well, sort of. It's there so that the spawned backend scripts will
> have the top-level of the tree in their PYTHONPATH, so that they will
> use the development version of pkgme libs instead of the system
> installed one.

Does that only apply if not using virtualenv + setup.py develop?

Thanks,

James

 * debug stuff done in another branch, merged to trunk
 * don't prompt for gpg passphrase during test run
 * actual tests to show we create these files
 * don't run lintian, because, you know, why bother?
 * raise an error if retcode is non-zero
 * basic infrastructure for showing nice errors, rather than stack traces
 * lots of debugging stuff

37 + except PkgmeError, e:
38 + sys.stderr.write("ERROR: %s" % (e,))
39 + return 3

Don't we have a --debug option now? Could it turn this in to a "raise" or something?

78 + if stdout:
79 + trace.debug('D: %s' % (stdout.rstrip('\n'),))

Why the prefix on the message there?

Otherwise this looks great, thanks. It's certainly a valuable feature to have, even
if there are some more details to work out later.

Thanks,

James

Changed the error to raise when debug is set.

The prefix is to make the output more readable. I started without a prefix, then added one and thought the output was more useful.