On Tue, Jan 24, 2017 at 01:36:59PM +0100, Michal Novotny wrote:
On Tue, Jan 24, 2017 at 11:44 AM, Pierre-Yves Chibon
<pingou(a)pingoured.fr>
wrote:
On Tue, Jan 24, 2017 at 11:30:33AM +0100, Michal Novotny wrote:
>Â Â Yes, untested changes got into production. We are sorry. I am
currently
>Â Â working on a fix.
If that made it up to production, then we'll need to adjust fedmsg_meta
to
support these messages as well.
fedmsg_meta should work for all our published messages (published and
stored on
datagrepper).
Okay, these messages have all empty body (the "msg" attribute), e.g
`build.end`:
{
"source_name": "datanommer",
"i": 3,
...
"timestamp": 1485166967.0,
"msg_id": "2017-db23086c-2001-496e-a59f-10b92c466ae5",
"topic": "org.fedoraproject.prod.copr.build.end",
"source_version": "0.6.5",
"signature":
"HJEOhP93vZFUk3cjIzggxZqIT+IRaLpKF/t21Kn0AcQ9B1VJEe+myerAAJMZfuXppGQqsFyzcPFx\nu+9p7geI5NqNxnN+diUXNlxbXN9/VN0X3vX7U4mbc0/zLyGcbKWIn/pcskbM5qYC2lJHLov0pMwq\nrbq/B0N3CxBL2og0Fj8=\n",
"msg": {}
}
Does fedmsg_meta need to be adjusted even so?
Yeah, fedmsg_meta should be able to process all the messages stored in
datagrepper, otherwise it breaks datagrepper and a few other apps.
So we'll end up with some not so useful message: "a build ended in copr" but
we
should add them.
The easiest for this type of changes is to use test-driven-development, add the
new message to the test, fill in the expectations and then adjust the processor
to handle those cases.
Pierre