Closed tailhook closed 7 years ago
withScmChangesOnly flag doesn't prevent pulling from repository. nci tries to get new changes and then skip build if only withScmChangesOnly is true and no scm changes.
There should another reason e.g. maybe it's wrong hook and when nci checks for changes they are absent.
Well, I assumed this from this line of code
Also nci-base-webhook
just does builds.create, and I clearly see that builds triggered by user do git pull
and builds triggered by scheduler too. But not builds from webhook.
Any other pointers? Should I close this one and open an issue in nci-base-webhook?
Scheduler also does only builds.create
I just checked github-webhook:
git pull
works fine
Okay, here are three builds:
package verwalter-render-0.3.4-3-g94dfaba-amd64 is already in repository
means we have already built it (you can see the whole log at the link). The version is got from git describe
in the build94dfaba
from error message in build 38 linked above: http://mglawica.org:7000/builds/37In github logs I see hook with code 200 that initiated the build, webhook request content follows. Headers:
Request URL: http://mglawica.org:7000/webhooks/verwalter/github/xxxx
Request method: POST
content-type: application/json
Expect:
User-Agent: GitHub-Hookshot/0b0c52f
X-GitHub-Delivery: 34db4680-33bd-11e6-95f2-f0a6c5e21353
X-GitHub-Event: push
Body (partial, note revisions):
{
"ref": "refs/heads/master",
"before": "94dfabaf8baf45ad9490dea68c5c403b233b4980",
"after": "2b8c9557e444ea6dc98b670f1086b9524d93f4cf",
...
The only issue I see at github log is the following "push" event with a tag, which returns error code 406. I assume that it's fine.
Well, there is a git pull
in the log, so I'm not sure what the issue here. Maybe it tries to pull too fast? I.e. we need to sleep a little? :) Any ideas how to debug the issue?
@tailhook could you check that with latest nci (1.0.9 currently)? Some things changed, especially the way nci implements work with git scm.
Yes, it works now. Not sure since what version, but I'm closing it anyway.
great to hear that, thanks for checking it
On 21 Apr 2017, at 19:56, Paul Colomiets notifications@github.com wrote:
Yes, it works now. Not sure since what version, but I'm closing it anyway.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.
I think this flag should only skip build if there are no changes. But it looks like if the
withScmChangesOnly
is false, nci doesn't even update the repo to find out whether the latter has any changes.In particular, I have a problem with git webhook, where I can't set
withScmChangesOnly
flag at all (it's always false). So it always rebuilds old version when webhook is triggered (builds from scheduler are okay, though)