@danielsz I am getting a strange error, could you explain possibly?
No, sorry, but then again, I'm not the right person to ask. Maybe @danielcompton?
Looks like it's trying to deploy a .finalized file?
Can you share the full deploy log?
@danielcompton where is the deploy log? or do you mean the stack trace?
@flyboarder all of the logs that were part of your deploy command
but the problem seems to be that you uploaded a .finalized
file which had no checksums
@flyboarder hmm, I believe we write a .finalized
file to the tmp upload repo, but it should be excluded from the checksum check. I'll take a look
@flyboarder it looks like the cookie store for the aether client is being reused between deploy calls. Are you doing these deploys from a repl?
if the cookies are reused, then the same tmp repo is used, which is what appears to be happening here (all the snapshot deploys are ending up in the same tmp)
not that deploying from the repl is a bad thing, I'm just trying to figure out the cause here
@tcrawley I think you are correct, i have only found reference to this file in the clojars-web repo, and I am deploying from boot continuously
ok, I'll take a look and see what we can do to fix this
I'd like to support that usage
in the meantime, you should be able to deploy by invoking boot from the command-line, if that's an option
yeah it works the first go round, but when the watch task reruns the push command it breaks
and it looks like this only affects snapshots, from looking at the code
ok, I think I see a way to fix it
I have not tried with releases, currently using snapshots as I dont want 50 new releases when im testing the deployment
right :)
actually, this would break with releases as well, now that I'm taking a closer look
oh man I spent all day yesterday making sure it wasnt me causing the propblem XD
sorry about that :(
@tcrawley do you want a github issue?
@flyboarder that would be swell, thanks!
@flyboarder thanks. We have a test that is doing almost the exact thing you are, but it succeeds :( Do you have a locally modified boot.aether
or a modified build of pomegranate
, perchance?
our test isn't using boot, but our invocation of aether looks the same as boot's
nope! im basically using this build pipeline boot watch build-jar push-snapshot
which works without the watch command
but on subsequent deployments it fails with that error
is push-snapshot
your task? or is it built-in?
it’s provided by bootlaces
, but it’s just a task that collects clojars creds and uses the built-in push task
https://github.com/adzerk-oss/bootlaces/blob/master/src/adzerk/bootlaces.clj#L76-L80
yeah, that does nothing special
hmm
I tracked the trace stack down to the aether/deploy call
so i thinks it’s the server
yes, definitely, we have a bug. I just want to add a test for it, so I know when I've fixed it/break it in the future
basically, I need to figure out how to get aether/deploy
to reuse a cookie store between calls
we do multiple deploys in the same test, but it doesn't share the cookies
but does under boot, apparently
that makes sense since the worker pod is constructed only once when the task it built
I agree, except two aether/deploy
calls right after each other in the clojars tests don't share :(
@tcrawley I fixed it by monkey-patching the push
task, now uses a new pod each time the task runs, fixed the issue!
heh, "fixed".
good deal. I may have you unfix it and try after I push a fix if you are cool with that
yep, I’m not submitting the patch up to boot, instead im including it with my custom task
coolio