Fetching deploy.sh keeps failing about whatwg.org

[annevk]

https://travis-ci.org/whatwg/fetch/builds/269117879

[foolip]

Is this a flaky failure? I can currently access https://resources.whatwg.org/build/deploy.sh just fine.

[foolip]

I've restarted the build to see.

[annevk]

FWIW, I already had restarted it four times while being able to access that resource just fine myself. It's rather unclear what's going on.

[foolip]

Might be because of the new deploy method for resources.whatwg.org, perhaps?

[foolip]

But I can't see exactly how that could be, since the resource is there.

[annevk]

Nah, this happened before this switch as well.

[foolip]

whatwg/fetch#587 didn't give much of a clue :/

[foolip]

Has this been happening for all repos?

[annevk]

Yeah, they just haven't been updated much recently so I haven't hit it a lot.

[annevk]

Saw it on whatwg/dom too and whatwg/encoding. https://travis-ci.org/whatwg/encoding/builds/265476353 is slightly different though. Maybe it's @sideshowbarker's server that is unresponsive and the problem is with fetching that happens in the deploy.sh script?

[domenic]

To me the first failing build looks like the deploy.sh fetch succeeds but the first curl inside deploy.sh fails. That would be this one:

SERVICE_WORKER_SHA=$(curl https://api.github.com/repos/whatwg/resources.whatwg.org/contents/standard-service-worker.js \
                     -H "Accept: application/vnd.github.v3+json" \
                     | grep -Po '(?<="sha": ")[^"]*') # Hacky JSON parsing but works for SHAs

That gives a 404 since the repo move, so that's probably the issue here.

[foolip]

Ah, good, I can fix that.

[annevk]

That does not really explain the whatwg/encoding failure though, which was before the repository move.

[annevk]

Or the other failures before the repository move.

[foolip]

Do those failures look the same in the logs? Maybe try adding --verbose to them, might not all be the same cause...

[annevk]

They all look the same more or less. You can't really use verbose since you'd have to add that to the central script.

[annevk]

Anyway, I guess we can first fix the service-worker thing and go from there.

[annevk]

So @foolip sorted this out in #73.