Salut, Philippe. Sadly (or perhaps happily?), I am unable to reproduce this error on any browser I tried.
It’s possible there was a sequence of intermediate steps that led to this error. Are you still seeing the problem? Any other hints you can think of to guide our investigation?
It works now on the same browser (no restart, no change that I’m aware of).
Classify as: mysterious.
There were really no weird intermediate steps (again: that I’m aware of). I just opened https://beta.observablehq.com/@fil/reprojecting-vector-tiles and it was broken (Chrome). I then tried to open it under a private window (still broken) and a secondary browser (Firefox) (still broken). Then created the demo from scratch to make sure it was not that code. (broken).
But hey! https://beta.observablehq.com/@fil/reprojecting-vector-tiles is still broken on my two browsers and on @clhenrick’s browser too
it’s fixed now… and I didn’t change anything
Unfortunately I’m still unable to reproduce this error, but we did recently discover a bug in the runtime, and have deployed a fix. Is the error still occurring for you?
it seems to be working now; could it have been related to that bug?
In this same notebook there is a random error coming from cloudflare/nextzen; sometimes some tiles don’t load. I fixed things to ignore this error, and we can see in this example that the US west coast tile is missing (error 503 in the console).
(I’m not sure how much of this is related to the error we’ve seen with
Of course. I wonder if there was any possibility that this error caused the original one. (I don’t see how.)
The original notebook you posted (369db1eb34ba5e3e which imports @fil/base-map) doesn’t use Nextzen. It’s possible that unpkg was returning an error, but I think that would have broken both of the imports rather than just one, since require caches internally.
My best guess is that this was fixed by our new release of the Observable runtime (which was related to pruning the computation graph and affects the behavior of imports and import-with).