Recent travis failures

[dhalbert]

Travis builds have been failing recently. Example:
https://travis-ci.com/adafruit/circuitpython/jobs/178330132

0.58s$ LAST_TAG=`git ls-remote --quiet --tags --sort=version:refname | egrep -o "refs/tags/[0-9]+.*\$" | tail -n 1`
3.11s$ git fetch --shallow-exclude=4.0.0-alpha.1 || git fetch --unshallow
remote: Enumerating objects: 8166, done.
remote: Counting objects: 100% (8164/8164), done.
remote: Compressing objects: 100% (2944/2944), done.
remote: Total 7180 (delta 5557), reused 5582 (delta 4151), pack-reused 0
Receiving objects: 100% (7180/7180), 3.65 MiB | 17.09 MiB/s, done.
Resolving deltas: 100% (5557/5557), completed with 749 local objects.
fatal: shallow file has changed since we read it
fatal: error in object: unshallow 4dfba2f8acec7d76fa61c13fcec662b6d957f69a
The command "git fetch --shallow-exclude=4.0.0-alpha.1 || git fetch --unshallow" failed and exited with 128 during .

I tried this on my local Ubuntu machine (running bionic, not xenial) immediately seeing this failure, so I have the same commit, etc., with exactly the same git version (2.20.1) (I used a git ppa for that). I do not get the error

fatal: shallow file has changed since we read it

but my output is otherwise identical up to that point, including the exact same data such as Enumerating objects: 8166. So this is odd. I wonder if it's some kind of filesystem date stat issue or something.

Perhaps we should change back to depth: 6 or something like that for now to restore the builds. Any comments?

[dhalbert]

These problems are new. .travis.yml hasn't changed in a couple of months or so.