Infinte loop with filecache and check_files=True

[edwardlonghurst]

What happened:
Using filecache with check_files=True
Discovered using the abfs filesystem and replicated with the local filesystem.

1. First call: file downloads correcly to cache.
2. File is updated in Azure
3. Second call: CachingFileSystem._check_file correctly identifies there is a new version with detail["uid"] != self.fs.ukey(path)
4. File is downloaded to cache folder
5. cache file is rewritten by ``CachingFileSystem.save_cache" however the uid is not updated
6. WholeFileCacheFileSystem._open calls return self._open(path, mode) kicking off the loop again and repeatedly failing the check in step 3

What you expected to happen:
cache file update includes saving the new uid

Minimal Complete Verifiable Example:
The following prints only Hello, then blows the stack

import fsspec

tf = open("testfile.txt", "w")
tf.write('Hello\n')
tf.flush()

fs = fsspec.filesystem("filecache", target_protocol="file", check_files=True)
with fs.open("file://testfile.txt") as fsfile:
    print(fsfile.read())

tf.write('World\n')
tf.flush()

with fs.open("file://testfile.txt") as fsfile:
    print(fsfile.read())

tf.close()

Anything else we need to know?:
Adding
c["uid"] = cache[k]["uid"]
below
c["time"] = max(c["time"], cache[k]["time"])
in CachingFileSystem.save_cache appears to solve the problem, however I do not know enough about fsspec to know if it introduces other problems

Environment:

• fsspec version: 2021.7.0
• Python version: 3.8.8
• Operating System: Windows 10
• Install method (conda, pip, source): conda

[martindurant]

Sorry to be late to notice this. What you say sounds reasonable, and I'd be happy to see the fix in a PR. Your example code can become a test.

[martindurant]

Ping - were you thinking to submit code to fix this, @edwardlonghurst ?