fix: Address flaky integration test

[aiven-anton]

Trying to figure this one out, I read up again on the documentation of StreamWriter.read() and took extra note of this passage.

[...] return at most n available bytes as soon as at least 1 byte is
available in the internal buffer.

After switching to use .readexactly() instead, the error systematically stops occurring.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (c57a4bd) 96.67% compared to head (c36685e) 96.67%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #130   +/-   ##
=======================================
  Coverage   96.67%   96.67%           
=======================================
  Files          14       14           
  Lines        1023     1023           
  Branches      139      139           
=======================================
  Hits          989      989           
  Misses         26       26           
  Partials        8        8           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[eliax1996]

Make sense! Great finding, I'm wondering if the IO read method could have the same behaviour. I tried to open the class in typing but I cannot see any guarantee nor doc about the read method.
Do you know where we can look for it?

[aiven-anton]

@eliax1996 IO is just a protocol so it doesn't really have opinions of how many bytes the method should return.

Note that there was also a previous related fix for all the lower-level read functions: #117. Prior to that, some of them would silently accept fewer bytes than expected, but the fix makes them fail loudly.

[aiven-anton]

IO is just a protocol so it doesn't really have opinions of how many bytes the method should return.

And this would probably require dependent types to express with the type system 😁