Updater: target hash calculation should not read the whole file in memory

[jku]

Description of issue or feature request:

This is not a bug (although it could be on memory limited client device), but a performance improvement:

After PR #1202 there is only one place where the updater loads the whole target file in memory. We should avoid doing that as targets could be very large and memory could be limited.

Current behavior:

_check_hashes() does this:

digest_object = securesystemslib.hash.digest(algorithm)
digest_object.update(file_object.read())
computed_hash = digest_object.hexdigest()

Expected behavior:
something like handwaves

digest_object = securesystemslib.hash.digest(algorithm)
while True:
    chunk = file_object.read(CHUNK_SIZE)
    if not chunk:
        break
    digest_object.update(chunk)
computed_hash = digest_object.hexdigest()

or even more simply just let SSLib handle this with its default chunk size:

digest_object = securesystemslib.hash.digest_fileobject(file_object, algorithm)
computed_hash = digest_object.hexdigest()

[trishankatdatadog]

Good find, thanks! On top of that, shouldn't securesystemslib already offer a function for checking hashes?

[jku]

shouldn't securesystemslib already offer a function for checking hashes?

Yes my last example uses a sslib function for calculating a fileobject hash, I just forgot to include 'securesystemslib.hash.' prefix there (now fixed)

Or do you mean actually comparing the hashes?