refactor fileio [CPP-472]

[notoriaga]

With a 1mb file, before:

(base) {9:47}~/projects/console_pp:main ✗ ➭ time cargo fileio write ./test-file-1mb /persistent/steve.txt tcp 10.1.52.164 --port 55555
    Finished release [optimized + debuginfo] target(s) in 0.12s
     Running `target/release/fileio write ./test-file-1mb /persistent/steve.txt tcp 10.1.52.164 --port 55555`
Writing 100.00%...
File written successfully.
cargo fileio write ./test-file-1mb /persistent/steve.txt tcp 10.1.52.164    1.37s user 1.10s system 0% cpu 6:52.37 total

After:

(base) {9:38}~/projects/console_pp:steve/fileio-fix ✗ ➭ time cargo fileio write ./test-file-1mb /persistent/steve.txt tcp 10.1.52.164 --port 55555
    Finished release [optimized + debuginfo] target(s) in 0.17s
     Running `target/release/fileio write ./test-file-1mb /persistent/steve.txt tcp 10.1.52.164 --port 55555`
Writing 100.00%...
File written successfully (1048576 bytes).
cargo fileio write ./test-file-1mb /persistent/steve.txt tcp 10.1.52.164    0.55s user 0.36s system 1% cpu 57.986 total

I think the original was so slow because by not using the window_size we'd send writes for parts of the file that are really far apart causing a lot of seeking back and forth on the piksi's end, but not 100% sure.

[john-michaelburke]

Trying to think of scenarios where this may happen. If someone were to shoot off an upgrade process then additionally try to send a file to the device with the bin possibly? If the result was benign maybe debug would make more sense here?

[notoriaga]

Yeah that's sort of the scenario I was thinking of. Should be harmless if it happens so I'll downgrade it to a debug

[john-michaelburke]

This is a great addition we should probably implement this throughout the app as well if we continue seeing reliability issues.

[notoriaga]

Yeah it makes using gdb a bit easier. You can also set the stack size for the thread. The default is 2mb, we could probably shave a bit of memory usage off too by going thru and trying smaller sizes.