vteromero/vtenc

VTEnc C library

vtenc's Issues

Always free allocated memory in case of error

When a controlled error occurs in a API function and it needs to stop the execution, we must ensure that any allocated memory is freed before exiting.

Specifically, the *encode* and *decode* functions should always call encctx_close and decctx_close before returning.

At the moment, there are some leak memory when running the unit tests:

valgrind --leak-check=yes ./tests/unit/unit_tests

ERROR SUMMARY: 21 errors from 21 contexts (suppressed: 0 from 0)

Release 0.1.0

• Tests for random sequences #15
• Refactor of count_zeros_at_bit_pos* functions to improve encoding performance #16
• Fix bug on wrongly using memcmp on some testing programs #17 #14
• Encode lower bits for clusters of length 1 #18

Question about decoding speed

Looks like a great library, I'm exciting to give it a spin!

When I looked at the benchmarks, I was impressed by the high encoding speed. Could you provide an insight why the decoding speed lags an order or magnitude behind the other implementations? Is this purely because the current state of the implementation or is there fundamental algorithmic limitation?

memcmp wrongly used

memcmp compare stream of bytes, not arrays of other bigger types like uint16_t, uint32_t or uint64_t. You can still use it with those types, but you need to properly calculate the number of bytes to compare For example:

const size_t len = 4;
uint32_t a[len] = {0, 1, 2, 3};
uint32_t b[len] = {0, 10, 20, 30};

// WRONG: will output "Arrays are equal"
if (memcmp(a, b, len) == 0) {
  printf("Arrays are equal\n");
} else {
  printf("Arrays are different\n");
}

// CORRECT: will output "Arrays are different"
if (memcmp(a, b, len * sizeof(uint32_t)) == 0) {
  printf("Arrays are equal\n");
} else {
  printf("Arrays are different\n");
}

There are a couple of places where memcmp is being used incorrectly:

• tests/gov2.c
• tests/timestamps.c