Jan Boon

No idea why the automated test is failing. This shouldn't affect anything existing since all the behavior is behind the file type flags. Unless there's some corrupt data existing in...

> ``` > ====== results ====== > tests/test_exhaustion.toml:353:failure: test_exhaustion#5#1 (LFS_BLOCK_CYCLES=5) failed > tests/test_exhaustion.toml:[44]:warn: max wear: 19 cycles > tests/test_exhaustion.toml:[45]:warn: avg wear: 0 cycles > tests/test_exhaustion.toml:451:warn: min wear: 0 cycles >...

~~Some thoughts. Perhaps it may be interesting to separate the *size* into *size* and *capacity*, using reserve to set the capacity, truncate to set the size, and then adding support...

Happy with this patch as it is now. Does the job so far in testing for our use case.

> Given that a B-tree would not require invasive pointers in the data-blocks, it seems a B-tree would allow you to create the `lfs_file_reserve` without needing a new data-structure. This...

Traverse will read the first bytes of each block in each file back to front to detect the allocated blocks. So, if your read size equals your block size, it...

Just use a bigger block size that's a multiple of 512, and recalculate appropriately in your implementation callbacks to remap multiple hardware blocks to one lfs block.

@sslupsky Nope. Your buffers only need to be sized to fit your minimum read and write sizes. Your block size can be as large as you like, and should be...

@sslupsky Yes. The OP is using a tiny block size that equals his minimum read size. He should increase his block size, otherwise any traversal will be completely reading every...

It'll read the first 8 bytes in each block of each file, yes. It's just a linked chain of block pointers from back to front for each file. That's how...