Repository navigation
Example in find_chunk_edge() can panic 'range start index' #29
Description
Using some example code here:
https://docs.rs/hash-roll/0.3.0/hash_roll/trait.Chunk.html#tymethod.find_chunk_edge
It turns out that discard_ct could result in data.len() + 1 in size as seen in this code:
https://docs.rs/hash-roll/0.3.0/src/hash_roll/fastcdc.rs.html#75
The only happens when the boundary is on the last byte and can result in i equaling data.len() causing &data[..data.len() + 1] resulting in a panic.
The fix is simple, I might suggest adding something like this to the docs:
if discard_ct > data_buf.len() {
break;
}Thanks for the hard work in providing this library!
Activity
safinaskar commentedon Jun 17, 2023
@jmesmon , example in https://docs.rs/hash-roll/0.3.0/hash_roll/trait.Chunk.html#tymethod.find_chunk_edge seems to be absolutely incorrect. It initializes
prev_cutto 0 and then never sets it again. So it seems something is really wrong here. So I'm afraid to use this library, because there is no example I can rely onsafinaskar commentedon Jun 17, 2023
Also, it seems that
&orig_data[prev_cut..g_cut]instead of&orig_data[prev_cut..cut_point]should be. There may be other bugs in this code, so, please, proof read itcodyps commentedon Oct 31, 2026
The FastCDC forced maximum boundary now returns an offset within the supplied buffer, and both APIs use exclusive end offsets. The example also preserves the trailing incomplete chunk. Fixed in ee44de0 and included in #36; regression and fragmented-input tests cover the reported failure. This issue will close when #36 merges. The published 0.3.0 documentation remains unchanged until a new release.