On Thu, Sep 29, 2022 at 12:24:22PM +0200, Stefan Sperling wrote:
> Well, our loops certainly take great care not to index result->chunks.head
> in the way you are proposing. For example:
> 
>         for (c_idx = cc->chunk.start; c_idx < cc->chunk.end; c_idx++) {
>                 const struct diff_chunk *c = &result->chunks.head[c_idx];
> 
> This keeps c_idx < chunk.end, not <= chunk.end.
> 
> Can you show what you are seeing in the debugger? What are the contents
> of the final chunk? In particular, what is that chunk's type?

Some more hints:

Keep in mind that an arraylist might have more memory allocated than
needed to represent the elements of the array (the elements being diff
chunks in this case).

What is important is that we never index at or beyond the chunk arraylist's
len field. If there is a chunk at index 'len' it should not be considered
valid. Because it is not part of the user-visible data stored on the array.

The 'chunk.end' index points to the first chunk array element following this
chunk. If chunk.end equals the chunk arraylist's len, we are at end-of-file.