bug: Integer underflow risk in reader fetch partition_point

[rich7420]

Describe the bug

The fetch method uses partition_point to find the corresponding merged range, but doesn't check if partition_point returns 0. When partition_point returns 0, 0 - 1 causes integer underflow (becomes usize::MAX), leading to out-of-bounds access and panic.

opendal/core/src/types/read/reader.rs

Line 179 in fc5725a

let idx = merged_ranges.partition_point(|v| v.start <= range.start) - 1;

Steps to Reproduce

none

Expected Behavior

Check if partition_point returns 0 before subtracting 1, and validate that each range is within the merged ranges.

Additional Context

No response

Are you willing to submit a PR to fix this bug?

• Yes, I would like to submit a PR.

[rich7420]

I could submit PR as you think it should be solved. ~

[Xuanwo]

Thank you for pointing that out. I'm interested in knowing in which case this will happen.

[rich7420]

I think the bug occurs when partition_point returns 0, which happens when all merged_ranges have a start value greater than the current range.start.

[Xuanwo]

merged_ranges have a start value greater than the current range.start.

How could this happen? Could you share an example? merged_ranges are generated from the given Vec<range>.

[rich7420]

@Xuanwo Thank you for the clarification. You're absolutely right - I apologize for the misunderstanding. Since merged_ranges are generated from the input ranges, partition_idx returning 0 should not be possible in practice. I'll close this unnecessary issue and PR.

[Xuanwo]

@Xuanwo Thank you for the clarification. You're absolutely right - I apologize for the misunderstanding. Since merged_ranges are generated from the input ranges, partition_idx returning 0 should not be possible in practice. I'll close this unnecessary issue and PR.

Hi @rich7420, it’s fine to use the AI Code Agent to help with tasks, but we don’t encourage communicating directly with its responses. Sorry if I was wrong to assume the response was AI-generated. My goal is for us to build real communication so you can learn more and gain deeper context about the project, not just rely on AI-generated replies.

[rich7420]

thanks @Xuanwo ! I'll try to dig deeper! Thanks a lot for the reminding.

[Xuanwo]

thanks @Xuanwo ! I'll try to dig deeper! Thanks a lot for the reminding.

Having fun!