Dear Paul, Thank you for your feedback. You are correct that fts_read does not guarantee that fts_path remains valid after a failure. My initial approach relied on unspecified behavior, which could lead to unreliable results.
As a more robust alternative, I suggest tracking the last successfully visited path to provide context when reporting an error. Would you consider this approach acceptable, or do you have a preferred method for improving the diagnostics in these cases? I am happy to revise the patch and resubmit it for review. Best regards, Shubham Chakraborty On Sun, 19 Apr, 2026, 2:22 am Paul Eggert, <[email protected]> wrote: > I thought that fts_path had unreliable contents after fts_read fails, > which would mean the proposed patch would lead to unreliable behavior. > What am I missing? >
