On 2026-11-21 01:39, Bruno Haible wrote:
Also, maybe we should reduce the max even further than 2^31? In clisp/src/win32aux.d there is the comment (from 1998):
This sounds like a bug in a third-party NT 3.51 NFS client. This kind of bug, which sporadically mishandled user-mode buffers not aligned to a multiple of 4, was killed off in MS-Windows 2000's Driver Verifier, which caused such drivers to crash pretty quickly. I doubt whether we still have to worry about these ancient issues; and if we did, I suspect we might have to align read buffers and/or limit reads to 8 KiB to handle the unlikely worst cases (NFSv2 has 8 KiB limits on the wire) and that would have significant performance penalties.
