Thread

Re: Switch buffile.c/h to use pgoff_t

Michael Paquier <michael@paquier.xyz> — 2025-12-24T23:42:04Z

On Wed, Dec 24, 2025 at 05:01:57PM +0800, Chao Li wrote:
> Make sense, bytestowrite is not a file offset. So, in the current code,
> availbytes is not a file offset either, but it is defined as pgoff_t, which
> has the same confusion, right? Also bytestowrite is casted to pgoff_t, it's
> the same confusion again.

Yeah, actually this suggestion makes more sense.  availbytes is a
computation made of a maximal size and an offset, so defining it as an
offset from the start is kind of weird.

Now I don't think that your suggested set of changes could become more
consistent with a few more changes.  For example, what about pos and
nbytes in BufFile?  While ssize_t is more consistent with FileRead()
and FileWrite(), this code is written to care about signedness while
ssize_t has a stricter range per posix, hence could int64 be a better
choice for the whole interface?  int64 is already what we use for
BufFileSize(), which is due to the limit of MAX_PHYSICAL_FILESIZE of 
course.
--
Michael