Hi Patrick,
On Thu, Jan 30, 2014 at 4:30 PM, Patrick Schaaf <[email protected]> wrote:
> I don't see that. You mean by the call to ps_files_close() within
> ps_files_open(), and that close() call dropping the lock? That is
> conditional
> on either data->fd < 0 (no file open, just a superfluous call to
> ps_file_close()), or key != lastkey (and then a _different_ file will be
> opened and locked for writing).
>
Next request may read closed session data which is no longer valid.
When it is destroyed or regenerated session with new code, it will be
handled by
lazy_destroy. So it's not a problem.
When session is not destroyed or regenerated, close call in PS_WRITE
unlocks
file then next request starts reading it. This could happen before writing
data by
previous request.
I don't have access to the code right now, so I have to double check though.
Regards,
--
Yasuo Ohgaki
[email protected]