On 2/28/07, William A. Rowe, Jr. <[EMAIL PROTECTED]> wrote:
There's a signficant difference between not calling flush 'anytime soon' and lazy writes, however :)
No - not really.
If this was a flag to apr_sdbm_open, or was modified to interact with the existing locking logic, I'd have much more faith that this is a reasonable approach.
I would appreciate it if you present a specific case where this breaks something rather than vague generalities about how it might break something. As I said earlier, I view this as a very low-risk change as SDBM uses APR's file locking semantics and doesn't support concurrent reader/writer combinations. If you want that facility via BDB, you need newer versions of BDB (and even that tends to barf all over itself on load). If we do determine there is a case where it would break when SHARELOCK is present (which I'm doubtful works given it doesn't even try to flush), then we can either not set the buffering when APR_SHARELOCK is set, or move the flag into apr_dbm's usage of SDBM - which, I hope we agree, certainly doesn't permit multiple reader/writers with SDBM at all. -- justin
