> It would be, except that we won't handle more than ONE size_t's worth > of data at a time within a single bucket. > > The patch to normalize to one size_t's worth of data took a 71 line delta. > The patch to normalize to any off_t's worth of data was at 500 lines and > growing when I gave up. > > I've said it a dozen times on list, when someone writes a patch that > builds clean [and is correct, I suppose there is a distinction there :-] > then I'll entertain the change. I'm not surprised there have been no > takers, I tried myself long before I gave up and worked in size_t's. > > Note that sendfile would -never- succeed on more than a size_t's > worth ... for that matter, I believe the cap is 56MB by one of the > IBM coder's own experiments.
Humm... I would be suprised if sendfile has a limit like this. If it does, then it is a bug in the OS IMHO. In principle, sendfile should work on any file supported by the OS... Bill > > At 09:48 PM 4/1/2002, you wrote: > >This looks broken to me (at least on Windows). apr_off_t is an int64 and > >apr_size_t is an > >int. > > > >APU_DECLARE(apr_bucket *) apr_bucket_file_create(apr_file_t *fd, > > apr_off_t offset, > > apr_size_t len, > > apr_pool_t *p, > > apr_bucket_alloc_t *list) > > > >I think both the len and offset should be apr_off_t. > > > >Bill > >
