Re: cvs: php4 /sapi/apache2handler sapi_apache2.c

"Ilia A." <ilia@prohost.org> wrote: > On March 25, 2003 03:40 am, Moriyoshi Koizumi wrote: > > Why did you commit another fix though my patch was already there and maybe > > Ian was looking into this issue? :) > > Because the existing patch (latest CVS) did not work, when posting 30k or more > data it would usually stop after 7.8k or so. The result is because > ap_get_brigade() can return partial data and if you look at the docs for the > AP_MODE_READBYTES flags you'll see it will only read until CRLF is found OR > if the entire buffer is filled. Hmm..? my patch has not been committed yet, that was only in the list archives. Although I don't see no problem with your fix, I think you'd be better off asking the assignee how your patch is before the commit. > > > "Ilia Alshanetsky" <ilia@prohost.org> wrote: > > > + if (tlen == count_bytes || !len) { > > > + break; > > > + } > > > > Is this ok? I suppose tlen would't always be equal to count_bytes. > > Yeah this is ok, the idea is to fill the buffer to the end up until the point > there is either no more data to read (len == 0) or the buffer is full (tlen > == count_bytes). As I was making the same patch, there seems nothing documented like apr_brigade_flatten() sets len to zero in case the passed brigade is void. Therefore I think it's potentially dangerous. Moriyoshi