On Mon, 04 Oct 2004 10:35:47 +0200, jean-frederic clere
<[EMAIL PROTECTED]> wrote:
>
>
> Jeff Trawick wrote:
> > On 1 Oct 2004 16:03:09 -0000, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
> >
> >>jfclere 2004/10/01 09:03:09
> >>
> >> Modified: os config.m4
> >> os/unix os.h unixd.c
> >> Log:
> >> Move the few BS2000 specific in unixd.c
> >
> >
> >> Index: unixd.c
> >> ===================================================================
> >> RCS file: /home/cvs/httpd-2.0/os/unix/unixd.c,v
> >> retrieving revision 1.69
> >> retrieving revision 1.70
> >> diff -u -r1.69 -r1.70
> >> --- unixd.c 24 Apr 2004 19:42:52 -0000 1.69
> >> +++ unixd.c 1 Oct 2004 16:03:08 -0000 1.70
> >> @@ -457,11 +457,26 @@
> >> {
> >> apr_socket_t *csd;
> >> apr_status_t status;
> >> +#ifdef _OSD_POSIX
> >> + int sockdes;
> >> +#endif
> >>
> >> *accepted = NULL;
> >> status = apr_socket_accept(&csd, lr->sd, ptrans);
> >> if (status == APR_SUCCESS) {
> >> *accepted = csd;
> >> +#ifdef _OSD_POSIX
> >> + apr_os_sock_get(&sockdes, csd);
> >> + if (sockdes >= FD_SETSIZE) {
> >> + ap_log_error(APLOG_MARK, APLOG_WARNING, 0, NULL,
> >> + "new file descriptor %d is too large; you probably
> >> need "
> >> + "to rebuild Apache with a larger FD_SETSIZE "
> >> + "(currently %d)",
> >> + sockdes, FD_SETSIZE);
> >> + apr_socket_close(csd);
> >> + return APR_EINTR;
> >> + }
> >> +#endif
> >
> >
> > the old logic was removed on purpose
> >
> > an analogous check is in APR at the point where select() is actually
> > needed; an error will be returned from the apr poll call or the apr
> > send/recv call
> >
> > the APR check covers not just sockets but also pipes used to
> > communicate with CGIs
> >
> > it would be nice if there was a special APR error code to use there
> > with a unique error message though
> >
> > (btw, it is APR that would need to be rebuilt)
> >
> >
>
> In poll/unix/poll.c:
> +++
> if (fd >= FD_SETSIZE) {
> /* XXX invent new error code so application has a clue */
> return APR_EBADF;
> }
> +++
>
yes, introducing a unique error code for this special condition would
result in a meaningful error message that doesn't require browsing the
source to understand the situation
the handling of this is better left to APR so that it applies to all
platforms where a select() with this limitation is used; if we need to
make the information about needing to compile apr differently more
transparent, so be it, as multiple applications can take advantage of
such an enhancement
