Justin Erenkrantz wrote:
>So, I'm thinking the appropriate check is to stop at EOS or
>SENTINEL. There is no good reason to continue processing if we
>see an EOS. But, we have a potentially saved brigade to handle.
>So, what do we do? Perhaps this patch might be better:
>
>Index: modules/filters/mod_include.c
>===================================================================
>RCS file: /home/cvs/httpd-2.0/modules/filters/mod_include.c,v
>retrieving revision 1.166
>diff -u -r1.166 mod_include.c
>--- modules/filters/mod_include.c 2001/12/08 03:14:50 1.166
>+++ modules/filters/mod_include.c 2001/12/14 18:37:48
>@@ -2700,7 +2700,7 @@
> ap_escape_shell_cmd(r->pool, arg_copy));
> }
>
>- while (dptr != APR_BRIGADE_SENTINEL(*bb)) {
>+ while (dptr != APR_BRIGADE_SENTINEL(*bb) || !APR_BUCKET_IS_EOS(dptr)) {
> /* State to check for the STARTING_SEQUENCE. */
> if ((ctx->state == PRE_HEAD) || (ctx->state == PARSE_HEAD)) {
> int do_cleanup = 0;
>@@ -2941,6 +2941,22 @@
>
> ctx->state = PRE_HEAD;
> }
>+ }
>+
>+ /* We have nothing more to send, stop now. */
>+ if (APR_BUCKET_IS_EOS(dptr)) {
>+ /* Ensure that EOS is sent along. */
>+ APR_BUCKET_REMOVE(dptr);
>+ APR_BUCKET_INSERT_TAIL(*bb, dptr);
>
Thanks, this patch seems to fix the problem. One question,
though: why remove the bucket and then insert it at the end
of the same brigade? Is that just in case there's an EOS in
the middle of the brigade?
--Brian
