You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Dean Gaudet <dg...@arctic.org> on 1997/03/31 05:43:26 UTC

[PATCH] mod_negotiation: not copying headers, and more memory fixes

Petr's patch to merge r->notes, r->headers_out and r->err_headers_out
didn't call overlay_tables properly.  This is the only bug fix in this
patch, it's the last hunk.  I think this might actually fix a PR but I
don't have the time to find it. 

The other hunks are more destroy_sub_req()s that are similar to the other
cleanups that Petr's patch performed.  But they're not strictly necessary,
just nice to prevent bloat during a request.

Dean

Index: mod_negotiation.c
===================================================================
RCS file: /hot/repository/apache/src/mod_negotiation.c,v
retrieving revision 1.9
diff -c -3 -r1.9 mod_negotiation.c
*** mod_negotiation.c	1997/03/18 06:40:07	1.9
--- mod_negotiation.c	1997/03/31 03:07:22
***************
*** 764,770 ****
  	if (sub_req->handler && !sub_req->content_type)
  	  sub_req->content_type = CGI_MAGIC_TYPE;
  
! 	if (sub_req->status != HTTP_OK || !sub_req->content_type) continue;
  	
  	/* If it's a map file, we use that instead of the map
  	 * we're building...
--- 764,773 ----
  	if (sub_req->handler && !sub_req->content_type)
  	  sub_req->content_type = CGI_MAGIC_TYPE;
  
! 	if (sub_req->status != HTTP_OK || !sub_req->content_type) {
! 	    destroy_sub_req (sub_req);
! 	    continue;
! 	}
  	
  	/* If it's a map file, we use that instead of the map
  	 * we're building...
***************
*** 1791,1799 ****
      char *sub_vary;
  
      if (!variant->sub_req) {
          sub_req = sub_req_lookup_file(variant->file_name, r);
!         if (sub_req->status != HTTP_OK && sub_req->status != HTTP_MULTIPLE_CHOICES)
!             return sub_req->status;
          variant->sub_req = sub_req;
      }
      else 
--- 1794,1807 ----
      char *sub_vary;
  
      if (!variant->sub_req) {
+ 	int status;
+ 
          sub_req = sub_req_lookup_file(variant->file_name, r);
! 	status = sub_req->status;
!         if (status != HTTP_OK && status != HTTP_MULTIPLE_CHOICES) {
! 	    destroy_sub_req (sub_req);
!             return status;
! 	}
          variant->sub_req = sub_req;
      }
      else 
***************
*** 1955,1960 ****
--- 1963,1969 ----
          sub_req = sub_req_lookup_file (best->file_name, r);
          if (sub_req->status != HTTP_OK) {
             res = sub_req->status;
+ 	   destroy_sub_req (sub_req);
             goto return_from_multi;
          }
      }
***************
*** 1982,1990 ****
      r->content_language = sub_req->content_language;
      r->finfo = sub_req->finfo;
      /* copy output headers from subrequest, but leave negotiation headers */
!     overlay_tables(r->pool, sub_req->notes, r->notes);
!     overlay_tables(r->pool, sub_req->headers_out, r->headers_out);
!     overlay_tables(r->pool, sub_req->err_headers_out, r->err_headers_out);
      avail_recs = (var_rec *)neg->avail_vars->elts;
      for (j = 0; j < neg->avail_vars->nelts; ++j) {
          var_rec *variant = &avail_recs[j];
--- 1991,1999 ----
      r->content_language = sub_req->content_language;
      r->finfo = sub_req->finfo;
      /* copy output headers from subrequest, but leave negotiation headers */
!     r->notes = overlay_tables(r->pool, sub_req->notes, r->notes);
!     r->headers_out = overlay_tables(r->pool, sub_req->headers_out, r->headers_out);
!     r->err_headers_out = overlay_tables(r->pool, sub_req->err_headers_out, r->err_headers_out);
      avail_recs = (var_rec *)neg->avail_vars->elts;
      for (j = 0; j < neg->avail_vars->nelts; ++j) {
          var_rec *variant = &avail_recs[j];


Re: [PATCH] mod_negotiation: not copying headers, and more memory fixes

Posted by Chuck Murcko <ch...@topsail.org>.
Dean Gaudet wrote:
> 
> Petr's patch to merge r->notes, r->headers_out and r->err_headers_out
> didn't call overlay_tables properly.  This is the only bug fix in this
> patch, it's the last hunk.  I think this might actually fix a PR but I
> don't have the time to find it.
> 
> The other hunks are more destroy_sub_req()s that are similar to the other
> cleanups that Petr's patch performed.  But they're not strictly necessary,
> just nice to prevent bloat during a request.
> 
+1. This is really weird code.
-- 
chuck
Chuck Murcko
The Topsail Group, West Chester PA USA
chuck@topsail.org