On 07/29/2007 10:35 AM, wrote: > Author: niq > Date: Sun Jul 29 01:35:56 2007 > New Revision: 560689 > > URL: http://svn.apache.org/viewvc?view=rev&rev=560689 > Log: > Fix protocol handling in mod_deflate input filter > PR 23287 > > Modified: > httpd/httpd/trunk/CHANGES > httpd/httpd/trunk/modules/filters/mod_deflate.c >
> Modified: httpd/httpd/trunk/modules/filters/mod_deflate.c > URL: > http://svn.apache.org/viewvc/httpd/httpd/trunk/modules/filters/mod_deflate.c?view=diff&rev=560689&r1=560688&r2=560689 > ============================================================================== > --- httpd/httpd/trunk/modules/filters/mod_deflate.c (original) > +++ httpd/httpd/trunk/modules/filters/mod_deflate.c Sun Jul 29 01:35:56 2007 > @@ -665,22 +665,52 @@ > return ap_get_brigade(f->next, bb, mode, block, readbytes); > } > > - /* Let's see what our current Content-Encoding is. > - * If gzip is present, don't gzip again. (We could, but let's not.) > + /* Check whether request body is gzipped. > + * > + * If it is, we're transforming the contents, invalidating > + * some request headers including Content-Encoding. > + * > + * If not, we just remove ourself. > */ > encoding = apr_table_get(r->headers_in, "Content-Encoding"); > - if (encoding) { > + if (encoding && *encoding) { > const char *tmp = encoding; > > - token = ap_get_token(r->pool, &tmp, 0); > - while (token && token[0]) { > - if (!strcasecmp(token, "gzip")) { > - found = 1; > - break; > - } > - /* Otherwise, skip token */ > - tmp++; > + /* check the usual/simple case first */ > + if (!strcasecmp(encoding, "gzip")) { > + found = 1; > + apr_table_unset(r->headers_in, "Content-Encoding"); > + } > + else { > token = ap_get_token(r->pool, &tmp, 0); > + while (token && token[0]) { > + if (!strcasecmp(token, "gzip")) { > + char *new_encoding = apr_pstrdup(r->pool, encoding); > + char *ptr = ap_strstr(new_encoding, token); > + size_t sz = 5*sizeof(char); > + if (ptr == new_encoding) { > + /* remove "gzip:" at start */ > + memmove(ptr, ptr+sz, sz); Maybe I am missing the point, but if gzip is at the start why moving memory and not simply increasing new_encoding by 5? > + } > + else { > + /* remove ":gzip" as found */ > + memmove(ptr-sizeof(char), ptr+4*sizeof(char), > sz); Is this correct? memmove maybe bcopy on platforms where no memmove is available and as I understand the man pages of memmove and bcopy it would only copy sz bytes from behind gzip to the old position of the gzip string. What if you have Content-Encoding: compress:gzip:deflate (OK, a silly case, but still valid). As far as I understand this would result in compress:defl:deflate And BTW, as the order of the encodings in Content-Encoding must be the order in which they have been applied, don't we need to ensure that gzip needs to be at the *end* of the content encoding string, as we have to decode things in the reverse order? Otherwise we would try to decode a gzip encoding when there is still another encoding on top of gzip. Regards Rüdiger
