[patch] Don't return success and an empty brigade in core_input_filter's getline (trunk)

trunk-core_filter_getline_eof.patch (text/plain), 2.08 KB, created by Yann Ylavic on 2013-09-17 11:43:44 UTC

(hide)

Description: Don't return success and an empty brigade in core_input_filter's getline (trunk)

Filename:

MIME Type:

Creator: Yann Ylavic

Created: 2013-09-17 11:43:44 UTC

Size: 2.08 KB

patch

obsolete

>Index: include/http_core.h >=================================================================== >--- include/http_core.h (revision 1523591) >+++ include/http_core.h (working copy) >@@ -698,6 +698,14 @@ apr_status_t ap_core_input_filter(ap_filter_t *f, > ap_input_mode_t mode, apr_read_type_e block, > apr_off_t readbytes); > apr_status_t ap_core_output_filter(ap_filter_t *f, apr_bucket_brigade *b); >+/** >+ * When ap_core_input_filter with AP_MODE_GETLINE gets an immediate EOF from >+ * the socket, should it return APR_SUCCESS with an empty brigade (legacy) or >+ * return APR_EOF (default) ? >+ */ >+#ifndef AP_CORE_INPUT_FILTER_GETLINE_EOF >+#define AP_CORE_INPUT_FILTER_GETLINE_EOF 1 >+#endif > > > AP_DECLARE(const char*) ap_get_server_protocol(server_rec* s); >Index: server/core_filters.c >=================================================================== >--- server/core_filters.c (revision 1523591) >+++ server/core_filters.c (working copy) >@@ -152,6 +152,25 @@ apr_status_t ap_core_input_filter(ap_filter_t *f, > if (APR_STATUS_IS_EAGAIN(rv) && block == APR_NONBLOCK_READ) { > rv = APR_SUCCESS; > } >+#if AP_CORE_INPUT_FILTER_GETLINE_EOF >+ /* When the socket is immediatly EOF, apr_brigade_split_line returns >+ * APR_SUCCESS with an empty brigade (socket_bucket_read morphes the >+ * socket into an empty buffer which in turn is stripped by >+ * apr_brigade_split_line). >+ * >+ * In blocking mode we should never return APR_SUCCESS with an empty >+ * brigade, we are APR_EOF now and for the next calls. >+ * >+ * XXX: Note that we are EOF in any blocking mode (although callers >+ * are used to empty brigade in non-blocking mode), should that break >+ * some logic, one can "&& block == APR_BLOCK_READ" here, but if it's >+ * OK, let's avoid a next call... >+ */ >+ else if (rv == APR_SUCCESS && APR_BRIGADE_EMPTY(b) >+ && APR_BRIGADE_EMPTY(ctx->b)) { >+ rv = APR_EOF; >+ } >+#endif > return rv; > } > View Attachment As Diff View Attachment As Raw

This is ASF Bugzilla: the Apache Software Foundation bug system. In case of problems with the functioning of ASF Bugzilla, please contact bugzilla-admin@apache.org. Please Note: this e-mail address is only for reporting problems with ASF Bugzilla. Mail about any other subject will be silently ignored.

• • Home
  • | New
  • | Browse
  • | Search
  • |
    [?]
  • | Reports
  • | Help
  • | New Account
  • | Log In
    Remember [x]
  • | Forgot Password
    Login: [x]