On Wednesday 19 January 2011 09:46 PM, C. Michael Pilato wrote:
On 01/19/2011 10:52 AM, Arwin Arni wrote:
Hi All,
Authz section names(and paths too) are parsed/loaded in a case insensitive way.
So they need to be compared in a case-insensitive way.
Following functions do authz,
1. libsvn_repos/authz.c:authz_get_path_access()
2. libsvn_repos/authz.c:authz_get_any_access() is called when the "path" is
NULL (for MKACTIVITY, DELETE).
'1' is leaving it to svn_config_enumerate2() to handle case (in)sensitiveness.
'2' is explicitly comparing each config item in a case sensitive way.
My patch is just fixing the '2' to check it in a case insensitive way.
So, just to be clear, now read and write access checks will both compare the
repository name in a case-insensitive way?
That's Correct.
Index: subversion/libsvn_repos/authz.c
===================================================================
--- subversion/libsvn_repos/authz.c (revision 1060836)
+++ subversion/libsvn_repos/authz.c (working copy)
@@ -398,7 +398,7 @@
/* Does the section apply to the query? */
if (section_name[0] == '/'
- || strncmp(section_name, b->repos_path,
+ || strncasecmp(section_name, b->repos_path,
strlen(b->repos_path)) == 0)
You've goofed up the indentation here.
Again.. extremely sorry.. I guess I'm still not completely accustomed to
the indentation conventions.. Here's a patch with the indentation fixed..
Regards,
Arwin Arni
Index: subversion/libsvn_repos/authz.c
===================================================================
--- subversion/libsvn_repos/authz.c (revision 1060836)
+++ subversion/libsvn_repos/authz.c (working copy)
@@ -398,8 +398,8 @@
/* Does the section apply to the query? */
if (section_name[0] == '/'
- || strncmp(section_name, b->repos_path,
- strlen(b->repos_path)) == 0)
+ || strncasecmp(section_name, b->repos_path,
+ strlen(b->repos_path)) == 0)
{
b->allow = b->deny = svn_authz_none;
Fix for Issue #3781 repo prefix rules in authz section is checked case
sensitively for write operations
* subversion/libsvn_repos/authz.c
(authz_get_any_access_parser_cb): Use strncasecmp() instead of strncmp()
Patch by: kameshj
Arwin Arni <arwin{_AT_}collab.net>
