Daniel Veillard wrote:
Windows is escaping all paths in the path_processing label from xmlCanonicPath, so test fails to match the origional unescaped path in uripOpen. Attached is a patch for xmlCanonicPath where if the uri structure created at the path_processing label does not contain a scheme, i assume this to be a local path, it returns the original path rather than the path from saveURI.On Tue, Jan 31, 2006 at 01:36:02PM -0800, Manish Marathe wrote:Libxml experts everywhere,Of all tests the URL passing test is failing: "failed the URL passing test for /path/to/a b.html## Schemas regression tests" , I just wanted to know if this is known to fail.On linux all runtest passes without error. So I don't know what this is about really.
Did not want to commit this without some review as I have no idea if this might break anything out there. Due to the custom pathing and I/O handling in PHP it has no effect on my regression tests there so not a good test case.
I had asked a similar question a couple of weeks ago, so you might want to check earlier messages for the discussion on that topic.Also on Linux variants there are many tests that run, do Windows have those many? or runtest.exe and runsuite.exe are the one's.
Rob
Index: uri.c
===================================================================
RCS file: /cvs/gnome/gnome-xml/uri.c,v
retrieving revision 1.76
diff -c -r1.76 uri.c
*** uri.c 15 Sep 2005 14:15:20 -0000 1.76
--- uri.c 1 Feb 2006 11:15:52 -0000
***************
*** 2419,2425 ****
xmlFreeURI(uri);
return(NULL);
}
! ret = xmlSaveUri(uri);
xmlFreeURI(uri);
#else
ret = xmlStrdup((const xmlChar *) path);
--- 2419,2431 ----
xmlFreeURI(uri);
return(NULL);
}
!
! if (uri->scheme == NULL) {
! ret = xmlStrdup((const xmlChar *) path);
! } else {
! ret = xmlSaveUri(uri);
! }
!
xmlFreeURI(uri);
#else
ret = xmlStrdup((const xmlChar *) path);