Re: [libvirt] [PATCH 2/2] xml: Surpress printing "domain.xml" for parse errors on XML strings
On 08/16/2011 06:49 AM, Peter Krempa wrote:
Patch a53ab1094e93f1b6d93ad9be63d8ccc5fd19a2a9 introduces printing
file name on XML errors. This corrects the URL string to be NULL and
therefrore to print an error message not containing bogus filename
s/therefrore/therefore/
"domain.xml".
---
src/conf/domain_conf.c | 2 +-
src/security/virt-aa-helper.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
+++ b/src/conf/domain_conf.c
@@ -7048,7 +7048,7 @@ virDomainDefParse(const char *xmlStr,
xmlDocPtr xml;
virDomainDefPtr def = NULL;
- if ((xml = virXMLParse(filename, xmlStr, "domain.xml"))) {
+ if ((xml = virXMLParse(filename, xmlStr, NULL))) {
I'm not sure I follow the difference between filename and url arguments.
In virXMLParseHelper, this is resulting in the difference between
xmlCtxtReadFile(pctxt, filename, NULL, flags) and xmlCtxtReadDoc(pctxt,
string, url, NULL, flags). If I'm understanding correctly, that means
that 'url' is treated as the filename owning the string being parsed
(that is, if you had manually read the file contents yourself and pass
the contents as a string, instead of caling xmlCtxtReadFile directly,
then you would use the filename of xmlCtxtReadFile as the url of
xmlCtxtReadDoc).
I guess passing NULL as the url is okay - it just means that any errors
reported by the parser are no longer tied to a specific filename (which
is okay, because we are intercepting the error reporting, and reporting
it on behalf of just the string line number anyway).
I wonder if this patch is complete, or if there are more instances where
we are passing a url to xml parse functions even though the string was
generated rather than being tied to a real file. tools/virsh.c in
particular may need some cleanups related to this, although I don't see
it using virXMLParse.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org