On 06/08/2016 04:10 AM, Ján Tomko wrote:
On Wed, Jun 08, 2016 at 08:49:37AM +0200, Peter Krempa wrote:
> On Tue, Jun 07, 2016 at 20:07:28 +0200, Ján Tomko wrote:
>> Store all the data related to RNG validation in one structure to
>> allow splitting virXMLValidateAgainstSchema.
>> ---
>> src/util/virxml.c | 47 +++++++++++++++++++++++++++--------------------
>> src/util/virxml.h | 10 ++++++++++
>> 2 files changed, 37 insertions(+), 20 deletions(-)
> [...]
>
>> @@ -177,6 +179,14 @@ int virXMLInjectNamespace(xmlNodePtr node,
>> const char *uri,
>> const char *key);
>>
>> +typedef struct _virXMLValidator {
>> + xmlRelaxNGParserCtxtPtr rngParser;
>> + xmlRelaxNGPtr rng;
>> + xmlRelaxNGValidCtxtPtr rngValid;
>> + virBuffer buf;
>> + char *schemafile;
>> +} virXMLValidator, *virXMLValidatorPtr;
> This is too revolutionary. I'd stick with the few extra typedefs as we
> do (almost) everywhere.
The revolution has already been started by commit ff89e9b with a fitting
summary:
conf: move virDomainDeviceInfo definition from domain_conf.h to device_conf.h
The revolution started long before that (and it wasn't me who filed the
declaration, although I did fire the latest shot):
grep "^typedef struct.*{$" src/*/*.[ch]
I really see no point in the ancient notation (nor in the extra _virBlah
naming of the struct when it's never otherwise used (and could have just
as easily been named virBlah to match the typedef - those are different
namespaces), although I forgot to remove those when I moved
virDomainDeviceInfo). Maybe there was some compiler in days of yore that
required it?
> ACK with the above changed.
I have changed it to honour the tradition.
Death to the redundant oppressors!!