On 02/20/2012 03:15 AM, Daniel Veillard wrote:
On Mon, Feb 20, 2012 at 09:59:23AM +0100, Jiri Denemark wrote:
> On Mon, Feb 20, 2012 at 11:26:50 +0800, Daniel Veillard wrote:
>> diff --git a/src/util/virfile.h b/src/util/virfile.h
>> index ec1e90b..184677c 100644
>> --- a/src/util/virfile.h
>> +++ b/src/util/virfile.h
>> @@ -58,10 +58,10 @@ typedef virFileWrapperFd *virFileWrapperFdPtr;
>>
>> int virFileDirectFdFlag(void);
>>
>> -enum {
>> +enum virFileWrapperFdFlags {
>> VIR_FILE_WRAPPER_BYPASS_CACHE = (1 << 0),
>> VIR_FILE_WRAPPER_NON_BLOCKING = (1 << 1),
>> -} virFileWrapperFdFlags;
>> +};
>
> Actually the error was missing typedef (again, shame on me) but this patch
> fixes it as well and since we use these flags OR-ed, it's unlikely we will
> ever need to use the type anywhere.
Well let's make the typedef explicit, that's cleaner, ACK in advance
if you want to make that patch :-)
Is this enough of a preferred style that we should go ahead and add a
note in HACKING about the style preference, as well as add a syntax
check and convert the offenders?
Overall, we currently favor raw enums over typedefs:
$ git grep '^enum' | wc
518 1226 25319
$ git grep '^typedef enum' | wc
140 488 6520
But those numbers are misleading; in all of our public headers:
$ git grep '^enum' include/ |wc
0 0 0
$ git grep '^typedef enum' include/ |wc
79 237 3475
I decided not to push a patch adding the typedef, after all, at least
not until we decide whether it makes a difference style wise.
--
Eric Blake eblake(a)redhat.com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org