Re: [libvirt] [PATCH 01/15] util: Introduce VIR_DEFINE_AUTOPTR_FUNC for virStorageAuthDef

On 2/7/19 11:54 AM, Andrea Bolognani wrote: > On Thu, 2019-02-07 at 17:36 +0100, Erik Skultety wrote: >> On Thu, Feb 07, 2019 at 05:11:24PM +0100, Andrea Bolognani wrote: >>> On Thu, 2019-02-07 at 16:55 +0100, Erik Skultety wrote: >>>> On Thu, Feb 07, 2019 at 04:24:13PM +0100, Andrea Bolognani wrote: >>>>> Please keep the semicolon! If a macro is used like a function, then >>>>> its call sites should also look like those of a function. >>>> >>>> Except VIR_DEFINE_AUTOPTR_FUNC macro is not used like a function, it defines a >>>> function. >>> >>> Sure, and the way you make that happen is by writing >>> >>> MACRO_NAME(argument_one, argument_two); >>> >>> How is that not using it like a function? :) >> >> I may need to replace my dictionary, because the way I understand the >> expression "like a function" is that the macro is called like function and it >> behaves like a function, i.e. returns a value, IOW by using the macro its >> expansion will perform the usual set of operation on the stack that a function >> call involves (push parameters, return address, jump into function, >> pop the return value...) > > I guess abort() is not a function either then, since it doesn't have > any parameters to push or values to return! :P > > Anyway, the point is that we have already started mandating the use > of semicolon after other macros that expand to definitions, such as > VIR_ENUM_DECL(), VIR_ENUM_IMPL(), and VIR_ONCE_GLOBAL_INIT(): we > should do the same for VIR_DEFINE_AUTOPTR_FUNC() and increase > consistency instead of pushing in the opposite direction. > Since the issue is consistency, how about a patch that adds the ; to the existing VIR_DEFINE_AUTOPTR_FUNC's that don't have it? Ironically, I found one that has it: VIR_DEFINE_AUTOPTR_FUNC(virSEVCapability, virSEVCapabilitiesFree);