On Tue, May 28, 2024 at 3:53 PM Eduard Zingerman <eddyz87@xxxxxxxxx> wrote:
>
> [...]
>
> > > +/* Dumps C language definition or forward declaration for type **id**:
> > > + * - returns 1 if type is printable;
> > > + * - returns 0 if type is non-printable.
> >
> > does it also return <0 on error?
>
> Right
>
> >
> > > + */
> >
> > let's follow the format of doc comments, see other APIs. There is
> > @brief, @param, @return and so on.
>
> Will do
>
> > pw-bot: cr
> >
> >
> > > +LIBBPF_API int btf_dump__dump_one_type(struct btf_dump *d, __u32 id, bool fwd);
> >
> > not a fan of a name, how about we do `btf_dump__emit_type(struct
> > btf_dump *d, __u32 id, struct btf_dump_emit_type_opts *opts)` and have
> > forward declaration flag as options? We have
> > btf_dump__emit_type_decl(), this one could be called
> > btf_dump__emit_type_def() as well. WDYT?
>
> `btf_dump__emit_type_def` seems good and I can make it accept options
> with forward as a flag.
>
> However, in such a case the following is also a contender:
>
> struct btf_dump_type_opts {
> __u32 sz;
> bool skip_deps; /* flags picked so that by default */
> bool forward_only; /* the behavior matches non-opts variant */
> };
>
> LIBBPF_API int btf_dump__dump_type_opts(struct btf_dump *d, __u32 id,
> struct btf_dump_type_opts *opts);
>
>
> I find this contender more ugly but a bit more consistent.
> Wdyt?
You'll also need "skip_semicolon" which makes this even uglier. Which
is why I'd not do it as an extension to btf_dump__dump_type() API.
>
> [...]
