[vlc-devel] [PATCH 2/7] core: utils: add vlc_qsort

[Hugo Beauzée-Luyssen]

On Tue, Jan 22, 2019, at 10:34 AM, Thomas Guillem wrote:
> On Mon, Jan 21, 2019, at 18:09, Hugo Beauzée-Luyssen wrote:
> > 
> > Any reason not to use "thread_local" ?
> 
> No, no idea, just copied the code from Rémi compat code.
> 

Since _Thread_local might be missing on some platform, I'd be in favor of using thread_local instead

> > 
> > > +{
> > > +    int (*compar)(const void *, const void *, void *);
> > > +    void *arg;
> > > +} state;
> > > +
> > > +static int compar_wrapper(const void *a, const void *b)
> > > +{
> > > +    return state.compar(a, b, state.arg);
> > > +}
> > > +
> > > +/* Follow the upcoming POSIX prototype, coming from GNU/libc.
> > > + * Note that this differs from the BSD prototype. */
> > > +
> > > +VLC_WEAK void vlc_qsort(void *base, size_t nmemb, size_t size,
> > > +                        int (*compar)(const void *, const void *, void *),
> > > +                        void *arg)
> > 
> > Any particular reason to use VLC_WEAK instead of the usual per-platform 
> > implementation?
> 
> It's more future proof for me. We could add more utility functions to 
> misc/utils.c. Some of these functions could be implemented differently 
> depending of the OS or not. With VLC_WEAK, we don't have to play with 
> conditions in the Makefile.am, this seems more convenient.
> 

Fair enough, the question is mostly that we don't use it much but more often use multiple source files. Your point is very valid IMO, I just find weak symbols harder to reason about, or less intuitive, than a plain different source file, but I'm absolutely fine with your solution :)

> As you want, I don't have strong opinions about wether we should use 
> VLC_WEAK in that case.
> 


-- 
  Hugo Beauzée-Luyssen
  hugo at beauzee.fr