Re: [PATCH 1/1] global: provide memrchr implementation for macOS

[Markus Mayer <code@mmayer.net>]

Hi Alex,

> This code seems overly complex to me.  A do-while seems unintuitive when
> you're not performing at least 1 iteration, and having an if that is
> basically the same operation as the while condition seems weird.  Maybe
> I'm missing something, but it looks like it can be written more readable as:

I have no objections to changing the code. For my initial submission,
I figured I'd use an implementation that comes directly from Apple's
own open source server. But if the consensus is to tweak the code to
make it smaller / faster / more readable, that's alright by me. I only
want CGIT to build on macOS. :-)

> void *
> memrchr(const void *s, int c, size_t n)
> {
>         unsigned char        ch;
>         const unsigned char  *cp;
>
>         ch = c;
>
>         cp = s;
>         for (cp += n; n != 0; n--) {
>                 cp--;
>                 if (*cp == ch)
>                         return (void *) cp;
>         }
>
>         return NULL;
> }

Let me play around with this.

> Clang produces similar assembly code for both.  GCC however produces
> smaller code for the for loop.

Thanks for the feedback.

Regards,
-Markus