Re: [PATCH 0/5] Remove a few hurdles of compiling with clang

[Simon Marchi <simon dot marchi at polymtl dot ca>]

On 2017-06-13 21:22, Eli Zaretskii wrote:
> See, I don't consider the proposed solution to be elegant, because it
> tweaks a perfectly valid code to placate a stupid compiler warning.

It replaces perfectly valid code with some other perfectly valid code.  
That the compiler warning is stupid is your opinion.  I think it's 
useful (even though it may trigger false positives sometimes).

> Someone at some point might rightfully ask why didn't we use
> ALL_DEBUG_ADDRESS_REGISTERS instead, and might even reinstate the code
> you are about to change. So I prefer to have a seemingly "ugly"
> workaround, which nonetheless points out exactly which warning of what
> compiler caused it.  That way, at some future point, when clang
> hopefully gets its act together, we could revisit the issue and see
> that the workaround is no longer needed.

We can always add comments like

  /* Don't use ALL_DEBUG_ADDRESS_REGISTERS here to silence Clang's 
-Wfor-loop-analysis warning.  */

like we have right now for GCC warnings:

breakpoint.c:14747:      /* Initialize it just to avoid a GCC false 
warning.  */

Even though I have a slight preference for not silencing warnings when 
possible, it's really not a strong opinion, I would also be fine with 
the #pragma if that's what people prefer.

Simon