[C++ Patch / RFC] PR 48420

[Paolo Carlini <paolo dot carlini at oracle dot com>]

Hi,

this issue seems pretty easy to deal with: submitter complains that we 
warn for

void foo(int* p);

void bar() {
  foo(false);
}

and we do *not* for:

void foo(int* p);

void bar() {
  const bool kDebugMode = false;
  foo(kDebugMode);
}

thus I tried using decl_constant_var_p / integral_constant_value and it 
worked fine, as expected. However - I may be wrong - but I'm not seeing 
*much* of this kind of idea around for diagnostic purposes, I'm afraid 
I'm missing something. It is actually Ok?

Anyway, testsuite passes on x86_64-linux (bootstrap pending, important 
because the W is on by default!)

Thanks,
Paolo.

//////////////////

Attachment: CL_48420
Description: Text document

Attachment: patch_48420
Description: Text document