RE: [PATCH] Modify tests to work with clang

[Taylor Simpson]

Without volatile, clang turns that statement into a call to abort.

It is discussed here (Chris Lattner's blog from 2011)
http://blog.llvm.org/2011/05/what-every-c-programmer-should-know.html

Here is the relevant paragraph
If you're using an LLVM-based compiler, you can dereference a "volatile" null 
pointer to get a crash if that's what you're looking for, since volatile loads 
and stores are generally not touched by the optimizer. There is currently no 
flag that enables random NULL pointer loads to be treated as valid accesses or 
to make random loads know that their pointer is "allowed to be null".

Taylor


-----Original Message-----
From: Alex Bennée <address@hidden>
Sent: Friday, November 15, 2019 6:17 AM
To: Taylor Simpson <address@hidden>
Cc: address@hidden
Subject: Re: [PATCH] Modify tests to work with clang

-------------------------------------------------------------------------
CAUTION: This email originated from outside of the organization.
-------------------------------------------------------------------------

Taylor Simpson <address@hidden> writes:

> Signed-off-by: Taylor Simpson <address@hidden>
> ---
>  tests/tcg/multiarch/float_helpers.c | 13 ++++++++++++-
>  tests/tcg/multiarch/linux-test.c    |  2 +-
>  2 files changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/tests/tcg/multiarch/float_helpers.c
> b/tests/tcg/multiarch/float_helpers.c
> index 8ee7903..437247c 100644
> --- a/tests/tcg/multiarch/float_helpers.c
> +++ b/tests/tcg/multiarch/float_helpers.c
<snip>
> diff --git a/tests/tcg/multiarch/linux-test.c
> b/tests/tcg/multiarch/linux-test.c
> index 673d7c8..edfc02c 100644
> --- a/tests/tcg/multiarch/linux-test.c
> +++ b/tests/tcg/multiarch/linux-test.c
> @@ -485,7 +485,7 @@ static void test_signal(void)
>      act.sa_flags = SA_SIGINFO;
>      chk_error(sigaction(SIGSEGV, &act, NULL));
>      if (setjmp(jmp_env) == 0) {
> -        *(uint8_t *)0 = 0;
> +        *(volatile uint8_t *)0 = 0;

This looks suspicious - volatile is almost never the answer to a question. What 
are we trying to achieve here?

>      }
>
>      act.sa_handler = SIG_DFL;


--
Alex Bennée