[COMMITTED] Do not pass NULL to memset in ssa_global_cache.

[Martin Liška]

On 11/14/21 14:15, Aldy Hernandez wrote:
> The code computing ranges in PHIs in the path solver reuses the
> temporary ssa_global_cache by calling its clear method.  Calling it on
> an empty cache causes us to call memset with NULL.
> 
> [The testcase doesn't fail without the patch.  I suppose it needs some
> usbsan magic, or to live somewhere else?]

Well, the actual test-case is the compiler itself as the source code.
Anyway, the UBSAN error happens for thousands of test-cases when run
during bootstrap-ubsan.mk config file. That said, I would remove
the added test-case.

Cheers,
Martin

> 
> Tested on x86-64 Linux.
> 
> gcc/ChangeLog:
> 
> 	PR tree-optimization/103229
> 	* gimple-range-cache.cc (ssa_global_cache::clear): Do not pass
> 	null value to memset.
> 
> gcc/testsuite/ChangeLog:
> 
> 	* gcc.dg/pr103229.c: New test.
> ---
>   gcc/gimple-range-cache.cc       |  3 ++-
>   gcc/testsuite/gcc.dg/pr103229.c | 10 ++++++++++
>   2 files changed, 12 insertions(+), 1 deletion(-)
>   create mode 100644 gcc/testsuite/gcc.dg/pr103229.c
> 
> diff --git a/gcc/gimple-range-cache.cc b/gcc/gimple-range-cache.cc
> index a63e20e7e49..b347edeb474 100644
> --- a/gcc/gimple-range-cache.cc
> +++ b/gcc/gimple-range-cache.cc
> @@ -651,7 +651,8 @@ ssa_global_cache::clear_global_range (tree name)
>   void
>   ssa_global_cache::clear ()
>   {
> -  memset (m_tab.address(), 0, m_tab.length () * sizeof (irange *));
> +  if (m_tab.address ())
> +    memset (m_tab.address(), 0, m_tab.length () * sizeof (irange *));
>   }
>   
>   // Dump the contents of the global cache to F.
> diff --git a/gcc/testsuite/gcc.dg/pr103229.c b/gcc/testsuite/gcc.dg/pr103229.c
> new file mode 100644
> index 00000000000..96ef9aff67c
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/pr103229.c
> @@ -0,0 +1,10 @@
> +// { dg-do compile }
> +// { dg-options "-O -w" }
> +
> +int main() {
> +  int i;
> +  for (; i;)
> +    ;
> +
> +  return 0;
> +}
>