On Thu, Mar 24, 2022 at 5:06 AM Alexandre Oliva via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
>
> The copies of identifiers, indended to associate hardening SSA
> temporaries to the original variables they refer to, end up causing
> -fcompare-debug to fail, because DECL_UIDs are not identical, and the
> nouid flag used in compare-debug dumps doesn't affect the uids in
> naked identifiers, so the divergence becomes apparent.
>
> This patch drops the naked identifiers. Though somewhat desirable,
> they're not necessary.
>
> Regstrapped on x86_64-linux-gnu. Ok to install?
OK.
>
> for gcc/ChangeLog
>
> PR debug/104564
> * gimple-harden-conditionals.cc (detach_value): Keep temps
> anonymous.
>
> for gcc/testsuite/ChangeLog
>
> PR debug/104564
> * c-c++-common/torture/harden-comp.c: Adjust.
> * c-c++-common/torture/harden-cond.c: Adjust.
> ---
> gcc/gimple-harden-conditionals.cc | 11 ++++-------
> gcc/testsuite/c-c++-common/torture/harden-comp.c | 2 +-
> gcc/testsuite/c-c++-common/torture/harden-cond.c | 2 +-
> 3 files changed, 6 insertions(+), 9 deletions(-)
>
> diff --git a/gcc/gimple-harden-conditionals.cc
> b/gcc/gimple-harden-conditionals.cc
> index be01f3ea8c44a..c7e5e077a74f6 100644
> --- a/gcc/gimple-harden-conditionals.cc
> +++ b/gcc/gimple-harden-conditionals.cc
> @@ -126,14 +126,11 @@ detach_value (location_t loc, gimple_stmt_iterator
> *gsip, tree val)
> return val;
> }
>
> - /* Create a SSA "copy" of VAL. This could be an anonymous
> - temporary, but it's nice to have it named after the corresponding
> - variable. Alas, when VAL is a DECL_BY_REFERENCE RESULT_DECL,
> - setting (a copy of) it would be flagged by checking, so we don't
> - use copy_ssa_name: we create an anonymous SSA name, and then give
> - it the same identifier (rather than decl) as VAL. */
> + /* Create a SSA "copy" of VAL. It would be nice to have it named
> + after the corresponding variable, but sharing the same decl is
> + problematic when VAL is a DECL_BY_REFERENCE RESULT_DECL, and
> + copying just the identifier hits -fcompare-debug failures. */
> tree ret = make_ssa_name (TREE_TYPE (val));
> - SET_SSA_NAME_VAR_OR_IDENTIFIER (ret, SSA_NAME_IDENTIFIER (val));
>
> /* Some modes won't fit in general regs, so we fall back to memory
> for them. ??? It would be ideal to try to identify an alternate,
> diff --git a/gcc/testsuite/c-c++-common/torture/harden-comp.c
> b/gcc/testsuite/c-c++-common/torture/harden-comp.c
> index 1ee0b3663443d..502f52e25be24 100644
> --- a/gcc/testsuite/c-c++-common/torture/harden-comp.c
> +++ b/gcc/testsuite/c-c++-common/torture/harden-comp.c
> @@ -11,4 +11,4 @@ f (int i, int j)
> /* { dg-final { scan-tree-dump-times "Adding reversed compare" 1 "hardcmp" }
> } */
> /* { dg-final { scan-tree-dump-times "__builtin_trap" 1 "hardcmp" } } */
> /* { dg-final { scan-tree-dump-times "_\[0-9\]* = i_\[0-9\]*\[(\]D\[)\] <
> j_\[0-9\]*\[(\]D\[)\];" 1 "hardcmp" } } */
> -/* { dg-final { scan-tree-dump-times "_\[0-9\]* = i_\[0-9\]* >= j_\[0-9\]*;"
> 1 "hardcmp" } } */
> +/* { dg-final { scan-tree-dump-times "_\[0-9\]* = _\[0-9\]* >= _\[0-9\]*;" 1
> "hardcmp" } } */
> diff --git a/gcc/testsuite/c-c++-common/torture/harden-cond.c
> b/gcc/testsuite/c-c++-common/torture/harden-cond.c
> index 86de8e155ed38..213b048b25af5 100644
> --- a/gcc/testsuite/c-c++-common/torture/harden-cond.c
> +++ b/gcc/testsuite/c-c++-common/torture/harden-cond.c
> @@ -15,4 +15,4 @@ f (int i, int j)
> /* { dg-final { scan-tree-dump-times "Adding reversed compare" 2 "hardcbr" }
> } */
> /* { dg-final { scan-tree-dump-times "__builtin_trap" 2 "hardcbr" } } */
> /* { dg-final { scan-tree-dump-times "if \[(\]i_\[0-9\]*\[(\]D\[)\] <
> j_\[0-9\]*\[(\]D\[)\]\[)\]" 1 "hardcbr" } } */
> -/* { dg-final { scan-tree-dump-times "if \[(\]i_\[0-9\]* >= j_\[0-9\]*\[)\]"
> 2 "hardcbr" } } */
> +/* { dg-final { scan-tree-dump-times "if \[(\]_\[0-9\]* >= _\[0-9\]*\[)\]" 2
> "hardcbr" } } */
>
>
> --
> Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
> Free Software Activist GNU Toolchain Engineer
> Disinformation flourishes because many people care deeply about injustice
> but very few check the facts. Ask me about <https://stallmansupport.org>
