Re: [Intel-gfx] [PATCH i-g-t 3/6] igt/core: Fix build warning

[Tvrtko Ursulin]

On 19/11/2021 13:53, Petri Latvala wrote:
On Fri, Nov 19, 2021 at 12:59:42PM +0000, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin <tvrtko.ursu...@intel.com>

.../lib/igt_thread.c: In function ‘__igt_unique____igt_constructor_l66’:
.../lib/igt_thread.c:68:9: warning: ‘pthread_setspecific’ expecting 1 byte in a 
region of size 0 [-Wstringop-overread]
    68 |         pthread_setspecific(__igt_is_main_thread, (void*) 0x1);
       |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Tvrtko Ursulin <tvrtko.ursu...@intel.com>
Cc: Petri Latvala <petri.latv...@intel.com>
---
  lib/igt_core.c   | 6 ++++--
  lib/igt_thread.c | 4 +++-
  2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/lib/igt_core.c b/lib/igt_core.c
index ec05535cd56e..acb9743c4a24 100644
--- a/lib/igt_core.c
+++ b/lib/igt_core.c
@@ -2752,6 +2752,8 @@ void igt_vlog(const char *domain, enum igt_log_level 
level, const char *format,
        char *line, *formatted_line;
        char *thread_id;
        const char *program_name;
+       const uintptr_t false_key = 0;
+       const uintptr_t true_key = 1;
        const char *igt_log_level_str[] = {
                "DEBUG",
                "INFO",
@@ -2796,9 +2798,9 @@ void igt_vlog(const char *domain, enum igt_log_level 
level, const char *format,
        }
  
  	if (line[strlen(line) - 1] == '\n')
-               pthread_setspecific(__vlog_line_continuation, (void*) false);
+               pthread_setspecific(__vlog_line_continuation, &false_key);
        else
-               pthread_setspecific(__vlog_line_continuation, (void*) true);
+               pthread_setspecific(__vlog_line_continuation, &true_key);

Quoting the usage of this:
         if (pthread_getspecific(__vlog_line_continuation)) {

That's going to give a non-null pointer in both cases.

Doh..




  
  	/* append log buffer */
        _igt_log_buffer_append(formatted_line);
diff --git a/lib/igt_thread.c b/lib/igt_thread.c
index 5bdda4102def..f5de2d57eaaa 100644
--- a/lib/igt_thread.c
+++ b/lib/igt_thread.c
@@ -64,6 +64,8 @@ bool igt_thread_is_main(void)
  }
  
  igt_constructor {
+       const uintptr_t key = 1;
+
        pthread_key_create(&__igt_is_main_thread, NULL);
-       pthread_setspecific(__igt_is_main_thread, (void*) 0x1);
+       pthread_setspecific(__igt_is_main_thread, &key);

This is fine, __igt_is_main_thread key will have non-null only on the
main thread. But still a bit sus slapping the address of a
function-local variable to setspecific, we might just be trading a
compiler warning for a new future one.

And this is then the same - why do you say this one is okay? :)

Okay I wasn't sufficiently focused while trying to fix this. No idea 
then apart for playing with pragmas.

Regards,

Tvrtko