[Bug c/102539] New: [11/12 regression] -Wmaybe-uninitialized false positive, invalid <unknown> location

[Bug c/102539] New: [11/12 regression] -Wmaybe-uninitialized false positive, invalid <unknown> location

[slyfox at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102539

            Bug ID: 102539
           Summary: [11/12 regression] -Wmaybe-uninitialized false
                    positive, invalid <unknown> location
           Product: gcc
           Version: 11.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: slyfox at gcc dot gnu.org
  Target Milestone: ---

Minimal example is extracted from aerospike-server-4.2.0.4:

$ cat a.c.c
void msg_set_buf(const int *);
void clustering_hb_event_listener_buffer_size(void) {
  unsigned long succession_buffer = (unsigned long)__builtin_alloca(42);
  msg_set_buf((const int*)succession_buffer);
}

$ LANG=C /nix/store/...-gcc-wrapper-12.0.0/bin/gcc -fno-strict-aliasing -Wall
-std=gnu99 -o a.o -c a.c.c -O2
a.c.c: In function 'clustering_hb_event_listener_buffer_size':
a.c.c:4:3: warning: '<unknown>' may be used uninitialized
[-Wmaybe-uninitialized]
    4 |   msg_set_buf((const int*)succession_buffer);
      |   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
a.c.c:1:6: note: by argument 1 of type 'const int *' to 'msg_set_buf' declared
here
    1 | void msg_set_buf(const int *);
      |      ^~~~~~~~~~~

There are two issues here (both are probably related):
1. '<unknown>' location is reported instead of some form of 'succession_buffer'
reference
2. there should be no warning about uninitialized data, it's the same pointer
passed around

$ LANG=C /nix/store/...-gcc-wrapper-12.0.0/bin/gcc -v
Using built-in specs.
COLLECT_GCC=/nix/store/...-gcc-12.0.0/bin/gcc
COLLECT_LTO_WRAPPER=/nix/store/...-gcc-12.0.0/libexec/gcc/x86_64-unknown-linux-gnu/12.0.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with:
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 12.0.0 20210919 (experimental) (GCC)

Affected compilers:
- 12.0.0 20210919 snapshot
- 11.1.0 release

Unaffected compilers:
- 10.3.0 release

[Bug c/102539] [11/12 regression] -Wmaybe-uninitialized false positive, invalid <unknown> location

[slyfox at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102539

--- Comment #1 from Sergei Trofimovich <slyfox at gcc dot gnu.org> ---
Created attachment 51524
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51524&action=edit
a.c.c.orig

a.c.c.orig is the unreduced preprocessed file with the same warning

[Bug tree-optimization/102539] [11/12 regression] -Wmaybe-uninitialized false positive, invalid <unknown> location

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102539

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Target Milestone|---                         |11.3
   Last reconfirmed|                            |2021-09-30
             Status|UNCONFIRMED                 |NEW
          Component|c                           |tree-optimization
           Keywords|                            |diagnostic

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The reduced testcase does not match at all the original testcase really.

Here is one which closer represents the original testcase:
void msg_set_buf(const int *);
void clustering_hb_event_listener_buffer_size(int a) {
  void* succession_buffer = a>16? __builtin_malloc (a) : __builtin_alloca(a);
  msg_set_buf((const int*)succession_buffer);
  if (a>16) __builtin_free (succession_buffer);
}

[Bug tree-optimization/102539] [11/12 regression] -Wmaybe-uninitialized false positive, invalid <unknown> location

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102539

Martin Sebor <msebor at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |INVALID
                 CC|                            |msebor at gcc dot gnu.org
             Status|NEW                         |RESOLVED

--- Comment #3 from Martin Sebor <msebor at gcc dot gnu.org> ---
-Wmaybe-uninitialized was enhanced to warn when a pointer to uninitialized
memory is passed to a function taking a const-qualified pointer:

  ... In addition, passing a pointer (or in C++, a reference) to an
uninitialized object to a const-qualified function argument is also diagnosed
by this warning.

That's also what seems to happen in the test case.  I get slightly different
output but the essence of the warning seems the same:

$ gcc -O2 -S -Wall -fdump-tree-uninit -fno-strict-aliasing pr102539.c
fabric/clustering.c: In function ‘msg_succession_list_field_set.constprop’:
fabric/clustering.c:3750:2: warning: ‘<unknown>’ may be used uninitialized
[-Wmaybe-uninitialized]
In file included from fabric/clustering.c:36:
/home/slyfox/dev/git/nixpkgs/source/cf/include/msg.h:195:6: note: by argument 3
of type ‘const uint8_t *’ {aka ‘const unsigned char *’} to ‘msg_set_buf’
declared here
tmp$ /build/gcc-master/gcc/xgcc -B /build/gcc-master/gcc -O2 -S -Wall
-fdump-tree-uninit -fno-strict-aliasing -std=gnu99 pr102539.c
fabric/clustering.c: In function ‘msg_succession_list_field_set.constprop’:
fabric/clustering.c:3750:2: warning: ‘<unknown>’ may be used uninitialized
[-Wmaybe-uninitialized]
In file included from fabric/clustering.c:36:
/home/slyfox/dev/git/nixpkgs/source/cf/include/msg.h:195:6: note: by argument 3
of type ‘const uint8_t *’ {aka ‘const unsigned char *’} to ‘msg_set_buf’
declared here


The <unknown> is a known cosmetic flaw with some middle end warnings.  It
should be fixed but we don't need to keep track of it here.

[Bug tree-optimization/102539] [11/12 regression] -Wmaybe-uninitialized false positive, invalid <unknown> location

[slyfox at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102539

--- Comment #4 from Sergei Trofimovich <slyfox at gcc dot gnu.org> ---
Ah, that's very clever! Thank you!