[Bug c/109530] New: Warning "is used uninitialized" raised for an initialized variable.

[Bug c/109530] New: Warning "is used uninitialized" raised for an initialized variable.

[patrick.pelissier at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109530

            Bug ID: 109530
           Summary: Warning "is used uninitialized" raised for an
                    initialized variable.
           Product: gcc
           Version: 12.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: patrick.pelissier at gmail dot com
  Target Milestone: ---

For the following program:

#include <stdio.h>
#include <setjmp.h>

jmp_buf buf;
jmp_buf buf1;

typedef struct ry_s {
  struct ry_s *next;
} ry_t[1];

struct ry_s *m_global;

static inline void
ry_init(ry_t state)
{
  m_global = state;
}

static inline void
ry_clear(ry_t state)
{
  m_global = state->next;
}

int func2(void)
{
  for(_Bool b = 1 ; b ; b = 0)
    for(ry_t test1 ; b ; ry_clear(test1), b = 0 )
      if (ry_init(test1), setjmp (buf) == 0)
        {
          FILE *f = fopen("tmp","wt");
          if ((setjmp (buf1) == 0) )
            {
              fprintf(f, "This is a text\n");
              fclose(f);
            }
        }
  return 0;
}


When running with "-O2  -Wall -Wextra", I get the following warning:

<source>: In function 'func2':
<source>:31:17: warning: 'f' is used uninitialized [-Wuninitialized]
   31 |           FILE *f = fopen("tmp","wt");

whereas f is obviously initialized in the statement.
I am confused by this warning.

[Bug c/109530] Warning "is used uninitialized" raised for an initialized variable.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109530

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The warning is confusing but it is partly correct.
Across a setjmp, you might need volatile variables.
Gcc does not know that fprintf nor fclose will call longjmp.

[Bug middle-end/109530] Warning "is used uninitialized" raised for an initialized variable.

[schwab@linux-m68k.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109530

--- Comment #2 from Andreas Schwab <schwab@linux-m68k.org> ---
It also doesn't see that setjmp()==0 means that the block is not reached a
second time.

[Bug middle-end/109530] Warning "is used uninitialized" raised for an initialized variable.

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109530

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-04-17
             Blocks|                            |24639
           Keywords|                            |missed-optimization
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
                 CC|                            |rguenth at gcc dot gnu.org

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Well, it actually diagnoses

<bb 2> [local count: 1073741824]:
f_4(ab) = f_9(D);
...

:/

So this is really an artifact of imperfect SSA form where we fail to
prune the PHI for 'f' before the first setjmp (into-SSA rewrite doesn't
honor scopes).

It's ethread that produces this copy by threading the loop header checks
and turning their PHIs into copies.

Btw, uninit diagnostics rely on the default defs eventually appearing in
PHIs for the may uninit case so we cannot simply skip such copies.

SSA_NAME_OCCURS_IN_ABNORMAL_PHI is also quite sticky, it's there from
original SSA and not removed after inlining ry_init.  When removing
an abnormal edge we'd have to check all other uses of destination PHI
arguments to validate clearing the flag.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=24639
[Bug 24639] [meta-bug] bug to track all Wuninitialized issues