[Bug c/98900] New: Wrong static analyses warning when optimisation is on

[Bug c/98900] New: Wrong static analyses warning when optimisation is on

[oleg.pekar.2017 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98900

            Bug ID: 98900
           Summary: Wrong static analyses warning when optimisation is on
           Product: gcc
           Version: 8.3.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: oleg.pekar.2017 at gmail dot com
  Target Milestone: ---

Created attachment 50095
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=50095&action=edit
Preprocessor output for source file test1.c

gcc version: 8.3.1
system type: RedHat 8.1, Kernel 4.18.0-147.3.1.el8_1.x86_64

Problem: Wrong static analyses warning when optimisation is on.
Input source: attached test1.i
Description: in this code address of a variable local to a block is used
outside of the block. With no optimisation there's no warning while the code
still can cause a problem. But with optimisation static analyses provides a
confusing warning - it complains about a variable not used in the context of
the warning.

Command line with no optimisation: gcc -Wall -save-temps test1.c
Output: nothing

Command line with optimisation: gcc -Wall -save-temps -O2 test1.c
Output: 
test1.c: In function ‘main’:
test1.c:11:7: warning: ‘x’ is used uninitialized in this function
[-Wuninitialized]
     g = *p;
     ~~^~~~

[Bug middle-end/98900] misleading -Wuninitialized using a variable outside its lifetime with -O2

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98900

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Blocks|                            |24639
                 CC|                            |msebor at gcc dot gnu.org
          Component|c                           |middle-end
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
            Summary|Wrong static analyses       |misleading -Wuninitialized
                   |warning when optimisation   |using a variable outside
                   |is on                       |its lifetime with -O2
   Last reconfirmed|                            |2021-01-31
           Keywords|                            |diagnostic

--- Comment #1 from Martin Sebor <msebor at gcc dot gnu.org> ---
The warning could be clearer but at the point in the pipeline it's issued there
isn't enough information in the IL to differentiate the specific nature of the
problem (variable used after the end of its lifetime) from a plain
uninitialized read:

$ gcc -O1 -S -Wall -fdump-tree-uninit=/dev/stdout pr98850.c

;; Function main (main, funcdef_no=0, decl_uid=1943, cgraph_uid=1,
symbol_order=1) (executed once)

test1.c: In function ‘main’:
test1.c:11:7: warning: ‘x’ is used uninitialized [-Wuninitialized]
int main ()
{
  int x;

  <bb 2> [local count: 1073741824]:
  g = x_3(D);
  return 0;

}

But after changing the example to the one below doesn't make the warning more
informative even though the IL now has sufficient information (the CLOBBER) to
make it clear the variable isn't uninitialized at the time it's read but rather
its lifetime has ended.  So with that I agree that things can be improved.

$ cat pr98850.c && gcc -O1 -S -Wall -fdump-tree-uninit=/dev/stdout pr98850.c
int g = 10;

void f (int*);

int main()
{
    int* p = 0;
    {
        int x = 15;
        p = &x;
        f (p);
    }

    g = *p;
}

;; Function main (main, funcdef_no=0, decl_uid=1945, cgraph_uid=1,
symbol_order=1) (executed once)

pr98850.c: In function ‘main’:
pr98850.c:14:9: warning: ‘x’ is used uninitialized [-Wuninitialized]
   14 |     g = *p;
      |         ^~
pr98850.c:9:13: note: ‘x’ declared here
    9 |         int x = 15;
      |             ^
int main ()
{
  int x;
  int _1;

  <bb 2> [local count: 1073741824]:
  x = 15;
  f (&x);
  x ={v} {CLOBBER};
  _1 = x;
  g = _1;
  return 0;

}


Referenced Bugs:

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

[Bug middle-end/98900] misleading -Wuninitialized using a variable outside its lifetime with -O2

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98900

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=81524

--- Comment #2 from Martin Sebor <msebor at gcc dot gnu.org> ---
See also pr81524 for a superset of these problems.

[Bug middle-end/98900] misleading -Wuninitialized using a variable outside its lifetime with -O2

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98900

--- Comment #3 from Martin Sebor <msebor at gcc dot gnu.org> ---
GCC 12 issues -Wdangling-poinnter for this bug now at all optimization levels. 
It still issues -Wuninitialized at -O1 and above, but with the dangling pointer
warning I don't think it's a big deal.  It should of course still be possible
to suppress it.

$ cat pr98900.c && gcc -S -Wall pr98900.c
int g = 10;

int main()
{
    int* p = 0;
    {
        int x = 15;
        p = &x;
    }

    g = *p;

    return 0;
}
pr98900.c: In function ‘main’:
pr98900.c:11:9: warning: using dangling pointer ‘p’ to ‘x’
[-Wdangling-pointer=]
   11 |     g = *p;
      |         ^~
pr98900.c:7:13: note: ‘x’ declared here
    7 |         int x = 15;
      |             ^