[Bug gcov-profile/106090] New: [GCOV] Wrong coverage for loop statements

[Bug gcov-profile/106090] New: [GCOV] Wrong coverage for loop statements

[njuwy at smail dot nju.edu.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106090

            Bug ID: 106090
           Summary: [GCOV] Wrong coverage for loop statements
           Product: gcc
           Version: 10.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: gcov-profile
          Assignee: unassigned at gcc dot gnu.org
          Reporter: njuwy at smail dot nju.edu.cn
                CC: marxin at gcc dot gnu.org
  Target Milestone: ---

-:    0:Source:test.c
        -:    0:Graph:test.gcno
        -:    0:Data:test.gcda
        -:    0:Runs:1
        -:    1:volatile int cnt = 0;
        -:    2:
        -:    3:__attribute__((noinline, noclone)) static int
        5:    4:last (void)
        -:    5:{
        5:    6:  return ++cnt % 5 == 0;
        -:    7:}
        -:    8:
        -:    9:__attribute__((noinline, noclone)) static void
        6:   10:do_it (void)
        -:   11:{
        6:   12:  asm volatile ("" : : "r" (&cnt) : "memory");
        6:   13:}
        -:   14:
        1:   15:static void f1 (void)
        -:   16:{
        1:   17:  do_it();
        4:   18:  for (;; do_it())
        -:   19:    {
        5:   20:      if (last ())
        1:   21:        break;
        -:   22:    }
        1:   23:  do_it ();
        1:   24:}
        -:   25:
        -:   26:int
        1:   27:main ()
        -:   28:{
        1:   29:  f1 ();
        -:   30:}
Should the coverage of line 18 be 5?

[Bug gcov-profile/106090] [GCOV] Wrong coverage for loop statements

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106090

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |INVALID
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #1 from Martin Liška <marxin at gcc dot gnu.org> ---
No, it's correct as it only contains do_it() call that is triggered the same
number times as line 20 minus one.

[Bug gcov-profile/106090] [GCOV] Wrong coverage for loop statements

[njuwy at smail dot nju.edu.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106090

--- Comment #2 from Yang Wang <njuwy at smail dot nju.edu.cn> ---
(In reply to Martin Liška from comment #1)
> No, it's correct as it only contains do_it() call that is triggered the same
> number times as line 20 minus one.

(In reply to Martin Liška from comment #1)
> No, it's correct as it only contains do_it() call that is triggered the same
> number times as line 20 minus one.

        -:    0:Source:test.c
        -:    0:Graph:test.gcno
        -:    0:Data:test.gcda
        -:    0:Runs:1
        -:    1:volatile int cnt = 0;
        -:    2:
        -:    3:__attribute__((noinline, noclone)) static int
        5:    4:last (void)
        -:    5:{
        5:    6:  return ++cnt % 5 == 0;
        -:    7:}
        -:    8:
        -:    9:__attribute__((noinline, noclone)) static void
        6:   10:do_it (void)
        -:   11:{
        6:   12:  asm volatile ("" : : "r" (&cnt) : "memory");
        6:   13:}
        -:   14:
        1:   15:static void f1 (void)
        -:   16:{
        1:   17:  do_it();
        4:   18:  for (; 0<1; do_it())
        -:   19:    {
        5:   20:      if (last ())
        1:   21:        break;
        -:   22:    }
        1:   23:  do_it ();
        1:   24:}
        -:   25:
        -:   26:int
        1:   27:main ()
        -:   28:{
        1:   29:  f1 ();
        -:   30:}

do_it() is indeed called 4 times,but I wonder how the coverage statistiscs of
the for-statement is calculated. It doesn't become 5 untill the for-stamentment
is completed.
        -:    0:Source:test.c
        -:    0:Graph:test.gcno
        -:    0:Data:test.gcda
        -:    0:Runs:1
        -:    1:volatile int cnt = 0;
        -:    2:
        -:    3:__attribute__((noinline, noclone)) static int
        5:    4:last (void)
        -:    5:{
        5:    6:  return ++cnt % 5 == 0;
        -:    7:}
        -:    8:
        -:    9:__attribute__((noinline, noclone)) static void
        6:   10:do_it (void)
        -:   11:{
        6:   12:  asm volatile ("" : : "r" (&cnt) : "memory");
        6:   13:}
        -:   14:
        1:   15:static void f1 (void)
        -:   16:{
        1:   17:  do_it();
        5:   18:  for (int a=0;0<1;do_it())
        -:   19:    {
        5:   20:      if (last ())
        1:   21:        break;
        -:   22:    }
        1:   23:  do_it ();
        1:   24:}
        -:   25:
        -:   26:int
        1:   27:main ()
        -:   28:{
        1:   29:  f1 ();
        -:   30:}

[Bug gcov-profile/106090] [GCOV] Wrong coverage for loop statements

[njuwy at smail dot nju.edu.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106090

--- Comment #3 from Yang Wang <njuwy at smail dot nju.edu.cn> ---
(In reply to Yang Wang from comment #2)
> (In reply to Martin Liška from comment #1)
> > No, it's correct as it only contains do_it() call that is triggered the same
> > number times as line 20 minus one.
> 
> (In reply to Martin Liška from comment #1)
> > No, it's correct as it only contains do_it() call that is triggered the same
> > number times as line 20 minus one.
> 
>         -:    0:Source:test.c
>         -:    0:Graph:test.gcno
>         -:    0:Data:test.gcda
>         -:    0:Runs:1
>         -:    1:volatile int cnt = 0;
>         -:    2:
>         -:    3:__attribute__((noinline, noclone)) static int
>         5:    4:last (void)
>         -:    5:{
>         5:    6:  return ++cnt % 5 == 0;
>         -:    7:}
>         -:    8:
>         -:    9:__attribute__((noinline, noclone)) static void
>         6:   10:do_it (void)
>         -:   11:{
>         6:   12:  asm volatile ("" : : "r" (&cnt) : "memory");
>         6:   13:}
>         -:   14:
>         1:   15:static void f1 (void)
>         -:   16:{
>         1:   17:  do_it();
>         4:   18:  for (; 0<1; do_it())
>         -:   19:    {
>         5:   20:      if (last ())
>         1:   21:	break;
>         -:   22:    }
>         1:   23:  do_it ();
>         1:   24:}
>         -:   25:
>         -:   26:int
>         1:   27:main ()
>         -:   28:{
>         1:   29:  f1 ();
>         -:   30:}
> 
> do_it() is indeed called 4 times,but I wonder how the coverage statistiscs
> of the for-statement is calculated. It doesn't become 5 untill the
> for-stamentment is completed.
>         -:    0:Source:test.c
>         -:    0:Graph:test.gcno
>         -:    0:Data:test.gcda
>         -:    0:Runs:1
>         -:    1:volatile int cnt = 0;
>         -:    2:
>         -:    3:__attribute__((noinline, noclone)) static int
>         5:    4:last (void)
>         -:    5:{
>         5:    6:  return ++cnt % 5 == 0;
>         -:    7:}
>         -:    8:
>         -:    9:__attribute__((noinline, noclone)) static void
>         6:   10:do_it (void)
>         -:   11:{
>         6:   12:  asm volatile ("" : : "r" (&cnt) : "memory");
>         6:   13:}
>         -:   14:
>         1:   15:static void f1 (void)
>         -:   16:{
>         1:   17:  do_it();
>         5:   18:  for (int a=0;0<1;do_it())
>         -:   19:    {
>         5:   20:      if (last ())
>         1:   21:	break;
>         -:   22:    }
>         1:   23:  do_it ();
>         1:   24:}
>         -:   25:
>         -:   26:int
>         1:   27:main ()
>         -:   28:{
>         1:   29:  f1 ();
>         -:   30:}

Besides, GCov doesn't seem to provide voerage statistics for "for(;;)" .
        -:    0:Source:test.c
        -:    0:Graph:test.gcno
        -:    0:Data:test.gcda
        -:    0:Runs:1
        -:    1:volatile int cnt = 0;
        -:    2:
        -:    3:__attribute__((noinline, noclone)) static int
        5:    4:last (void)
        -:    5:{
        5:    6:  return ++cnt % 5 == 0;
        -:    7:}
        -:    8:
        -:    9:__attribute__((noinline, noclone)) static void
        6:   10:do_it (void)
        -:   11:{
        6:   12:  asm volatile ("" : : "r" (&cnt) : "memory");
        6:   13:}
        -:   14:
        1:   15:static void f1 (void)
        -:   16:{
        1:   17:  do_it();
        -:   18:  for (; ; )
        -:   19:    {
        5:   20:      if (last ())
        1:   21:        break;
        4:   22:     do_it();
        -:   23:    }
        1:   24:  do_it ();
        1:   25:}
        -:   26:
        -:   27:int
        1:   28:main ()
        -:   29:{
        1:   30:  f1 ();
        -:   31:}

[Bug gcov-profile/106090] [GCOV] Wrong coverage for loop statements

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106090

--- Comment #4 from Martin Liška <marxin at gcc dot gnu.org> ---
> do_it() is indeed called 4 times,but I wonder how the coverage statistiscs
> of the for-statement is calculated. It doesn't become 5 untill the
> for-stamentment is completed.

We count a line every time there exists a jump from a different line that jumps
to the line.
Thus for `for(;;)` there are no statements we should jump to.

[Bug gcov-profile/106090] [GCOV] Wrong coverage for loop statements

[njuwy at smail dot nju.edu.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106090

--- Comment #5 from Yang Wang <njuwy at smail dot nju.edu.cn> ---
(In reply to Martin Liška from comment #4)
> > do_it() is indeed called 4 times,but I wonder how the coverage statistiscs
> > of the for-statement is calculated. It doesn't become 5 untill the
> > for-stamentment is completed.
> 
> We count a line every time there exists a jump from a different line that
> jumps to the line.
> Thus for `for(;;)` there are no statements we should jump to.

Thanks for your reply, I have figured it out. "0<1" is used as a conditional
judgment of permanent true, and its execution is optimized away. 
I also find that two previously reported issues(#100980,#101618) have
disappeared in the new version.Could you mind doing me a favor to take a look
if you have a minute?

[Bug gcov-profile/106090] [GCOV] Wrong coverage for loop statements

[njuwy at smail dot nju.edu.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106090

--- Comment #6 from Yang Wang <njuwy at smail dot nju.edu.cn> ---
(In reply to Martin Liška from comment #4)
> > do_it() is indeed called 4 times,but I wonder how the coverage statistiscs
> > of the for-statement is calculated. It doesn't become 5 untill the
> > for-stamentment is completed.
> 
> We count a line every time there exists a jump from a different line that
> jumps to the line.
> Thus for `for(;;)` there are no statements we should jump to.

Thanks for your confirmation! I will continue to track the progress of open
bugs.