[llvm-bugs] [Bug 51034] New: Unable to call K&R declared function with float arguments in debugger after 6c9559b67b91966bfeff9e17808a3e84a92e64a0

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=51034

            Bug ID: 51034
           Summary: Unable to call K&R declared function with float
                    arguments in debugger after
                    6c9559b67b91966bfeff9e17808a3e84a92e64a0
           Product: new-bugs
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: douglas_yung at playstation.sony.com
                CC: htmldeveloper at gmail.com, llvm-bugs at lists.llvm.org

Internally we run the GDB test suite and a couple of tests in there started
failing recently which I bisected to commit
6c9559b67b91966bfeff9e17808a3e84a92e64a0. The root of the problem seems to be
that after 6c9559b67, when trying to invoke a function from a debugging session
in gdb, if the function was declared in a K&R style, it does not seem to pass
float arguments to the function being called.

For example, consider the following code:
/* test.c */
#include <stdio.h>

#define DELTA (0.001)

float float_val1 = 3.14159;
float float_val2 = -2.3765;

int t_float_values (float_arg1, float_arg2)                                     
  float float_arg1, float_arg2;                                                 
{
  printf("float_arg1 = %f, float_arg2 = %f\n", float_arg1, float_arg2);
  return ((float_arg1 - float_val1) < DELTA
          && (float_arg1 - float_val1) > -DELTA
          && (float_arg2 - float_val2) < DELTA
          && (float_arg2 - float_val2) > -DELTA);
}

int foo1() {
  return t_float_values(3.14159,-2.3765);
}

int main(int argc, char *argv[]) {
  return foo1();
}
/* test.c */

If you compile that code with debug information and then debug the binary using
gdb with the following commands, you can see the problem:
 1. b main
 2. r
 3. p foo1()
 4. p t_float_values(3.14159,-2.3765)

The return value of the commands in 3 and 4 should be the same. When test.c is
compiled with a compiler prior to 6c9559b67, that is indeed the case. But if
compiled with a compiler that includes 6c9559b67, the return value of 4 is 0
instead of 1 like it should be. Additionally, from the printf() statement, you
can see that the values of float_arg1 and float_arg2 are not being passed into
the function when it is invoked in 4.

GDB session using binary built by a "good" compiler:
Reading symbols from test.good.elf...
(gdb) b main
Breakpoint 1 at 0x401266: file test.c, line 23.
(gdb) r
Starting program: /home/dyung/sandbox/gdb/test.good.elf 

Breakpoint 1, main (argc=1, argv=0x7fffffffe458) at test.c:23
23        return foo1();
(gdb) p foo1()
float_arg1 = 3.141590, float_arg2 = -2.376500
$1 = 1
(gdb) p t_float_values(3.14159,-2.3765)
float_arg1 = 3.141590, float_arg2 = -2.376500
$2 = 1


GDB session using binary built by a "bad" compiler:
Reading symbols from test.bad.elf...
(gdb) b main
Breakpoint 1 at 0x401266: file test.c, line 23.
(gdb) r
Starting program: /home/dyung/sandbox/gdb/test.bad.elf 

Breakpoint 1, main (argc=1, argv=0x7fffffffe458) at test.c:23
23        return foo1();
(gdb) p foo1()
float_arg1 = 3.141590, float_arg2 = -2.376500
$1 = 1
(gdb) p t_float_values(3.14159,-2.3765)
float_arg1 = 0.000000, float_arg2 = 0.000000
$2 = 0

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20210709/f78c5975/attachment.html>