Quick question about "print EXPR"

Quick question about "print EXPR"

[Brian Desany]

I read in the docs that "print EXPR" should work with function calls.

Why is it that no matter what argument I pass to "print floor(234.567)", I always get an answer of
"1"?

...
7               yada = 2.5;
(gdb) n
8               haha = floor(yada);
(gdb) 
9               printf("%0.1f %0.1f\n",yada, haha);
(gdb) p yada
$1 = 2.5
(gdb) p haha
$2 = 2
(gdb) p floor(yada)
$3 = 1
(gdb) l
4       int main(int argc, char *argv[]) {
5               double yada;
6               double haha;
7               yada = 2.5;
8               haha = floor(yada);
9               printf("%0.1f %0.1f\n",yada, haha);
10              return 0;
11      }
...


Thanks-
-Brian.



		
__________________________________ 
Do you Yahoo!? 
Read only the mail you want - Yahoo! Mail SpamGuard. 
http://promotions.yahoo.com/new_mail 

Re: Quick question about "print EXPR"

[Peter Barada]

>Why is it that no matter what argument I pass to "print floor(234.567)", I always get an answer of
>"1"?

Did you include <math.h> in your program?  If not you don't have a
prototype for floor that indicates it takes a double and returns a
double...

-- 
Peter Barada
peter@the-baradas.com

Re: Quick question about "print EXPR"

[Brian Desany]

> Did you include <math.h> in your program? 

Yep, I did (the program itself works properly).

FWIW the only lines of the program that aren't listed in my original post are:
#include <stdio.h>
#include <math.h>

And just in case it matters, I compiled using the command "cc -lm -g test.c".
If I compile with "gcc -lm -g test.c", "p floor(whatever)" is always 16 (rather than 1 as in the
original post).

[bdesany]$ cc -v
Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-81)
[bdesany]$ gcc -v
Reading specs from /home/bdesany/sys/lib/gcc/i686-pc-linux-gnu/3.4.2/specs
Configured with: ./configure --prefix=/home/bdesany/sys
Thread model: posix
gcc version 3.4.2

Thanks-
-Brian.




		
__________________________________ 
Do you Yahoo!? 
Take Yahoo! Mail with you! Get it on your mobile phone. 
http://mobile.yahoo.com/maildemo 

Re: Quick question about "print EXPR"

[Andrew Cagney]

Brian Desany wrote:
>>Did you include <math.h> in your program? 
> 
> 
> Yep, I did (the program itself works properly).
> 
> FWIW the only lines of the program that aren't listed in my original post are:
> #include <stdio.h>
> #include <math.h>
> 
> And just in case it matters, I compiled using the command "cc -lm -g test.c".
> If I compile with "gcc -lm -g test.c", "p floor(whatever)" is always 16 (rather than 1 as in the
> original post).

Wonder if GDB can see this.  What does:

(gdb) ptype floor

print?

Andrew

> [bdesany]$ cc -v
> Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
> gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-81)
> [bdesany]$ gcc -v
> Reading specs from /home/bdesany/sys/lib/gcc/i686-pc-linux-gnu/3.4.2/specs
> Configured with: ./configure --prefix=/home/bdesany/sys
> Thread model: posix
> gcc version 3.4.2
> 
> Thanks-
> -Brian.
> 
> 
> 
> 
> 		
> __________________________________ 
> Do you Yahoo!? 
> Take Yahoo! Mail with you! Get it on your mobile phone. 
> http://mobile.yahoo.com/maildemo 
> 

RE: Quick question about "print EXPR"

[Brian Desany]

 
> Brian Desany wrote:
> >>Did you include <math.h> in your program? 
> > 
> > 
> > Yep, I did (the program itself works properly).
> > 
> > FWIW the only lines of the program that aren't listed in my 
> original post are:
> > #include <stdio.h>
> > #include <math.h>
> > 
> > And just in case it matters, I compiled using the command 
> "cc -lm -g test.c".
> > If I compile with "gcc -lm -g test.c", "p floor(whatever)" 
> is always 
> > 16 (rather than 1 as in the original post).
> 
> Wonder if GDB can see this.  What does:
> 
> (gdb) ptype floor
> 
> print?

type = int ()

I was wrong about "p floor(whatever)" always returning a constant value. It
returns a different value depending apparently on what line of the program
I'm currently stopped on:

Breakpoint 1, main (argc=1, argv=0xbfffcf54) at test.c:9
9                        yada = 2.5;
(gdb) p floor(123.456)
$1 = 0
(gdb) n
10                       haha = floor(yada);
(gdb) p floor(123.456)
$2 = 0
(gdb) n
11                       printf("%0.1f %0.1f\n",yada, haha);
(gdb) p floor(123.456)
$3 = 3199
(gdb) n
2.5 -0.0
12                       return 0;
(gdb) p floor(123.456)
$4 = 9
(gdb) 

> 
> Andrew
> 
> > [bdesany]$ cc -v
> > Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
> > gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-81) 
> [bdesany]$ gcc 
> > -v Reading specs from 
> > /home/bdesany/sys/lib/gcc/i686-pc-linux-gnu/3.4.2/specs
> > Configured with: ./configure --prefix=/home/bdesany/sys 
> Thread model: 
> > posix gcc version 3.4.2
> > 
> > Thanks-
> > -Brian.

Re: Quick question about "print EXPR"

[Andrew Cagney]

Brian Desany wrote:
[sorry, going backwards through my e-mail]

>>Brian Desany wrote:
>>
>>>>Did you include <math.h> in your program? 
>>>
>>>
>>>Yep, I did (the program itself works properly).
>>>
>>>FWIW the only lines of the program that aren't listed in my 
>>
>>original post are:
>>
>>>#include <stdio.h>
>>>#include <math.h>
>>>
>>>And just in case it matters, I compiled using the command 
>>
>>"cc -lm -g test.c".
>>
>>>If I compile with "gcc -lm -g test.c", "p floor(whatever)" 
>>
>>is always 
>>
>>>16 (rather than 1 as in the original post).
>>
>>Wonder if GDB can see this.  What does:
>>
>>(gdb) ptype floor
>>
>>print?
> 
> 
> type = int ()

Which hopefully explains why things go wrong - GDB doesn't know the 
function's signature and hence invokes it incorrectly.  Here GDB is 
probably pulling the return value out of an integer register instead of 
floating-point register or stack address (as you step through the code 
that integer register's value would change and hence the value would 
change).

There are two things from here:

- is the debug info stabs?
I believe that only dwarf>=2 debug info can describe the above.
=> Should GDB issue a warning when calling a function with no visible 
prototype?

- is the debug info dwarf
Check the input file (you'll need to look at binutils's readelf program) 
to see if the info is there.  Beyond that we'd have to think out why GDB 
can't see it.

Andrew


> I was wrong about "p floor(whatever)" always returning a constant value. It
> returns a different value depending apparently on what line of the program
> I'm currently stopped on:
> 
> Breakpoint 1, main (argc=1, argv=0xbfffcf54) at test.c:9
> 9                        yada = 2.5;
> (gdb) p floor(123.456)
> $1 = 0
> (gdb) n
> 10                       haha = floor(yada);
> (gdb) p floor(123.456)
> $2 = 0
> (gdb) n
> 11                       printf("%0.1f %0.1f\n",yada, haha);
> (gdb) p floor(123.456)
> $3 = 3199
> (gdb) n
> 2.5 -0.0
> 12                       return 0;
> (gdb) p floor(123.456)
> $4 = 9
> (gdb) 
> 
> 
>>Andrew
>>
>>
>>>[bdesany]$ cc -v
>>>Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
>>>gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-81) 
>>
>>[bdesany]$ gcc 
>>
>>>-v Reading specs from 
>>>/home/bdesany/sys/lib/gcc/i686-pc-linux-gnu/3.4.2/specs
>>>Configured with: ./configure --prefix=/home/bdesany/sys 
>>
>>Thread model: 
>>
>>>posix gcc version 3.4.2
>>>
>>>Thanks-
>>>-Brian.
> 
> 
>