Difference in -E output

Difference in -E output

[Tom Kacvinsky]

[-- Attachment #1: Type: text/plain, Size: 1206 bytes --]

I am working on an issue that involves our tool depending on the
preprocessed output from gcc -E.

This simple test program exhibits the problem

#include <stdlib.h>
int main() {
  return EXIT_SUCCESS;
}

With some elision, this preprocesses to:

# 3 "test.c"

int main() {

  return

# 4 "test.c" 3 4

        0

# 4 "test.c"

                    ;

}


Notice how when the macro EXIT_SUCCESS is expanded, we get several lines
around the 0?


This is happening for us with GCC 8.3.  But using something as early as
4.8.5 (yeah, I know,

bad idea), we get this:


int main() {

  return 0;

}


I am pretty sure this is an intentional change.  What I'd like to know is
if there is a way of

reverting back to the former behavior?  I looked at


https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html#Preprocessor-Options


But didn't see anything that jumped out at me and said "this is how you can
work around it"


I did a modicum of bi-secting to see that this change was introduced
somewhere between gcc
4.9.3 and 5.4.0.  I can go deeper to figure out which version actually
broke it if you would find

it helpful.


If anyone has some ideas on this, I'd appreciate hearing them.


Thanks,


Tom

Re: Difference in -E output

[Segher Boessenkool]

On Thu, Oct 27, 2022 at 04:22:11PM -0400, Tom Kacvinsky via Gcc-help wrote:
> I am working on an issue that involves our tool depending on the
> preprocessed output from gcc -E.

> # 4 "test.c" 3 4
> 
>         0
> 
> # 4 "test.c"
> 
>                     ;
> 
> }
> 
> 
> Notice how when the macro EXIT_SUCCESS is expanded, we get several lines
> around the 0?

I don't get any of the blank lines you have though?

> This is happening for us with GCC 8.3.  But using something as early as
> 4.8.5 (yeah, I know,
> 
> bad idea), we get this:
> 
> 
> int main() {
> 
>   return 0;
> 
> }
> 
> 
> I am pretty sure this is an intentional change.  What I'd like to know is
> if there is a way of
> 
> reverting back to the former behavior?  I looked at

-P will do what you want (as a side effect though).  -traditional also
works (and that one is more intentional).  Neither of those does *only*
this, but either may be usable for your purposes.

Have fun,


Segher

Re: Difference in -E output

[Tom Kacvinsky]

[-- Attachment #1: Type: text/plain, Size: 1272 bytes --]

On Thu, Oct 27, 2022 at 6:30 PM Segher Boessenkool <
segher@kernel.crashing.org> wrote:

> On Thu, Oct 27, 2022 at 04:22:11PM -0400, Tom Kacvinsky via Gcc-help wrote:
> > I am working on an issue that involves our tool depending on the
> > preprocessed output from gcc -E.
>
> > # 4 "test.c" 3 4
> >
> >         0
> >
> > # 4 "test.c"
> >
> >                     ;
> >
> > }
> >
> >
> > Notice how when the macro EXIT_SUCCESS is expanded, we get several lines
> > around the 0?
>
> I don't get any of the blank lines you have though?
>
> > This is happening for us with GCC 8.3.  But using something as early as
> > 4.8.5 (yeah, I know,
> >
> > bad idea), we get this:
> >
> >
> > int main() {
> >
> >   return 0;
> >
> > }
> >
> >
> > I am pretty sure this is an intentional change.  What I'd like to know is
> > if there is a way of
> >
> > reverting back to the former behavior?  I looked at
>
> -P will do what you want (as a side effect though).  -traditional also
> works (and that one is more intentional).  Neither of those does *only*
> this, but either may be usable for your purposes.
>

A colleague informed me that -ftrack-macro-expansion=0 does exactly what I
want.

Said colleague also told me to call them the next time a question like this
came up.  Haha