Re: [ast-users] Arithmetic assignment side-effects

[Roland Mainz]

On Sun, Aug 4, 2013 at 12:04 AM, Dan Douglas <ormaaj@gmail.com> wrote:
> Is it specified what the value of x should be after this expression?
>
> x=0; : $((x+=x=1))
>
> Bash, ksh93, mksh, posh say 1. zsh, dash, busybox say 2. Clang and gcc both
> throw warnings about it, but both plus icc agree on 2.

Just curious: Is that x86-specific or is the result always the same on
other architectures, too ? Maybe there is something in ISO C1X/C99
which actually defines or recommends a specific compiler behaviour.
Maybe we should ask the "clang" people about the "why ?" ...

> <stdin>:1:42: warning: unsequenced modification and access to 'x' 
> [-Wunsequenced]
> int main() { int x=0; printf("%d\n", x+=x=1); return 0; }
>                                       ~~ ^
> 1 warning generated.
> 2

Mhhh... I tried this on SuSE 12.3/AMD64/64bit and Solaris 11 with Sun Studio 12:
-- snip --
$ cat -n x.c
     1  #include <stdlib.h>
     2  #include <stdio.h>
     3
     4  int main(int ac, char *av[])
     5  {
     6          float x;
     7
     8          x=0;
     9          ((x+=x=1));
    10
    11          printf("x=%f\n", x);
    12
    13          return(EXIT_SUCCESS);
    14  }
    15
$ cc -std=c99 -Wall x.c # or cc -xc99=%all x.c for Sun Studio
x.c: In function ‘main’:
x.c:9:5: warning: operation on ‘x’ may be undefined
$ ./a.out
x=2.000000
-- snip --

AFAIK the next best thing is to issue a warning for ksh -n that
((var1+=var1=othervaror_const)) creates undefined results (maybe we
get arithmetic optimisations in the future and then this will bite
back) ...

----

Bye,
Roland

-- 
  __ .  . __
 (o.\ \/ /.o) roland.mainz@nrubsig.org
  \__\/\/__/  MPEG specialist, C&&JAVA&&Sun&&Unix programmer
  /O /==\ O\  TEL +49 641 3992797
 (;O/ \/ \O;)