julbro@auto-trol.com (Julie Brown) (06/18/91)
Can anyone explain to me why the following does not work
(at least on the Sun):
q = q--;
'q' does not get decremented. No matter what the order of
operations, I would expect 'q' to get decremented at some
point. It doesn't make a whole lot of sense to do this until
you put it in a more meaningful context:
#include <stdio.h>
main()
{
int count = 5;
count = (count > 0) ? count-- : 0;
printf("count %d\n", count);
}
the output from this program is :
count 5
pre-increment works just fine (q = --q;).
just curious.
--
Julie Brown
julbro@auto-trol.COM Auto-trol Technology Corporation
{...}ncar!ico!auto-trol!julbro 12500 North Washington Street
(303) 252-2856 Denver, CO 80241-2404rgm@ocf.berkeley.edu (Rob Menke) (06/18/91)
In article <1991Jun17.230838.9628@auto-trol.com> julbro@auto-trol.com (Julie Brown) writes: Can anyone explain to me why the following does not work (at least on the Sun): q = q--; It's quite simple: set-value ('=') has a lower precedence than post-decrement. So, the expression on the RHS is evaluated (it equals 'q'), q is decremented, then q is set to the stored value from the RHS. Net result: no decrement. You're better off with: q = (q > 0) ? q - 1 : 0; -- "It is a pity that you Autobots die so | Robert Menke easily... otherwise one might get a | rgm@OCF.berkeley.edu sense of satisfaction..." | ...!ucbvax!OCF!rgm
worley@compass.com (Dale Worley) (06/18/91)
In article <1991Jun17.230838.9628@auto-trol.com> julbro@auto-trol.com (Julie Brown) writes:
Can anyone explain to me why the following does not work
(at least on the Sun):
q = q--;
pre-increment works just fine (q = --q;).
According to Ansi C, neither of these is valid, because they both
modify q twice between two sequence points (during a statement). The
fact that the second happens to do what you want it to is luck.
BTW, what *is* it that you want to do? "q--" is equivalent to "q = q-1"
already, so you don't have to assign the result to q...
Dale Worley Compass, Inc. worley@compass.com
--
Start the project after the deadline -- the time pressure is less then.steve@taumet.com (Stephen Clamage) (06/18/91)
rgm@ocf.berkeley.edu (Rob Menke) writes: >In article <1991Jun17.230838.9628@auto-trol.com> julbro@auto-trol.com >(Julie Brown) writes: > Can anyone explain to me why the following does not work > q = q--; >It's quite simple: set-value ('=') has a lower precedence than >post-decrement. So, the expression on the RHS is evaluated (it equals >'q'), q is decremented, then q is set to the stored value from the >RHS. Net result: no decrement. This is not a correct explanation. We have been through endless comments on another thread, the evaluation of if( (i = 1) == (i = 2) ) and the situation here is the same. In the expression q = q-- variable q is assigned a value in two places with no intervening sequence point. The result of the double assignment between sequence points is undefined. A compiler could reasonably evaluate the expression as if you had written q = q; q--; or as if you had written tmp = q; q--; q = tmp; In fact, since the result is specifically "undefined" (a technical term in ANSI C), the compiler is free to do whatever it likes, such as send you mail at run time telling you not to modify a variable twice between sequence points. -- Steve Clamage, TauMetric Corp, steve@taumet.com
timr@gssc.UUCP (Tim Roberts) (06/19/91)
In article <RGM.91Jun17184532@avalanche.Berkeley.EDU> rgm@ocf.berkeley.edu (Rob Menke) writes: >In article <1991Jun17.230838.9628@auto-trol.com> julbro@auto-trol.com >(Julie Brown) writes: > > Can anyone explain to me why the following does not work > (at least on the Sun): > > q = q--; > >It's quite simple: set-value ('=') has a lower precedence than >post-decrement. So, the expression on the RHS is evaluated (it equals >'q'), q is decremented, then q is set to the stored value from the >RHS. Net result: no decrement. Although we need to be aware that this result is simply coincidence, and some other compiler might produce different results. Following the lengthy and rather heated debate about the results of undefined statements in ANSI C, I hope all you viewers out there at home recognized that the results of q = q--; are UNDEFINED, because, like the statement which started this whole mess, this statement modifies a variable twice without an intervening sequence point. >You're better off with: > q = (q > 0) ? q - 1 : 0; Correct. The results of this are well-defined. -- timr@gssc.gss.com Tim N Roberts, CCP Graphic Software Systems Beaverton, OR This is a very long palindrome. .emordnilap gnol yrev a si sihT
Dave.Harris@f14.n15.z1.fidonet.org (Dave Harris) (06/20/91)
>From: julbro@auto-trol.com (Julie Brown) >Can anyone explain to me why the following does not work >(at least on the Sun): > q = q--; As I have been reminded several times in the last week myself, this statement is undefined. Modifing an address multiple time between sequence points is undefined. Both pre and post yeild 4 when I try this with what I use (TC++). Being undefined, your machine could yeild or do anything including crash on this statement (so I've been told). q-- all by its lonesome self is sufficient. >'q' does not get decremented. No matter what the order of >operations, I would expect 'q' to get decremented at some >point. It doesn't make a whole lot of sense to do this until >you put it in a more meaningful context: In machine language terms your compiler is possibly doing something like: R1 = Q Dec Q Q = R1 where Q your variable and R1 is a register. Just speculating of course. Fiddle with the optimizer switches and see if it acts differently. >#include <stdio.h> >main() >{ > int count = 5; > count = (count > 0) ? count-- : 0; > printf("count %d\n", count); >} >the output from this program is : > count 5 count = (count > 0) ? count-1 : 0; count = max(0,count-1); count -= (count>0); no shortage of ways to do this. >pre-increment works just fine (q = --q;). Dec Q R1 = Q Q = R1 Speculating again of course. -- Uucp: ...{gatech,ames,rutgers}!ncar!asuvax!stjhmc!15!14!Dave.Harris Internet: Dave.Harris@f14.n15.z1.fidonet.org