The issue is somewhat complicated, because even in C there is an interpretation that says that the C standard does not quite define the behavior here. However, gcc and jBASE are generally thought to be correct. You also have to be careful with optimizations such as O4, which can change such things. As a general rule of thumb, you will find that when xlc differs from other compilers, it is marginally justified but usually (but not always) 'wrong'. The only exception to this rule is that if it is differing from an HP compiler, you can usually (but not always) bet your last cent on the HP compiler doing something stupid. :-)
In general, the best way to handle such issues is to avoid them, be more explicit with parenthesis and break down the expression. In fact there is a more important reason why you should do that and this is that 'clever' expressions such as this are beyond the ability of most to just read the code and know exactly what it is doing. Hence, for the sake of maintainability, we should never be 'clever' unless it is the only way. The optimizers will take care of almost everything these days. Besides which, the usual problem with a program is algorithmic, which neither a clever line of code, nor an optimizer (at least, not as yet) can fix. Jim > -----Original Message----- > From: [email protected] [mailto:[email protected]] On > Behalf Of Chenai > Sent: Monday, September 06, 2010 1:14 AM > To: jBASE > Subject: Re: Pre-increment > > Hi, > > Actually eh, the issue isnt quite settled. On AIX 5.3 I cut and paste your > program and the output is 9 not 10 for the pre-increment. I compiled the > program with xlc without passing any compiler options. > jbase 5 on the same server using the same compiler gives a result of > 10 for the OP's code. What am I missing? > > > 0001 #include <stdio.h> > 0002 > 0003 int main(int cnt, char * argc[]) > 0004 { > 0005 int c; > 0006 > 0007 c = 3; > 0008 printf("Expr is %d\n, c is %d\n\n", ++c + ++c, c); > 0009 c = 3; > 0010 printf("Expr is %d\n, c is %d\n\n", c++ + c++, c); > 0011 > 0012 return 0; > 0013 } > > Expr is 9 > , c is 5 > > Expr is 6 > , c is 5 > > > Chenai. > > On Sep 5, 9:57 pm, "Jim Idle" <[email protected]> wrote: > > Greg, > > > > Thank you for your proof of Fermat's last theorem, your mistake is at > > line 6. > > > > I have been trying to teach you C for so long now that I cannot > > remember how long it is, but if this is an example of how much you > > have learned, then perhaps I spent about 20 minutes on your lessons? > > Run the program I attached and not your incorrect program. Mine > > confirms that cnt is 5 and that therefore the expression is 10, which your C > program is merely confirming. > > as well as confirming that one should not try to write C at 9PM on a > > Sunday after a few beers and a curry. unless you are me. > > > > If a prior version of jBASE did not return 10 in this instance then it > > was incorrect. In fact, I think I remember fixing this bug in 4.x (and > > specifically only at the major point release). or maybe I don't. > > Perhaps you remember me doing this. nah, it was more than 20 minutes > > ago ;-) (c.f. The Invisible Gorilla) > > > > Anyway, your program is evaluating the expression and storing it in > > the variable 'res', but then your printf statement is outputting the > > value of cnt (I think you are missing a letter perhaps? =:?o), which > > as cnt is has been incremented twice by then, will, as you acidulously > > demonstrate, have the value 5, thus proving that C thinks that 3 +1 +1 is 5. > > > > So, you have fallen for the same issue only worse, in that the ++var > > will always happen before the binary operator + and so this will be 5 + 5 = > 10. > > However your mistake is even worse because you are not even printing > > out the correct variable. If you are having trouble typing in the > > program that I attached to my email in the first place, then please > > let me know, so I can show you how to type in October when I am back > > in London. I was going to visit with you of course, but I am not sure > > if I should be associating with junior C programmers ;-) > > > > Now, it could well be that the original parser at some point in 3.x > > did not have the correct precedence on the pre_increment operator and > > so it gives the wrong answer. However, the answer that the OP is > > getting from the latest jBASE is the correct one. > > > > In future my eld padwan, you should concentrate on handing out the res > > due to your old master, otherwise you risk looking like a > > > > As I presume that pointing out that your C program is utter crap and > > merely verifies my much cleverer program, I take it that this is > > sufficient demolition of your argument and that you will now be buying > > the second and third rounds as well as the first, as well of course being. > > > > Yours respectfully, etc etc, > > > > Ben Franklin. > > > > From: [email protected] [mailto:[email protected]] On > Behalf > > Of Gregory Cooper > > Sent: Sunday, September 05, 2010 12:13 PM > > To: [email protected] > > Subject: RE: Pre-increment > > > > Jim me eld mucker, > > > > All very well and good, but .. > > > > On jBASE 3.x, the answer given is 9 (and then 15 for the second test). > > This is what Xze (??) was expecting and differs to what he saw on > > jBASE 5.x , answers of 10 and 16. > > > > So no matter what your argument, the fact we get different answers on > > different jBASE releases shows that something is wrong. As you were > > responsible for the jBASE 3.x compiler and not for the jBASE 5.x > > compiler, I would obviously expect the jBASE 3.x answer to be correct > > ;-) > > > > You then talk about "reference this in C if you don't believe me". > > Now, not as though I didn't believe you or anything, but writing a > > simple C program on Suse 11.2 gives this > > > > main() > > > > { > > > > int cnt , res ; > > > > cnt = 3 ; > > > > res = ++cnt + ++cnt ; > > > > printf("cnt = %d\n",cnt); > > > > } > > > > cnt = 5 > > > > So that doesn't match up with either of the jBASE releases or what you > > had said about C. > > > > So there is obviously some behaviour that is hard to pin down with > > these pre and post increments. > > > > I am sure I am missing something here, and await your demolishment of > > this email to prove you right again ! ;-) > > > > Greg > > > > From: [email protected] [mailto:[email protected]] On > Behalf > > Of Jim Idle > > Sent: 02 September 2010 17:11 > > To: [email protected] > > Subject: RE: Pre-increment > > > > This is correct, reference this in C if you don't believe me. > > Technically it is because the operator is unary right associative and is 'pre' > increment. > > Hence, the instruction set generated is (roughly): > > > > INC CNT in place ; CNT is now 4 > > > > INC CNT in place ; CNT is now 5 > > > > LD CNT ; 5 > > > > ADD CNT ; Add 5 > > > > And the answer is 10. I presume that you can derive the sequence for > > your second example from this. > > > > Now contrast that with 'post' increment. > > > > Jim > > > > #include <stdio.h> > > > > int > > > > main(int cnt, char * argc[]) > > > > { > > > > int c, c1; > > > > c = 3; > > > > printf("Expr is %d\n, c is %d\n\n", ++c + ++c, c); > > > > c = 3; > > > > printf("Expr is %d\n, c is %d\n\n", c++ + c++, c); > > > > } > > > > From: [email protected] [mailto:[email protected]] On > Behalf > > Of [ Xze ] > > Sent: Wednesday, September 01, 2010 5:35 AM > > To: [email protected] > > Subject: Pre-increment > > > > Hi all! > > > > My question pertains to the way that jbase treats pre-increments > > > > Consider the following routine: > > > > 0004 PROGRAM MY.MATH > > 0005 > > 0006 CNT = 3 > > 0007 RES = ++CNT + ++CNT > > 0008 > > 0009 PRINT 'RES = ':RES > > 0010 RETURN > > 0011 END > > > > The output result is RES = 10 > > > > if i change line 007 to RES = ++CNT + ++CNT + ++CNT, then the output > > result is RES = 16 > > > > Can anybody explain this behaviour? (I expected the first case to > > output RES = 9 and the second RES = 15) > > > > My system configuration: > > > > OS : AIX 5.3 > > jbase : Major 5.0 , Minor 20 , Patch 0364 (Change 85159) > > > > Regards > > > > -- > > Please read the posting guidelines > > at:http://groups.google.com/group/jBASE/web/Posting%20Guidelines > > > > IMPORTANT: Type T24: at the start of the subject line for questions > > specific to Globus/T24 > > > > To post, send email to [email protected] To unsubscribe, send > > email to [email protected] > > For more options, visit this group > > athttp://groups.google.com/group/jBASE?hl=en > > > > -- > > Please read the posting guidelines > > at:http://groups.google.com/group/jBASE/web/Posting%20Guidelines > > > > IMPORTANT: Type T24: at the start of the subject line for questions > > specific to Globus/T24 > > > > To post, send email to [email protected] To unsubscribe, send > > email to [email protected] > > For more options, visit this group > > athttp://groups.google.com/group/jBASE?hl=en > > > > -- > > Please read the posting guidelines > > at:http://groups.google.com/group/jBASE/web/Posting%20Guidelines > > > > IMPORTANT: Type T24: at the start of the subject line for questions > > specific to Globus/T24 > > > > To post, send email to [email protected] To unsubscribe, send > > email to [email protected] > > For more options, visit this group > > athttp://groups.google.com/group/jBASE?hl=en > > -- > Please read the posting guidelines at: > http://groups.google.com/group/jBASE/web/Posting%20Guidelines > > IMPORTANT: Type T24: at the start of the subject line for questions specific > to Globus/T24 > > To post, send email to [email protected] To unsubscribe, send email > to [email protected] > For more options, visit this group at > http://groups.google.com/group/jBASE?hl=en -- Please read the posting guidelines at: http://groups.google.com/group/jBASE/web/Posting%20Guidelines IMPORTANT: Type T24: at the start of the subject line for questions specific to Globus/T24 To post, send email to [email protected] To unsubscribe, send email to [email protected] For more options, visit this group at http://groups.google.com/group/jBASE?hl=en
