www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - Screwed up decrement of bytes and shorts

reply Stewart Gordon <smjg_1998 yahoo.com> writes:
Using DMD 0.120, Windows 98SE.

---------
import std.stdio;

void main() {
	byte x = 9;
	writefln(x);
	writefln(--x);
}
---------
9
6618376
---------

The bug also happens with ubyte, short and ushort.  It disappears when:
- the writefln(x) statement is removed
- an intermediate variable of same-size or smaller type is used, like this

     writefln(x);
     byte y = --x;
     writefln(y);

but it still happens if the intermediate variable is a larger integer type.

Using --x as an array index screws up similarly, tending to cause an 
ArrayBoundsError:

---------
import std.stdio;

int[12] MONTH_OFFSET = [
     -1, 30, 58, 89, 119, 150, 180, 211, 242, 272, 303, 333
];

void main() {
     byte month = 9;

     writefln(month);
     int off = MONTH_OFFSET[--month];
     writefln(off);
}
----------

Presumably ++ has the same bug, but I didn't get round to testing it. 
Nor did I get round to testing postincrement/decrement....

Stewart.

-- 
My e-mail is valid but not my primary mailbox.  Please keep replies on 
the 'group where everyone may benefit.
Apr 18 2005
next sibling parent Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Stewart Gordon schrieb am Mon, 18 Apr 2005 11:32:20 +0100:
 Using DMD 0.120, Windows 98SE.

I can't reproduce this with Linux and dmd 0.110 .. 0.121. Does the sample below reproduce the error on your system? void dummy(...){ } int main(){ byte x=9; dummy(x); dummy(--x); printf("%i\n", x); return 0; } Thomas -----BEGIN PGP SIGNATURE----- iD8DBQFCY7Gz3w+/yD4P9tIRAp+MAKDMVD28qXt/8g1Ex3YqYHMmkCgKfwCfdfjD OomRyDods8qiQBhdYqppGz4= =Rc1e -----END PGP SIGNATURE-----
Apr 18 2005
prev sibling next sibling parent Tom S <h3r3tic remove.mat.uni.torun.pl> writes:
Stewart Gordon wrote:

 Using DMD 0.120, Windows 98SE.
 
 ---------
 import std.stdio;
 
 void main() {
     byte x = 9;
     writefln(x);
     writefln(--x);
 }
 ---------
 9
 6618376
 ---------

Confirmed. I'm getting the same bug with DMD 0.121 on WinXP SP2 Yet the numbers are different --------- 9 1244936 --------- What's specific about the second number is that its hexadecimal representation == 12FF08h. The last byte == 8. With your numbers, it's 64FD08h, so again the last byte == 8. When using a short, I'm getting 120008, so again, the last word == 8. For some reason the compiler is confusing that bytes and shorts with ints. The bug goes away when compiling with ' -o '. This runs as expected: --------- import std.stdio; void main() { byte x = 9; writefln(x); --x; writefln(x); } --------- -- Tomasz Stachowiak /+ a.k.a. h3r3tic +/
Apr 18 2005
prev sibling parent reply Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Stewart Gordon schrieb am Mon, 18 Apr 2005 11:32:20 +0100:
 Using DMD 0.120, Windows 98SE.

 ---------
 import std.stdio;

 void main() {
 	byte x = 9;
 	writefln(x);
 	writefln(--x);
 }
 ---------
 9
 6618376
 ---------

 The bug also happens with ubyte, short and ushort.  It disappears when:
 - the writefln(x) statement is removed
 - an intermediate variable of same-size or smaller type is used, like this

      writefln(x);
      byte y = --x;
      writefln(y);

 but it still happens if the intermediate variable is a larger integer type.

 Using --x as an array index screws up similarly, tending to cause an 
 ArrayBoundsError:

 ---------
 import std.stdio;

 int[12] MONTH_OFFSET = [
      -1, 30, 58, 89, 119, 150, 180, 211, 242, 272, 303, 333
 ];

 void main() {
      byte month = 9;

      writefln(month);
      int off = MONTH_OFFSET[--month];
      writefln(off);
 }

Added to DStress as http://dstress.kuehne.cn/run/opPostDec_01.d http://dstress.kuehne.cn/run/opPostDec_02.d http://dstress.kuehne.cn/run/opPostDec_03.d http://dstress.kuehne.cn/run/opPostDec_04.d http://dstress.kuehne.cn/run/opPostDec_05.d http://dstress.kuehne.cn/run/opPostDec_06.d http://dstress.kuehne.cn/run/opPostDec_07.d http://dstress.kuehne.cn/run/opPostDec_08.d http://dstress.kuehne.cn/run/opPostDec_09.d http://dstress.kuehne.cn/run/opPostDec_10.d http://dstress.kuehne.cn/run/opPostDec_11.d http://dstress.kuehne.cn/run/opPostInc_01.d http://dstress.kuehne.cn/run/opPostInc_02.d http://dstress.kuehne.cn/run/opPostInc_03.d http://dstress.kuehne.cn/run/opPostInc_04.d http://dstress.kuehne.cn/run/opPostInc_05.d http://dstress.kuehne.cn/run/opPostInc_06.d http://dstress.kuehne.cn/run/opPostInc_07.d http://dstress.kuehne.cn/run/opPostInc_08.d http://dstress.kuehne.cn/run/opPostInc_09.d http://dstress.kuehne.cn/run/opPostInc_10.d http://dstress.kuehne.cn/run/opPostInc_11.d http://dstress.kuehne.cn/run/opPreDec_01.d http://dstress.kuehne.cn/run/opPreDec_02.d http://dstress.kuehne.cn/run/opPreDec_03.d http://dstress.kuehne.cn/run/opPreDec_04.d http://dstress.kuehne.cn/run/opPreDec_05.d http://dstress.kuehne.cn/run/opPreDec_06.d http://dstress.kuehne.cn/run/opPreDec_07.d http://dstress.kuehne.cn/run/opPreDec_08.d http://dstress.kuehne.cn/run/opPreDec_09.d http://dstress.kuehne.cn/run/opPreDec_10.d http://dstress.kuehne.cn/run/opPreDec_11.d http://dstress.kuehne.cn/run/opPreInc_01.d http://dstress.kuehne.cn/run/opPreInc_02.d http://dstress.kuehne.cn/run/opPreInc_03.d http://dstress.kuehne.cn/run/opPreInc_04.d http://dstress.kuehne.cn/run/opPreInc_05.d http://dstress.kuehne.cn/run/opPreInc_06.d http://dstress.kuehne.cn/run/opPreInc_07.d http://dstress.kuehne.cn/run/opPreInc_08.d http://dstress.kuehne.cn/run/opPreInc_09.d http://dstress.kuehne.cn/run/opPreInc_10.d http://dstress.kuehne.cn/run/opPreInc_11.d Thomas -----BEGIN PGP SIGNATURE----- iD8DBQFCZITN3w+/yD4P9tIRAkaAAKC837Jp7swBMZn/T02/rQyDFEhNMwCfR0MG H0EZjuNcd7DZ8LWHm0wuKZY= =OxKb -----END PGP SIGNATURE-----
Apr 18 2005
parent Stewart Gordon <smjg_1998 yahoo.com> writes:
Thomas Kuehne wrote:
<snip>
 Added to DStress as
 http://dstress.kuehne.cn/run/opPostDec_01.d

These testcases aren't quite adequate - they test only the final value of the variable, not the value of the increment/decrement expressions. But this reproduced the problem: ---------- void dummy(...) {} void main() { byte x = 9; dummy(x); int y = --x; assert (y == 8); assert (x == 8); } ---------- I also found the exact same problem with +=. Probably -= as well, but strangely *= didn't show it. We should have some testcases for various op= as well as both prefix and postfix ++ and --. Stewart. -- My e-mail is valid but not my primary mailbox. Please keep replies on the 'group where everyone may benefit.
Apr 19 2005