• =?UTF-8?B?VGhvbWFzIEvDvGhuZQ==?= (81/81) May 14 2005 -----BEGIN PGP SIGNED MESSAGE-----
• Walter (1/1) May 14 2005 Yeah, I was thinking that too. You're right.

=?UTF-8?B?VGhvbWFzIEvDvGhuZQ==?= writes:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The current version/debug system has some undefined/unenforced corners.


version(A){
	pragma(msg, "everytime");
}else{
	version=A;
}


The "everytime" message will always be displayed.



version=A;

version(A){
	version=B;
}

version(B){
	pragma(msg, "sometimes");
}



version(A){
	version=B;
}

version=A;

version(B){
	pragma(msg, "sometimes");
}


The content of sample 3 is a reordered sample 2.
Sample 2 displays the message whereas sample 3 displays nothing.


I suggest a query-change lock mechanism to close those corners.

The compiler has to ensure that re-evaluating a conditional -
"version(A)" - will always have the same result.


version(A){
	pragma(msg, "A: true");
}
version=A; // ***illegal, the status of A was changed***




version=A;
version(A){
	pragma(msg, "A: true");
}
version=A; // ***legal, the status of A didn't change ***




version=2;
version(2){
	pragma(msg, "level >= 2");
}
version=3; // ***legal***

true: newLevel >= highestQuery
true: highestQuery >= current level



version=1;
version(2){
	pragma(msg, "level >= 2");
}
version=3; // ***illegal***

true: newLevel >= highestQuery
false: highestQuery >= currentLevel



version=1;
version(2){
	pragma(msg, "level >= 2");
}
version=0; // ***illegal***

false: newLevel >= highestQuery
true: highestQuery >= currentLevel


Thomas
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (MingW32)

iD8DBQFChkio3w+/yD4P9tIRAoCvAKCwFPw5SrEpAc0XXaq6yyU76Gy1GQCgpMrs
M0sLKyNIqWi1iIEQ0G3qizE=
=r5pg
-----END PGP SIGNATURE-----

"Walter" <newshound digitalmars.com> writes:

Yeah, I was thinking that too. You're right.