Re: unsigned < 0
Iain Buclaw:
> You also need to watch out for code like this too:
> if (T.min < 0) { ...
> As that could possibly trigger off unsigned < 0 warnings too.
I was not talking about warnings. I was talking about changing the D language,
turning that into a _error_ if x is unsigned.
Bye,
bearophile
Re: unsigned < 0
== Quote from bearophile (bearophileh...@lycos.com)'s article
> > To avoid troubles in generic code you need a little workaround:
> > if (__traits(isUnsigned, x) || x >= 0) { ...
> That's not good enough yet. The first part of the test needs to be done in a
static if.
> Bye,
> bearophile
You also need to watch out for code like this too:
if (T.min < 0) { ...
As that could possibly trigger off unsigned < 0 warnings too.
Regards
Iain
Re: unsigned < 0
> To avoid troubles in generic code you need a little workaround:
> if (__traits(isUnsigned, x) || x >= 0) { ...
That's not good enough yet. The first part of the test needs to be done in a
static if.
Bye,
bearophile
Re: unsigned < 0
> Do you like this little change in the D2 language?
To avoid troubles in generic code you need a little workaround:
if (__traits(isUnsigned, x) || x >= 0) { ...
Bye,
bearophile
unsigned < 0
Recently I have shown some bugs (and added few enhancement requests):
http://www.digitalmars.com/webnews/newsgroups.php?art_group=digitalmars.D&article_id=129069
Among those there is my suggestion to statically disallow unsigned<0
comparisons:
http://d.puremagic.com/issues/show_bug.cgi?id=5539
So that's a little change in the D2 language itself, it's not a warning as the
one GCC shows on C code.
Recently Iain Buclaw has shown some patches for DMD:
http://d.puremagic.com/issues/show_bug.cgi?id=5490
Three cases were of unsigned < 0, like:
+++ b/src/optimize.c
@@ -351,7 +351,7 @@ Expression *AddrExp::optimize(int result)
{
TypeSArray *ts = (TypeSArray *)ve->type;
dinteger_t dim = ts->dim->toInteger();
-if (index < 0 || index >= dim)
+if (index >= dim)
error("array index %jd is out of bounds [0..%jd]", index,
dim);
e = new SymOffExp(loc, ve->var, index * ts->nextOf()->size());
e->type = type;
Do you like this little change in the D2 language?
Bye,
bearophile
