http://d.puremagic.com/issues/show_bug.cgi?id=3374
Don clugd...@yahoo.com.au changed:
What|Removed |Added
Keywords||patch
Platform|Other |All
Version|unspecified |2.033
OS/Version|Linux |All
--- Comment #1 from Don clugd...@yahoo.com.au 2009-10-09 00:32:48 PDT ---
The ICE only occurs if it's inside a function. BTW if you change 'auto' into
enum, and
'assert' into 'static assert', it also works with the patch in place.
void bug3374()
{
auto famousNamedConstants =
[ pi : 3.14, e : 2.71, moving sofa : 2.22 ];
assert(famousNamedConstants[e]==2.71);
}
There are two parts to this patch. The first part, in
ArrayInitializer::toExpression() is just for the ICE.
The ICE occurs because it's not checking for an ERROR type. (was NULL in D1).
This changes it from
ice-on-valid-code into rejects-valid. To completely fix the bug requires the
second part.
The second part, in ArrayInitializer::inferType(Scope *sc), is for the AA type
inference.
But, I had to do a semantic on the expression. Not sure if that's OK. This part
of the patch is for D1 as well.
PATCH against DMD 2.033. init.c, ArrayInitializer::toExpression(), line 407.
Index: init.c
===
--- init.c(revision 201)
+++ init.c(working copy)
@@ -404,6 +404,7 @@
Type *t = NULL;
if (type)
{
+if (type == Type::terror) goto Lno;
t = type-toBasetype();
switch (t-ty)
{
@@ -542,8 +543,18 @@
return type;
Lno:
-error(loc, cannot infer type from this array initializer);
-return Type::terror;
+Initializer *iz = (Initializer *)value.data[0];
+Expression *indexinit = (Expression *)index.data[0];
+if (iz indexinit)
+{ Type *t = iz-inferType(sc);
+indexinit = indexinit-semantic(sc);
+Type *indext = indexinit-type;
+t = new TypeAArray(t, indext);
+t = t-semantic(loc, sc);
+type = t;
+}
+return type;
+
}
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
--- You are receiving this mail because: ---