I just reopened 190 and updated 4753 with more details.  The fix for the former (one from the current development cycle, ie not yet released) caused the regression filed in the latter.  I'm not particularly familiar with the semantic passes of the compiler.  The symptoms here is that the new code is sending dmd into a recursive loop:

repeat in groups of 4 for frames 0..93588:
#93588 0x0813c0f4 in TypeTypeof::semantic (this=0x820e4c8, loc=..., sc=0x8242328) at mtype.c:5961
#93589 0x080da945 in VarExp::semantic (this=0x82424b0, sc=0x8242328) at expression.c:4219
#93590 0x080d51cc in DsymbolExp::semantic (this=0x8242488, sc=0x8242328) at expression.c:2312
#93591 0x080d4a08 in IdentifierExp::semantic (this=0x820e4a0, sc=0x8242328) at expression.c:2152

the end of the stack:
#93592 0x0813c0f4 in TypeTypeof::semantic (this=0x820e4c8, loc=..., sc=0x8242328) at mtype.c:5961
#93593 0x080af74d in VarDeclaration::semantic (this=0x82423b8, sc=0x8242328) at declaration.c:779
#93594 0x0817ef0c in TemplateValueParameter::semantic (this=0x820e528, sc=0x8242328) at template.c:3142
#93595 0x081790b6 in TemplateDeclaration::semantic (this=0x820e6a0, sc=0x820e9f8) at template.c:470
#93596 0x0812d66d in Module::semantic (this=0x820e068) at module.c:787
#93597 0x0812ac0b in main (argc=11, argv=0x82044d0) at mars.c:1168

Yes.. 93597 is how many frames it takes to blow up my stack. :)

Later,
Brad

The other known regression in the current development version is bug 4752. It can fixed by adding to parse.c, line 6292:

    precedence[TOKdotti] = PREC_primary;
    precedence[TOKdot] = PREC_primary;
+    precedence[TOKdottd] = PREC_primary;

Regression bug 4738 seems to be essentially the same problem, but I can't reproduce it with the current revision.

On 30 August 2010 09:49, Brad Roberts <braddr at puremagic.com> wrote:
> I just reopened 190 and updated 4753 with more details. ?The fix for the former (one from the current development cycle, ie not yet released) caused the regression filed in the latter.