Bug 4941 – Built-in tuple slice boundaries are not CTFE'd

Status
RESOLVED
Resolution
FIXED
Severity
normal
Priority
P2
Component
dmd
Product
D
Version
D2
Platform
All
OS
All
Creation time
2010-09-25T21:25:00Z
Last change time
2015-06-09T05:11:40Z
Keywords
patch, rejects-valid
Assigned to
nobody
Creator
rsinfu

Attachments

IDFilenameSummaryContent-TypeSize
776tuple-slice-ctfe.patchPatch against dmd r687, CTFE on tuple slice boundariestext/plain1662

Comments

Comment #0 by rsinfu — 2010-09-25T21:25:29Z
Type tuples can't be sliced with boundaries that involve CTFE-able function call(s). The following valid code doesn't compile (both D1 & D2): -------------------- template T(_...) { alias _ T; } size_t mid(size_t n) { return n/2; } alias T!(int, int, int)[0 .. mid($)] A; // Error: Integer constant expression expected instead of mid(3u) -------------------- Patch against dmd r687, makes it sure that TypeSlice boundaries get CTFE'd. --- src/mtype.c +++ src/mtype.c @@ -7787,11 +7787,11 @@ Type *TypeSlice::semantic(Loc loc, Scope *sc) TypeTuple *tt = (TypeTuple *)tbn; lwr = semanticLength(sc, tbn, lwr); - lwr = lwr->optimize(WANTvalue); + lwr = lwr->optimize(WANTvalue | WANTinterpret); uinteger_t i1 = lwr->toUInteger(); upr = semanticLength(sc, tbn, upr); - upr = upr->optimize(WANTvalue); + upr = upr->optimize(WANTvalue | WANTinterpret); uinteger_t i2 = upr->toUInteger(); if (!(i1 <= i2 && i2 <= tt->arguments->dim))
Comment #1 by rsinfu — 2010-09-25T22:11:04Z
Created attachment 776 Patch against dmd r687, CTFE on tuple slice boundaries The attached patch obsoletes the patch in comment #0. The new one applies the same fix to other relevant parts, and fixes following additional cases: template T(_...) { alias _ T; } size_t mid(size_t n) { return n/2; } alias T!(int, int)[0 .. mid($)] A; // A alias T!(1, 2, 3)[0 .. mid($)] B; // B void main() { T!(int, int, int) values; auto slice = values[0 .. mid($)]; // C }
Comment #2 by bugzilla — 2010-09-27T21:08:58Z