[Issue 2569] New: static arrays in CTFE functions don't compile

Jan 08, 2009

d-bugmail

Jul 21, 2009

Jul 22, 2009

Jul 23, 2009

Jul 23, 2009

Jul 24, 2009

Jul 24, 2009

Jul 24, 2009

Jul 24, 2009

Jul 24, 2009

Jul 24, 2009

Sep 03, 2009

http://d.puremagic.com/issues/show_bug.cgi?id=2569 Summary: static arrays in CTFE functions don't compile Product: D Version: 1.039 Platform: PC OS/Version: Windows Status: NEW Keywords: rejects-valid Severity: normal Priority: P2 Component: DMD AssignedTo: bugzilla@digitalmars.com ReportedBy: clugdbug@yahoo.com.au int foo() { int [3] a; return 0; } static x = foo(); --- bug.d(7): Error: cannot evaluate foo() at compile time Interestingly, in D2.023, the error message is displayed twice. --

http://d.puremagic.com/issues/show_bug.cgi?id=2569 --- Comment #1 from Don <clugdbug@yahoo.com.au> 2009-07-21 00:18:03 PDT --- This isn't complicated. The only reason it doesn't work is that BinExp::interpretAssignCommon in interpret.c doesn't deal with array assignment AT ALL. Adding a trivial hack like: if (e1->op == TOKslice) { return e2; } is enough to make most cases work. I'm working on a proper patch which will deal with array literals, etc. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=2569 Don <clugdbug@yahoo.com.au> changed: What |Removed |Added ---------------------------------------------------------------------------- Keywords| |patch --- Comment #2 from Don <clugdbug@yahoo.com.au> 2009-07-22 00:36:19 PDT --- Turned out to be slightly more involved than I thought, there are a few special cases. I think I've captured them all. Extended test case: ------------------- struct Foo { int m; } int foo() { short [4] a = [2, 3, 4, 6]; double [27] b = 2.0; cfloat[2] c; auto d = [2, 3, 4, 5]; Foo[2] e = [Foo(3), Foo(2)]; // ushort [3] f = [1, 2]; // Uncomment to generate a compile-time error. d[2..6] = 4; a[1] = 7; return a[0]-3; } static assert(foo()==-1); PATCH: interpret.c, line 1464 in BinExp::interpretAssignCommon(), before checking for the other cases. ------------ // Assignment/initialization of static arrays if (e1->op == TOKslice && ((SliceExp *)e1)->e1->op==TOKvar) { SliceExp * sexp = (SliceExp *)e1; VarExp *ve = (VarExp *)(sexp->e1); VarDeclaration *v = ve->var->isVarDeclaration(); Type *t = v->type->toBasetype(); if (t->ty == Tsarray){ size_t dim = ((TypeSArray *)t)->dim->toInteger(); if (e2->op == TOKarrayliteral) { // Static array assignment from literal ArrayLiteralExp *ae = (ArrayLiteralExp *)e2; // Ensure length is the same if (ae->elements->dim != dim) { error("Array length mismatch"); return e; } v->value = ae; return ae; } if (t->nextOf()->ty == e2->type->ty) { // Static array block assignment Expressions *elements = new Expressions(); elements->setDim(dim); for (size_t i = 0; i < dim; i++) elements->data[i] = e2; ArrayLiteralExp *ae = new ArrayLiteralExp(0, elements); ae->type = v->type; v->value = ae; return e2; } } } -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=2569 Dimitar Kolev <DimitarRosenovKolev@hotmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |DimitarRosenovKolev@hotmail | |.com --- Comment #4 from Dimitar Kolev <DimitarRosenovKolev@hotmail.com> 2009-07-23 07:41:12 PDT --- (In reply to comment #3) > The patch I posted was incomplete, so I'm withdrawing it. Done properly, it should support slicing assignment, and arrays initialized to void. This is a nasty little bugger. Any idea when the pAtch will be ready? -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------

July 24, 2009

[Issue 2569] static arrays in CTFE functions don't compile

Posted by Don
in reply to d-bugmail

Permalink

Don

Posted in reply to d-bugmail

Permalink

http://d.puremagic.com/issues/show_bug.cgi?id=2569





--- Comment #5 from Don <clugdbug@yahoo.com.au>  2009-07-24 00:01:29 PDT ---
(In reply to comment #4)
> (In reply to comment #3)
> > The patch I posted was incomplete, so I'm withdrawing it. Done properly, it should support slicing assignment, and arrays initialized to void.
> 
> This is a nasty little bugger. Any idea when the patch will be ready?

The patch is now complete for array literals, and also fixes bug#1948 and bug#3205. It also gives nicer error messages when it can't compile an assignment in CTFE. There are an unbelievable number of special cases.

The only thing that's missing is that it doesn't deal with assignment from
string literals. Which means that:
char [5] s = "abc"; // fails
char [6] t = ['a', 'b', 'c']; // ok


Here's my test case. All the tests below are working on my patched DMD:

struct S {
    int x;
    char y;
}

// Functions which should fail CTFE

int badfoo(){
   S[2] c;
   c[4].x=6;  // array bounds error
   return 7;
}

int badglobal = 1;

int badfoo2(){
   S[] c;
   c[7].x=6;  // uninitialized error
   return 7;
}

int badfoo3(){
   S[2] c;
   c[badglobal].x=6;  // global index error
   return 7;
}

int badfoo4(){
   static S[2] c;
   c[0].x=6;  // Cannot access static
   return 7;
}

int badfoo5(){
   S[] c = void;
   c[0].x=6;  // c is uninitialized, and not a static array.
   return 1;
}

int badfoo6()
{
    S[] b = [S(7), S(15), S(56), S(12)];
    b[-2..4] = S(17); // exceeding (negative) array bounds
    return 1;
}

int badfoo7()
{
    S[] b = [S(7), S(15), S(56), S(12), S(67)];
    b[1..4] = [S(17), S(4)]; // slice mismatch in dynamic array
    return 1;
}

int badfoo8()
{
    S[] b;
    b[1..3] = [S(17), S(4)]; // slice assign to uninitialized dynamic array
    return 1;
}


template Compileable(int z) { bool OK;}
static assert(!is(typeof(Compileable!(badfoo()).OK)));
static assert(!is(typeof(Compileable!(badfoo2()).OK)));
static assert(!is(typeof(Compileable!(badfoo3()).OK)));
static assert(!is(typeof(Compileable!(badfoo4()).OK)));
static assert(!is(typeof(Compileable!(badfoo5()).OK)));
static assert(!is(typeof(Compileable!(badfoo6()).OK)));
static assert(!is(typeof(Compileable!(badfoo7()).OK)));
static assert(!is(typeof(Compileable!(badfoo8()).OK)));

// Functions which should pass CTFE

int goodfoo1()
{
   int[8] w;    // use static array in CTFE
   w[]=7;       // full slice assign
   w[$-1]=538;  // use of $ in index assignment
   assert(w[6]==7);
   return w[7];
}
static assert(goodfoo1()==538);

int goodfoo2()
{
   S[4] w = S(101);  // Block-initialize array of structs
   w[$-2].x = 917; // use $ in index member assignment
   w[$-2].y = 58; // this must not clobber the prev assignment
   return w[2].x; // check we got the correct one
}
static assert(goodfoo2()==917);

int goodfoo3()
{
   S[4] w = void; // uninitialized array of structs
   w[$-2].x = 217; // initialize one member
   return w[2].x;
}
static assert(goodfoo3()==217);

int goodfoo4()
{
   S[4] b = [S(7), S(15), S(56), S(12)]; // assign from array literal
   assert(b[3]==S(12));
   return b[2].x-55;
}
static assert(goodfoo4()==1);

int goodfoo5()
{
    S[4] b = [S(7), S(15), S(56), S(12)];
    b[0..2] = [S(2),S(6)]; // slice assignment from array literal
    assert(b[3]==S(12));
    assert(b[1]==S(6));
    return b[0].x;
}
static assert(goodfoo5()==2);
static assert(goodfoo5()==2); // check for memory corruption

int goodfoo6()
{
    S[6] b = void;
    b[2..5] = [S(2),S(6), S(17)]; // slice assign to uninitialized var
    assert(b[4]==S(17));
    return b[3].x;
}
static assert(goodfoo6()==6);

int goodfoo7()
{
    S[8] b = void;
    b[2..5] = S(217); // slice assign to uninitialized var
    assert(b[4]==S(217));
    return b[3].x;
}
static assert(goodfoo7()==217);

int goodfoo8()
{
    S[] b = [S(7), S(15), S(56), S(12), S(67)];
    b[2..4] = S(17); // dynamic array block slice assign
    assert(b[3]==S(17));
    assert(b[4]==S(67));
    return b[0].x;
}
static assert(goodfoo8()==7);

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=2569 --- Comment #6 from Dimitar Kolev <DimitarRosenovKolev@hotmail.com> 2009-07-24 01:50:38 PDT --- Thanks great work. You got a beer from me. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=2569 --- Comment #7 from Don <clugdbug@yahoo.com.au> 2009-07-24 02:19:04 PDT --- Created an attachment (id=434) --> (http://d.puremagic.com/issues/attachment.cgi?id=434) Patch against DMD1.046. Works for DMD2.031 as well. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=2569 Max Samukha <samukha@voliacable.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |samukha@voliacable.com --- Comment #9 from Max Samukha <samukha@voliacable.com> 2009-07-24 03:14:39 PDT --- Thanks a lot! -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------

Forums