[Issue 4389] ICE(constfold.c, expression.c), or wrong code: string~=dchar in CTFE
https://issues.dlang.org/show_bug.cgi?id=4389 Andrei Alexandrescu and...@erdani.com changed: What|Removed |Added Version|D1 D2 |D2 --
[Issue 4389] ICE(constfold.c, expression.c), or wrong code: string~=dchar in CTFE
http://d.puremagic.com/issues/show_bug.cgi?id=4389 Don clugd...@yahoo.com.au changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED --- Comment #7 from Don clugd...@yahoo.com.au 2011-02-06 13:53:08 PST --- D1 fixed as well: https://github.com/D-Programming-Language/dmd/commit/2b51411e9e8b7ca3e62747ffb6f96c94089cebcc -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email --- You are receiving this mail because: ---
[Issue 4389] ICE(constfold.c, expression.c), or wrong code: string~=dchar in CTFE
http://d.puremagic.com/issues/show_bug.cgi?id=4389
--- Comment #6 from Don clugd...@yahoo.com.au 2011-01-19 23:59:47 PST ---
Fails on D1 because dchar, wchar implicitly convert to char, even if they don't
fit.
PATCH: expression.c, CatAssignExp::semantic(), line 8574. The two 'else'
clauses need to be swapped,
so that we check for an array~char before we check for implicit conversion.
{ // Append array
e2 = e2-castTo(sc, e1-type);
type = e1-type;
e = this;
}
--- SWAP THIS SECTION WITH THE NEXT ONE
else if ((tb1-ty == Tarray)
e2-implicitConvTo(tb1next)
)
{ // Append element
e2 = e2-castTo(sc, tb1next);
type = e1-type;
e = this;
}
---
else if (tb1-ty == Tarray
(tb1next-ty == Tchar || tb1next-ty == Twchar)
e2-implicitConvTo(Type::tdchar)
)
{ // Append dchar to char[] or wchar[]
e2 = e2-castTo(sc, Type::tdchar);
type = e1-type;
e = this;
/* Do not allow appending wchar to char[] because if wchar happens
* to be a surrogate pair, nothing good can result.
*/
}
---
else
{
if (tb1 != Type::terror tb2 != Type::terror)
error(cannot append type %s to type %s, tb2-toChars(),
tb1-toChars());
e = new ErrorExp();
}
ALTERNATE PATCH:
Incidentally, if dchar-char is disallowed, all DMD tests still pass. But this
would probably cause existing valid code to break, so probably the first
solution is better.
mtype.c, line 1570, MATCH TypeBasic::implicitConvTo(Type *to)
+ if ((ty == Tdchar || ty == Twchar) to-ty == Tchar)
+return MATCHnomatch;
+ if (ty == Tdchar to-ty == Twchar)
+ return MATCHnomatch;
}
else if (flags TFLAGSfloating)
{
// Disallow implicit conversion of floating point to integer
if (tob-flags TFLAGSintegral)
return MATCHnomatch;
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
--- You are receiving this mail because: ---
[Issue 4389] ICE(constfold.c, expression.c), or wrong code: string~=dchar in CTFE
http://d.puremagic.com/issues/show_bug.cgi?id=4389 Walter Bright bugzi...@digitalmars.com changed: What|Removed |Added CC||bugzi...@digitalmars.com --- Comment #4 from Walter Bright bugzi...@digitalmars.com 2010-12-27 18:54:58 PST --- The patch should work in D1, but does not. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email --- You are receiving this mail because: ---
[Issue 4389] ICE(constfold.c, expression.c), or wrong code: string~=dchar in CTFE
http://d.puremagic.com/issues/show_bug.cgi?id=4389 Walter Bright bugzi...@digitalmars.com changed: What|Removed |Added Version|D2 |D1 D2 --- Comment #5 from Walter Bright bugzi...@digitalmars.com 2010-12-27 19:27:54 PST --- http://www.dsource.org/projects/dmd/changeset/825 Leaving bug open until D1 is fixed. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email --- You are receiving this mail because: ---
[Issue 4389] ICE(constfold.c, expression.c), or wrong code: string~=dchar in CTFE
http://d.puremagic.com/issues/show_bug.cgi?id=4389
Don clugd...@yahoo.com.au changed:
What|Removed |Added
Keywords||ice-on-valid-code
Summary|CTFE: Appending dchar to|ICE(constfold.c,
|empty string creates|expression.c), or wrong
|invalid string. |code: string~=dchar in CTFE
--- Comment #1 from Don clugd...@yahoo.com.au 2010-11-23 00:56:25 PST ---
Actually constant folding of string ~= dchar is ALWAYS wrong, even if the
string is non-empty. All these tests fail. It can also cause an
ICE(constfold.c) if you attempt a further concatenation, or ICE(expression.c)
if it is mixed in (bug 3490).
It happens because string ~= dchar is treated specially in the backend, but not
in constfold.c.
-
int bug4389()
{
string s;
dchar c = '\u2348';
s ~= c;
assert(s.length==3);
dchar d = 'D';
s ~= d;
assert(s.length==4);
s = ;
s ~= c;
assert(s.length==3);
s ~= d;
assert(s.length==4);
string z;
wchar w = '\u0300';
z ~= w;
assert(z.length==2);
z = ;
z ~= w;
assert(z.length==2);
return 1;
}
static assert(bug4389());
// ICE(constfold.c)
int ice4389()
{
string s;
dchar c = '\u2348';
s ~= c;
s = s ~ xxx;
return 1;
}
static assert(ice4389());
// ICE(expression.c)
string ice4390()
{
string s;
dchar c = '`';
s ~= c;
s ~= c;
return s;
}
static assert(mixin(ice4390()) == ``);
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
--- You are receiving this mail because: ---
[Issue 4389] ICE(constfold.c, expression.c), or wrong code: string~=dchar in CTFE
http://d.puremagic.com/issues/show_bug.cgi?id=4389 --- Comment #3 from Don clugd...@yahoo.com.au 2010-11-23 12:33:23 PST --- Created an attachment (id=827) Patch against D2 svn 767 Although the patch is relatively long, it is very mundane. The patch adds these functions to utf.c (they are copied from druntime). These convert single chars from UTF32 to UTF8 or UTF16. void utf_encodeChar(unsigned char *s, dchar_t c); void utf_encodeWchar(unsigned short *s, dchar_t c); int utf_codeLengthChar(dchar_t c); int utf_codeLengthWchar(dchar_t c); Two other wrapper functions provide the interface that is actually used: int utf_codeLength(int sz, dchar_t c); void utf_encode(int sz, void *s, dchar_t c); Then, these functions are used in Cat() in constfold.c Instead of assuming a len of 1, it uses utf_codeLength; and instead of a memcpy, it uses utf_encode. The cases which are patched in Cat() are: TOKnull ~ dchar, dchar ~ TOKnull, TOKstring ~ dchar. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email --- You are receiving this mail because: ---
