const char *utf_validateString(unsigned char *s, size_t len)
{
size_t idx;
const char *err = NULL;
dchar_t dc;
for (idx = 0; idx < len; )
{
err = utf_decodeChar(s, len, &idx, &dc);
if (err)
break;
}
return err;
}
void PragmaDeclaration::semantic(Scope *sc)
{ // Should be merged with PragmaStatement
#if IN_LLVM
Pragma llvm_internal = LLVMnone;
std::string arg1str;
#endif
//printf("\tPragmaDeclaration::semantic '%s'\n",toChars());
if (ident == Id::msg)
{
if (args)
{
for (size_t i = 0; i < args->dim; i++)
{
Expression *e = (*args)[i];
sc = sc->startCTFE();
e = e->semantic(sc);
e = resolveProperties(sc, e);
sc = sc->endCTFE();
// pragma(msg) is allowed to contain types as well as expressions
e = ctfeInterpretForPragmaMsg(e);
if (e->op == TOKerror)
{ errorSupplemental(loc, "while evaluating pragma(msg, %s)", (*args)[i]->toChars());
return;
}
StringExp *se = e->toString();
if (se)
{
se = se->toUTF8(sc);
fprintf(stderr, "%.*s", (int)se->len, (char *)se->string);
}
else
fprintf(stderr, "%s", e->toChars());
}
fprintf(stderr, "\n");
}
goto Lnodecl;
}
else if (ident == Id::lib)
{
if (!args || args->dim != 1)
error("string expected for library name");
else
{
Expression *e = (*args)[0];
sc = sc->startCTFE();
e = e->semantic(sc);
e = resolveProperties(sc, e);
sc = sc->endCTFE();
e = e->ctfeInterpret();
(*args)[0] = e;
if (e->op == TOKerror)
goto Lnodecl;
StringExp *se = e->toString();
if (!se)
error("string expected for library name, not '%s'", e->toChars());
else
{
char *name = (char *)mem.malloc(se->len + 1);
memcpy(name, se->string, se->len);
name[se->len] = 0;
if (global.params.verbose)
fprintf(global.stdmsg, "library %s\n", name);
if (global.params.moduleDeps && !global.params.moduleDepsFile)
{
OutBuffer *ob = global.params.moduleDeps;
Module *imod = sc->instantiatingModule();
ob->writestring("depsLib ");
ob->writestring(imod->toPrettyChars());
ob->writestring(" (");
escapePath(ob, imod->srcfile->toChars());
ob->writestring(") : ");
ob->writestring((char *) name);
ob->writenl();
}
mem.free(name);
}
}
goto Lnodecl;
}
else if (ident == Id::startaddress)
{
if (!args || args->dim != 1)
error("function name expected for start address");
else
{
/* Bugzilla 11980:
* resolveProperties and ctfeInterpret call are not necessary.
*/
Expression *e = (*args)[0];
sc = sc->startCTFE();
e = e->semantic(sc);
sc = sc->endCTFE();
(*args)[0] = e;
Dsymbol *sa = getDsymbol(e);
if (!sa || !sa->isFuncDeclaration())
error("function name expected for start address, not '%s'", e->toChars());
}
goto Lnodecl;
}
else if (ident == Id::mangle)
{
if (!args || args->dim != 1)
error("string expected for mangled name");
else
{
Expression *e = (*args)[0];
e = e->semantic(sc);
e = e->ctfeInterpret();
(*args)[0] = e;
if (e->op == TOKerror)
goto Lnodecl;
StringExp *se = e->toString();
if (!se)
{
error("string expected for mangled name, not '%s'", e->toChars());
return;
}
if (!se->len)
error("zero-length string not allowed for mangled name");
if (se->sz != 1)
error("mangled name characters can only be of type char");
#if 1
/* Note: D language specification should not have any assumption about backend
* implementation. Ideally pragma(mangle) can accept a string of any content.
*
* Therefore, this validation is compiler implementation specific.
*/
for (size_t i = 0; i < se->len; )
{
utf8_t *p = (utf8_t *)se->string;
dchar_t c = p[i];
if (c < 0x80)
{
if (c >= 'A' && c <= 'Z' ||
c >= 'a' && c <= 'z' ||
c >= '0' && c <= '9' ||
c != 0 && strchr("$%().:?@[]_", c))
{
++i;
continue;
}
else
{
error("char 0x%02x not allowed in mangled name", c);
break;
}
}
if (const char* msg = utf_decodeChar((utf8_t *)se->string, se->len, &i, &c))
{
error("%s", msg);
break;
}
if (!isUniAlpha(c))
{
error("char 0x%04x not allowed in mangled name", c);
break;
}
}
#endif
}
}
#if IN_LLVM
else if ((llvm_internal = DtoGetPragma(sc, this, arg1str)) != LLVMnone)
{
// nothing to do anymore
}
#endif
else if (global.params.ignoreUnsupportedPragmas)
{
if (global.params.verbose)
{
/* Print unrecognized pragmas
*/
fprintf(global.stdmsg, "pragma %s", ident->toChars());
if (args)
{
for (size_t i = 0; i < args->dim; i++)
{
Expression *e = (*args)[i];
#if IN_LLVM
// ignore errors in ignored pragmas.
global.gag++;
unsigned errors_save = global.errors;
#endif
sc = sc->startCTFE();
e = e->semantic(sc);
e = resolveProperties(sc, e);
sc = sc->endCTFE();
e = e->ctfeInterpret();
if (i == 0)
fprintf(global.stdmsg, " (");
else
fprintf(global.stdmsg, ",");
fprintf(global.stdmsg, "%s", e->toChars());
#if IN_LLVM
// restore error state.
global.gag--;
global.errors = errors_save;
#endif
}
if (args->dim)
fprintf(global.stdmsg, ")");
}
fprintf(global.stdmsg, "\n");
}
}
else
error("unrecognized pragma(%s)", ident->toChars());
Ldecl:
if (decl)
{
for (size_t i = 0; i < decl->dim; i++)
{
Dsymbol *s = (*decl)[i];
s->semantic(sc);
if (ident == Id::mangle)
{
StringExp *e = (*args)[0]->toString();
char *name = (char *)mem.malloc(e->len + 1);
memcpy(name, e->string, e->len);
name[e->len] = 0;
unsigned cnt = setMangleOverride(s, name);
if (cnt > 1)
error("can only apply to a single declaration");
}
#if IN_LLVM
else
{
DtoCheckPragma(this, s, llvm_internal, arg1str);
}
#endif
}
}
return;
Lnodecl:
if (decl)
{
error("pragma is missing closing ';'");
goto Ldecl; // do them anyway, to avoid segfaults.
}
}
Expression *StringExp::castTo(Scope *sc, Type *t)
{
/* This follows copy-on-write; any changes to 'this'
* will result in a copy.
* The this->string member is considered immutable.
*/
StringExp *se;
Type *tb;
int copied = 0;
//printf("StringExp::castTo(t = %s), '%s' committed = %d\n", t->toChars(), toChars(), committed);
if (!committed && t->ty == Tpointer && t->nextOf()->ty == Tvoid)
{
error("cannot convert string literal to void*");
return new ErrorExp();
}
se = this;
if (!committed)
{ se = (StringExp *)copy();
se->committed = 1;
copied = 1;
}
if (type == t)
{
return se;
}
tb = t->toBasetype();
//printf("\ttype = %s\n", type->toChars());
if (tb->ty == Tdelegate && type->toBasetype()->ty != Tdelegate)
return Expression::castTo(sc, t);
Type *typeb = type->toBasetype();
if (typeb == tb)
{
if (!copied)
{ se = (StringExp *)copy();
copied = 1;
}
se->type = t;
return se;
}
if (tb->ty != Tsarray && tb->ty != Tarray && tb->ty != Tpointer)
{ if (!copied)
{ se = (StringExp *)copy();
copied = 1;
}
goto Lcast;
}
if (typeb->ty != Tsarray && typeb->ty != Tarray && typeb->ty != Tpointer)
{ if (!copied)
{ se = (StringExp *)copy();
copied = 1;
}
goto Lcast;
}
if (typeb->nextOf()->size() == tb->nextOf()->size())
{
if (!copied)
{ se = (StringExp *)copy();
copied = 1;
}
if (tb->ty == Tsarray)
goto L2; // handle possible change in static array dimension
se->type = t;
return se;
}
if (committed)
goto Lcast;
#define X(tf,tt) ((tf) * 256 + (tt))
{
OutBuffer buffer;
size_t newlen = 0;
int tfty = typeb->nextOf()->toBasetype()->ty;
int ttty = tb->nextOf()->toBasetype()->ty;
switch (X(tfty, ttty))
{
case X(Tchar, Tchar):
case X(Twchar,Twchar):
case X(Tdchar,Tdchar):
break;
case X(Tchar, Twchar):
for (size_t u = 0; u < len;)
{ unsigned c;
const char *p = utf_decodeChar((unsigned char *)se->string, len, &u, &c);
if (p)
error("%s", p);
else
buffer.writeUTF16(c);
}
newlen = buffer.offset / 2;
buffer.writeUTF16(0);
goto L1;
case X(Tchar, Tdchar):
for (size_t u = 0; u < len;)
{ unsigned c;
const char *p = utf_decodeChar((unsigned char *)se->string, len, &u, &c);
if (p)
error("%s", p);
buffer.write4(c);
newlen++;
}
buffer.write4(0);
goto L1;
case X(Twchar,Tchar):
for (size_t u = 0; u < len;)
{ unsigned c;
const char *p = utf_decodeWchar((unsigned short *)se->string, len, &u, &c);
if (p)
error("%s", p);
else
buffer.writeUTF8(c);
}
newlen = buffer.offset;
buffer.writeUTF8(0);
goto L1;
case X(Twchar,Tdchar):
for (size_t u = 0; u < len;)
{ unsigned c;
const char *p = utf_decodeWchar((unsigned short *)se->string, len, &u, &c);
if (p)
error("%s", p);
buffer.write4(c);
newlen++;
}
buffer.write4(0);
goto L1;
case X(Tdchar,Tchar):
for (size_t u = 0; u < len; u++)
{
unsigned c = ((unsigned *)se->string)[u];
if (!utf_isValidDchar(c))
error("invalid UCS-32 char \\U%08x", c);
else
buffer.writeUTF8(c);
newlen++;
}
newlen = buffer.offset;
buffer.writeUTF8(0);
goto L1;
case X(Tdchar,Twchar):
for (size_t u = 0; u < len; u++)
{
unsigned c = ((unsigned *)se->string)[u];
if (!utf_isValidDchar(c))
error("invalid UCS-32 char \\U%08x", c);
else
buffer.writeUTF16(c);
newlen++;
}
newlen = buffer.offset / 2;
buffer.writeUTF16(0);
goto L1;
L1:
if (!copied)
{ se = (StringExp *)copy();
copied = 1;
}
se->string = buffer.extractData();
se->len = newlen;
se->sz = tb->nextOf()->size();
break;
default:
assert(typeb->nextOf()->size() != tb->nextOf()->size());
goto Lcast;
}
}
#undef X
L2:
assert(copied);
// See if need to truncate or extend the literal
if (tb->ty == Tsarray)
{
int dim2 = ((TypeSArray *)tb)->dim->toInteger();
//printf("dim from = %d, to = %d\n", se->len, dim2);
// Changing dimensions
if (dim2 != se->len)
{
// Copy when changing the string literal
unsigned newsz = se->sz;
void *s;
int d;
d = (dim2 < se->len) ? dim2 : se->len;
s = (unsigned char *)mem.malloc((dim2 + 1) * newsz);
memcpy(s, se->string, d * newsz);
// Extend with 0, add terminating 0
memset((char *)s + d * newsz, 0, (dim2 + 1 - d) * newsz);
se->string = s;
se->len = dim2;
}
}
se->type = t;
return se;
Lcast:
Expression *e = new CastExp(loc, se, t);
e->type = t; // so semantic() won't be run on e
return e;
}