[WIP] Merge DMD r251: bugzilla 111 (appending a dchar to a char[]) This patch needs some work in the code generation, because of the runtime changes (functions "_d_arrayappendcd" and "_d_arrayappendwd" are added). This doesn't affect existing code though, it just makes with patch a little useless, because something like this: char [] s; s ~= '\u6211'; That failed to compile with a nice error message previously to this change, now fails with and ugly error message (a failed assertion). Apparently there is a regression introduced by this patch too, when compiling Dil I get this assertion message: ldc: /home/luca/tesis/ldc/gen/statements.cpp:132: virtual void ReturnStatement::toIR(IRState*): Assertion `p->topfunc()->getReturnType() == llvm::Type::getVoidTy(gIR->context())' failed. 0 ldc 0x08a91628 Thank god we have bisecting capabilities in VCSs now ;) --- dmd/expression.c | 47 +++++++++++++++++++++++++++++++++++++++++------ 1 files changed, 41 insertions(+), 6 deletions(-)

#include "mars.h"
#include "mtype.h"
#include "expression.h"

#include "gen/warnings.h"

void warnInvalidPrintfCall(Loc loc, Expression* arguments, size_t nargs)
{
    Expression* arg = arguments;

    // make sure first argument is a string literal, or we can't do much
    // TODO make it smarter ?
    if (arg->op != TOKstring)
        return; // assume valid

    StringExp* strexp = (StringExp*)arg;

    // not wchar or dhar
    if (strexp->sz != 1)
    {
        warning(loc, "printf does not support wchar and dchar strings");
        return;
    }

#if 0
    // check the format string
    const char* str = (char*)strexp->string;
    for (size_t i = 0; i < strexp->len; ++i)
    {
        // TODO
    }
#endif
}