cstruct initialisation dangerous

[skaller]

Currently I think cstructs are initialised like structs, with a reinterpret cast. Not sure atm. It is certainly the case if the fields are listed out out of order, the result corrupts memory. Memberwise initialisation should be used to ensure the correct members are initialised.

i think part of the reason is C89 provides no way to initialise a struct in an expression, which Felix has to do. However I think C++ now does and I think C99 designated initialisers could also be used possibly wrapped inside a C++ lambda, although exactly how a copy can be avoided I'm not sure.

Binding is complicated because it is split up in multiple places.

[skaller]

#include <stdio.h>

struct X {int a; double b; };
void f (X const &x) { printf ("a=%d b=%f\n",x.a,x.b); }
int main() {
  f( X{.a=1,.b=43.0} ); 
}

correct method

[skaller]

Now, test case:

requires header "struct X { double b; int a; };";
cstruct X { int a; double b; };
var x = X (a=1,b=2.3);
println$ x._strr;

works correctly, generating this:

     ptf-> x  = X{.a=1, .b=2.3}/* apply cstruct*/; //init

Notice the C struct decl is out of order with the cstruct decl.