translate-c fails silently when source code initializes a struct to zero

[kavika13]

If you have a C function that initializes a struct instance to { 0 } and run translate-c on it, then zig skips codegen for that function, and doesn't output any error messages.

This was confusing to me because the function didn't output at all. I only figured it out when I did a "binary chop" on the input file to figure out which particular line was keeping the function from being output.

Ideally this would just work, but if it fails to work it would be nice if a good error message was output.

Repro steps:

1. Create a C example file and save it as main.c:

typedef struct {
    int abc;
} SomeStruct;

int main() {
    SomeStruct blah;
    return 0;
}

2. zig translate-c main.c > main.zig
3. Change the line SomeStruct blah; to SomeStruct blah = { 0 }; inside main.c and save it
4. zig translate-c main.c > main.zig

Expected:
The implementation of fn main() exists in main.zig after both step 2 and step 4
or
An error message is displayed after step 4 that says translate-c cannot cope

Actual:
The implementation of fn main() exists in main.zig after step 2, but disappears after step 4. No error message is displayed.

[andrewrk]

Are you aware of the --verbose-cimport flag? This will probably shed some light.

One of the improvements that the self-hosted translate-c implementation (#1964) does is make these comments rather than usually hidden warnings.

[kavika13]

From --verbose-cimport, seems this is semi-intentional since the work to emit this code just isn't done yet, and there's no compiler-side work around in place:

main.c:6:23: warning: TODO handle C InitListExprClass
main.c:5:5: warning: unable to translate function

It does raise the question: should "warnings" be hidden if you don't specify verbose?