GCC - shouldn't a warning be issued when assig

问题:

I've recently set up a MinGW + MSYS environment on my laptop to check how things are with Netbeans C/C++ support. Everything seems to work fine, however, during my testing I have noticed a difference between GCC and Microsoft's cl.exe compiler.

Here's a sample program:

#include <stdio.h>
#include <stdlib.h>
#include <limits.h>

int main(void) {
    int i_max = INT_MAX;
    char c_max = CHAR_MAX, c;

    c = i_max;
    printf("i_max: %d, c_max: %d, c: %d\n", i_max, c_max, c);
    return EXIT_SUCCESS;
}

The output is:

i_max: 2147483647, c_max: 127, c: -1

As you can see in the code above, I assign an int to a char. Shouldn't this produce a warning that a possible data loss may occur? Microsoft's compiler (which I have configured to be very strict) does issue the warning while GCC doesn't.

Here are the GCC options I use:

-g -Werror -ansi -pedantic -Wall -Wextra

Am I missing some GCC option to make the compile time checks even stricter?

回答1:

You're looking for

-Wconversion

You'd have to ask a gcc developer for the specific reasons why some warnings aren't included in -Wall or -Wextra.

Anyway, these are the flags I use:

-Wall -Wextra -Wmissing-prototypes -Wmissing-declarations -Wshadow
-Wpointer-arith -Wcast-align -Wwrite-strings -Wredundant-decls -Wnested-externs
-Winline -Wno-long-long -Wconversion -Wstrict-prototypes

As other's have pointed out, the behaviour of -Wconversion changed with version 4.3 - the old warning about prototypes forcing a type conversion is now available as -Wtraditional-conversion.

回答2:

I didnt get a warning/error with -Wconversion either. However, if you give a pass with something like splint you get three warnings:

file.c: (in function main)
file.c:9:5: Assignment of int to char: c = i_max
  To make char and int types equivalent, use +charint.
file.c:10:52: Format argument 2 to printf (%d) expects int gets char: c_max
   file.c:10:32: Corresponding format code
file.c:10:59: Format argument 3 to printf (%d) expects int gets char: c
   file.c:10:39: Corresponding format code

Finished checking --- 3 code warnings

If you are serious about catching all errors you should use more than one tool.

回答3:

Theres a little nuance in your question that is not immediately obvious to me from the way it is worded.

If you think GCC (specifically GCC) should issue a warning here, then some compiler options might possibly help (see other replies).

If you think any compiler should issue a warning here (and I seem to read that sentiment in your question), then... well, "warnings" are not in any way mandatory or even de-facto unified. There's no "should" here. Assigning a larger-typed integer value to a smaller type without an explicit cast is pefectly legal in C. Overflow on conversion, produces implementation-defined behavior (it is not even UB :))

回答4:

-Wall doesn't quite mean -Wall, -Wextra is already under fire for being a little overly pedantic.

As Christoph said, you are looking for -Wconversion. Its good to really dig in to what -Wall and -Wextra actually turn on, and just specify the -W flags you want in your make file, especially if treating warnings as errors.

回答5:

In C, assigning an int to a char is legal.

As it's legal (but possibly dodgy) the different compiler vendors do different things when they encounter this code.

I guess MS is just being extra pedantic while the GCC guys have decided that it's not even worth a warning.

回答6:

I think it falls under the "Usual arithmetic conversions" (6.3.1.8) or "integer promotion rules" (5.1.2.3 (?)), but I can't find the specific text that says the behaviour you're seeing is expected.