Re: [ALACPP] Why won't g++ tell me I'm dumb?

[Josh Dybnis <jdybnis@xxxxxxxxx>]

Technically gcc is doing the right thing here. String literals aren't
actually constants. It's pretty obscure. I had to actually bust out the
standard on this. What it says is that string literals are static
arrays of type char, not const char. Attempting to modify a string
literal is "undefined", but it's not actually forbidden for an
implementation to support it. This is from the C standard, but I assume
that C++ retains this for compatibility.

-Josh

--- Kevin Scaldeferri <kevin@xxxxxxxxxxxxxxx> wrote:
> We had a bit of a puzzle at lunch today, trying to figure out why my 
> simple test case for my simple test function was dying a horrible 
> death.  Here's the distilled version of the code:
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> #include <string.h>
> 
> #include <boost/test/included/unit_test_framework.hpp>
> using namespace boost::unit_test_framework;
> 
> void strrev(char* s) {
>    int l = strlen(s);
>    char tmp;
>    int i;
> 
>    for ( i = 0 ; i < (l / 2)   ; ++i ) {
>      tmp = s[i];
>      s[i] = s[l - 1 - i];
>      s[l - 1 - i] = tmp;
>    }
> }
> 
> void strrevtest() {
>    char* t;
>    t = ""; strrev(t);BOOST_CHECK_EQUAL( std::strcmp(t,""), 0);
>    t = "a"; strrev(t);BOOST_CHECK_EQUAL( std::strcmp(t,"a"), 0);
>    t = "ab"; strrev(t);BOOST_CHECK_EQUAL( std::strcmp(t,"ba"), 0);
>    t = "asdf"; strrev(t);BOOST_CHECK_EQUAL( std::strcmp(t,"fdsa"),
> 0);
>    t = "asdfg"; strrev(t);BOOST_CHECK_EQUAL( std::strcmp(t,"gfdsa"),
> 0);
> }
> 
> test_suite*
> init_unit_test_suite ( int argc, char** argv) {
>    test_suite* test = BOOST_TEST_SUITE( "strrev test");
>    test->add( BOOST_TEST_CASE( &strrevtest ));
>    return test;
> }
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> This produces a memory access violation when it does the first swap. 
> 
> The reason is that I'm dumb and I was trying to modify a (const)
> string 
> literal.  Now, I can fix my test case by sticking in some strdup's,
> but 
> there was also a consensus that the C++ compiler ought to complain.  
> However, it really seems to just silently discard the const-ness, 
> despite the suggestions of the g++ man page.
> 
> Chris and I also tried with the Intel compiler, but that didn't 
> complain either.
> 
> So, is there some flag we're missing that would spit out a warning 
> about this const issue?
> 
> 
> 
> Kevin
> 
> _______________________________________________
> alacpp mailing list
> alacpp@xxxxxxxxxxx
> http://lists.ellipsis.cx/mailman/listinfo/alacpp

_______________________________________________
alacpp mailing list
alacpp@xxxxxxxxxxx
http://lists.ellipsis.cx/mailman/listinfo/alacpp