Skip to main content
Code Review

Return to Answer

added 1 character in body
Source Link
Solkar
Solkar
  • 183
  • 5

using the the new e.code() semantics yieldyields (with clang 3.4 and libc++ 1101) only the first line of

using the the new e.code() semantics yield (with clang 3.4 and libc++ 1101) only the first line of

using the the new e.code() semantics yields (with clang 3.4 and libc++ 1101) only the first line of

added 5 characters in body
Source Link
Solkar
Solkar
  • 183
  • 5

So even if one's lucklucky enough to have a std:: implementation that cares about iso §27.5.3.1.1 and actually derives ios_base::failure from system_error, the msg generated is too poor to be presented to users.

The only thing to be discussed is how the libc messages are best incorporated into wrapper classes.

So even if one's luck enough to have a std:: implementation that cares about iso §27.5.3.1.1 and actually derives ios_base::failure from system_error, the msg generated is too poor to be presented to users.

The only thing to be discussed is how the libc messages are incorporated into wrapper classes.

So even if one's lucky enough to have a std:: implementation that cares about iso §27.5.3.1.1 and actually derives ios_base::failure from system_error, the msg generated is too poor to be presented to users.

The only thing to be discussed is how the libc messages are best incorporated into wrapper classes.

Source Link
Solkar
Solkar
  • 183
  • 5

I currently don't see any seamless alternative to using errno and ::strerror;.

#include <stdexcept>
#include <system_error>
#include <string>
#include <iostream>

#include <cstring> // strerror

#include <fstream>

#if defined(_LIBCPP_VERSION) && (_LIBCPP_VERSION >= 1000)
#define HAS_IOS_BASE_FAILURE_DERIVED_FROM_SYSTEM_ERROR 1
#else
#define HAS_IOS_BASE_FAILURE_DERIVED_FROM_SYSTEM_ERROR 0
#endif

using std::cerr;
using std::cout;
using std::endl;


int main(/*int argc, char** argv*/) {
    int rv = EXIT_SUCCESS;
    errno = 0;
    
    try {
        std::ifstream ifs;
        ifs.exceptions(std::ios::badbit | std::ios::failbit);
        ifs.open("DOESN'T EXIST");

    } catch (const std::ios_base::failure& e) {
        
#if (HAS_IOS_BASE_FAILURE_DERIVED_FROM_SYSTEM_ERROR)
        //
        // e.code() is only available if the lib actually follows iso §27.5.3.1.1
        // and derives ios_base::failure from system_error
        // like e.g. libc++
        // http://llvm.org/viewvc/llvm-project/libcxx/trunk/include/ios?revision=193085&view=markup
        // (line 415)
        //
        // and not keeps on deriving it directly from runtime_error
        // like libstdc++
        // https://github.com/mirrors/gcc/blob/master/libstdc%2B%2B-v3/include/bits/ios_base.h#L209
        //
        cout << "libc++ error #"    << e.code().value()  
                << ' '              << e.code().message()
                << ',' << endl << "    ";
#endif
        cout << "libc error #"      << (rv = errno)
                << ": "             << ::strerror(errno)
                << endl;
                
        cout << "handled" << endl;
    } 
    
    return rv;
}

using the the new e.code() semantics yield (with clang 3.4 and libc++ 1101) only the first line of

libc++ error #1 unspecified iostream_category error,
    libc error #2: No such file or directory
handled

Live

So even if one's luck enough to have a std:: implementation that cares about iso §27.5.3.1.1 and actually derives ios_base::failure from system_error, the msg generated is too poor to be presented to users.

The only thing to be discussed is how the libc messages are incorporated into wrapper classes.