Good Morning,
I have come across something that I feel a shortcoming in the warnings
that gcc issues when compiling some C++ code. Calling it a bug seems a
bit harsh so I thought I would ask this mailing list of their opinion
before raising an issue. When I use a template to perform some sscanf
work on a string I loose some of the -Wformat warnings about the
expected argument types.
Here is a set of C++ code that demonstrates the warnings occurring for
sscanf and not happening for the template.
-----
#include <stdio.h>
template<class...Ts>
bool message_scanner(const char *message, const char *format, Ts
const&...ts)
{
size_t num_scanned(sscanf(message,format,ts...));
bool got_them_all(num_scanned == sizeof...(ts));
return got_them_all;
}
int main ()
{
bool correct(false);
short int anInt(0);
char aChar('c');
float aFloat(0.0);
char* message = "EXAMPLE001:01,02";
// No warnings are given about expected argument types for the
template
if(message_scanner(message,"<EXAMPLE%x:%x,%i",&anInt,&aChar,&aFloat))
{
correct = true;
}
// Warnings are given about expected argument types
if(2==sscanf(message,"<EXAMPLE%x:%x,%i",&anInt,&aChar,&aFloat))
{
correct = true;
}
return correct?0:1;
}
-----
Perhaps I am doing something wrong with the way I am handling this
template. There is probably a macro equivalent or some other way this
could be handled and retain the warnings but I found it interesting they
were lost here. I assume that is because templates are handled not by
code substitution but at a later stage so there is never a sscanf
function produced for GCC to check.
When I compile the above with GCC 4.8.4 I get:
$ gcc sscanf_template_warnings_example.cpp
sscanf_template_warnings_example.cpp:3:15: warning: variadic templates
only available with -std=c++11 or -std=gnu++11 [enabled by default]
template<class...Ts>
^
sscanf_template_warnings_example.cpp:4:75: warning: variadic templates
only available with -std=c++11 or -std=gnu++11 [enabled by default]
bool message_scanner(const char *message, const char *format, Ts
const&...ts)
^
sscanf_template_warnings_example.cpp: In function ‘bool
message_scanner(const char*, const char*, const Ts& ...)’:
sscanf_template_warnings_example.cpp:7:42: warning: variadic templates
only available with -std=c++11 or -std=gnu++11 [enabled by default]
bool got_them_all(num_scanned == sizeof...(ts));
^
sscanf_template_warnings_example.cpp: In function ‘int main()’:
sscanf_template_warnings_example.cpp:18:19: warning: deprecated
conversion from string constant to ‘char*’ [-Wwrite-strings]
char* message = "EXAMPLE001:01,02";
^
sscanf_template_warnings_example.cpp:27:64: warning: format ‘%x’ expects
argument of type ‘unsigned int*’, but argument 3 has type ‘short
int*’ [-Wformat=]
if(2==sscanf(message,"<EXAMPLE%x:%x,%i",&anInt,&aChar,&aFloat))
^
sscanf_template_warnings_example.cpp:27:64: warning: format ‘%x’ expects
argument of type ‘unsigned int*’, but argument 4 has type
‘char*’ [-Wformat=]
sscanf_template_warnings_example.cpp:27:64: warning: format ‘%i’ expects
argument of type ‘int*’, but argument 5 has type ‘float*’ [-Wformat=]
Note that all the warnings are about line 27, the sscanf line but none
are given about the templated version.
Do people think this is worthy of me raising a bug?
Is there a better way I could be handling the sscanf work without making
it so that the developer has to correct tally up the number of items
that should be scanned? There are a couple of dozen items in some of
the strings I must parse, so the odds of a mistake now or in a future
update are high.
--
Thomas Thorne <tafthorne@xxxxxxxxxxxxxx>
Attachment:
signature.asc
Description: This is a digitally signed message part
