[Bug c/50645] New: old issue - deprecated conversion from string to char*

[Bug c/50645] New: old issue - deprecated conversion from string to char*

[mike.c at rocketime dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50645

             Bug #: 50645
           Summary: old issue - deprecated conversion from string to char*
    Classification: Unclassified
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: mike.c@rocketime.com


While I know this is an old issue, I port a lot of code, so this keeps coming
up:  (I know u can change the warning level).

If I declare:

const char    *DowNames[] = {
        "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun",
        NULL
    };

as just char* instead of const char*, YES I WOULD LIKE TO GET A WARNING.

HOWEVER, if I have a function:

myfunction(char *buf) {}

I'd like to be able to call it like this 

myfunction("hello world") 

... WITH NO 'deprecated conversion from string to char*' WARNING

because char* as a parameter should be able to take a string as an argument --
this makes sense.

It's annoying to have to cast (char*)"hello world" every place in code, as
programmers always used to use, simply "hello world". (i know an option is to
change the warning level -- but that has other issues.)

PLEASE UNDEPRECATE!  (Keep things simple, keep backward compatibility, as the
old-school way is often best, designed that way for simplicity!)

[Bug c/50645] old issue - deprecated conversion from string to char*

[mike.c at rocketime dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50645

--- Comment #1 from mike.c at rocketime dot com 2011-10-07 00:35:20 UTC ---
The quick fix:

#pragma GCC diagnostic ignored "-Wwrite-strings"

... of course eliminates the warnings -- but logically, you DO WANT A WARNING
when you should be declaring

const char* (and you forget const)

but LOGICALLY you don't want I warning when passing a string literal to a char*
parameter in a function.

[Bug c/50645] old issue - deprecated conversion from string to char*

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50645

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-10-07 00:37:40 UTC ---
IIRC it was the C standard which declared it as deprecated.

[Bug c/50645] old issue - deprecated conversion from string to char*

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50645

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> 2011-10-07 00:41:25 UTC ---
(In reply to comment #0)
> 
> If I declare:
> 
> const char    *DowNames[] = {
>         "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun",
>         NULL
>     };
> 
> as just char* instead of const char*, YES I WOULD LIKE TO GET A WARNING.

So you want a warning if non-const pointers point to const data. Makes sense.

> HOWEVER, if I have a function:
> 
> myfunction(char *buf) {}
> 
> I'd like to be able to call it like this 
> 
> myfunction("hello world") 
> 
> ... WITH NO 'deprecated conversion from string to char*' WARNING

So you don't want a warning if a non-const pointer points to const data.
Huh?  How is that different from the earlier case.

> because char* as a parameter should be able to take a string as an argument --
> this makes sense.

Why?  If a function takes char* it means it can change the contents of the char
or chars at that address, why is a warning not appropriate if you call it with
an argument that might cause a runtime error if the function changes the data?

> It's annoying to have to cast (char*)"hello world" every place in code, as
> programmers always used to use, simply "hello world". (i know an option is to
> change the warning level -- but that has other issues.)

Or, if your function really doesn't change the data, you could always fix its
signature to be myfunction(const char*).

Or use C, instead of C++ (you do realise the warning you're complaining about
is a C++ warning, not a C one, right?)

> PLEASE UNDEPRECATE!  (Keep things simple, keep backward compatibility, as the
> old-school way is often best, designed that way for simplicity!)

That conversion is deprecated by the C++ standard, GCC cannot un-deprecate it.
Talk to the C++ committee (who will tell you to stop living in the 1970s and
get over it, type safety is a good thing.)

[Bug c/50645] old issue - deprecated conversion from string to char*

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50645

--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> 2011-10-07 00:44:12 UTC ---
(In reply to comment #1)
> but LOGICALLY you don't want I warning when passing a string literal to a char*
> parameter in a function.

Yes you do. A string literal is immutable. A function taking a non-const char*
implies it might modify the characters pointed to.  Modifying immutable data is
bad, mmmkay?

If your function won't change the data, change its signature to const char*.

[Bug c/50645] old issue - deprecated conversion from string to char*

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50645

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2011-10-07
     Ever Confirmed|0                           |1

--- Comment #5 from Jonathan Wakely <redi at gcc dot gnu.org> 2011-10-07 01:10:45 UTC ---
two points:

1) the component for this bug should surely be c++, since the "deprecated
conversion" warning is only issued by the C++ front end, the C front end says

passing argument 1 of ‘myfunction’ discards ‘const’ qualifier from pointer
target type

and even then, only with -Wwrite-strings, which is not enabled by -Wall -Wextra

2) unlike the PR's summary, the warning says "deprecated conversion from string
CONSTANT to 'char*'" (emphasis mine) which should be a pretty big clue that
it's warning you about a real problem.

I can't see any good reason this PR isn't INVALID, but will go for WAITING in
case you have a better case to make for changing the current behaviour, which
is entirely consistent with the C++ standard.