string literals in struct initializers with C++

string literals in struct initializers with C++

[J.T. Conklin]

In the process of moving from the gcc-3_2-branch to the CVS head, I found 
code with string literals in struct initializers no longer compiles using
g++.  I'm not sure whether this is just a tightening of the parser or a 
bug.

The following code:

        struct foo {
                char    x[20];
                int     y;
                int     z;
        };

        struct foo foos[] = {
                { "foo", 0, 2 },
                { "bar", 1, 3 }
        };

Now results in:
        foo.cc:10: error: invalid conversion from `const char*' to `char'
        foo.cc:10: error: invalid conversion from `const char*' to `char'

When it used to compile with older versions of g++.  It continues to
compile with gcc.

        --jtc

-- 
J.T. Conklin

Re: string literals in struct initializers with C++

[Michael Ritzert]

On Tuesday 15 October 2002 19:58, J.T. Conklin wrote:
> The following code:
>
>         struct foo {
>                 char    x[20];
>                 int     y;
>                 int     z;
>         };
>
>         struct foo foos[] = {
>                 { "foo", 0, 2 },
>                 { "bar", 1, 3 }
>         };
>
> Now results in:
>         foo.cc:10: error: invalid conversion from `const char*' to `char'
>         foo.cc:10: error: invalid conversion from `const char*' to `char'

I hit the same error message during my daily build of STLport. The simplified 
code looks like this:

-bash-2.05b$ gcc -v
Reading specs from 
/home/ritzert/gcc-HEAD/install/lib/gcc-lib/i386-unknown-freebsd4.5/3.3/specs
Configured with: /home/ritzert/gcc-HEAD/gcc/configure --enable-threads=posix 
--prefix=/home/ritzert/gcc-HEAD/install
Thread model: posix
gcc version 3.3 20021014 (experimental)
-bash-2.05b$ cat x.cpp
const char x[][2] = { "Hello", "world" };

-bash-2.05b$ gcc -c x.cpp
x.cpp:1: error: invalid conversion from `const char*' to `char'
x.cpp:1: error: invalid conversion from `const char*' to `char'

It compiled fine just a week ago. (Sorry, I can't be more precise; first there 
was the anoncvs update problem, then my build machine went down...).

Michael

Re: string literals in struct initializers with C++

[J.T. Conklin]

jtc@acorntoolworks.com (J.T. Conklin) writes:
> In the process of moving from the gcc-3_2-branch to the CVS head, I found 
> code with string literals in struct initializers no longer compiles using
> g++.  I'm not sure whether this is just a tightening of the parser or a 
> bug.

FWIW, The same or similar problem appears to effect the following code:

        #include <string>

        struct bar {
                std::string s;
                int y;
                int z;
        };

        bar bar_array[] = {
                { "abc", 1, 2 },
                { "def", 2, 4 },
                { "ghi", 3, 6 },
                { "klm", 4, 8 }
        };

Which results in:

        bar.cc:14: error: conversion from `const char[4]' to non-scalar type
           `bar' requested
        bar.cc:14: error: conversion from `const char[4]' to non-scalar type
           `bar' requested
        bar.cc:14: error: conversion from `const char[4]' to non-scalar type
           `bar' requested
        bar.cc:14: error: conversion from `const char[4]' to non-scalar type
           `bar' requested

Unlike the privious example, this one can't be worked around with an
extra layer of {}'s.

And since I neglected to mention it before, this is a g++ from the CVS head,
on QNX 6.2 (the changes I submitted for QNX in August with slight tweaks as
suggested by Zack and as necessary to to adapt to the CVS head.

        --jtc

-- 
J.T. Conklin

Re: string literals in struct initializers with C++

[Mark Mitchell]

--On Tuesday, October 15, 2002 10:58:05 AM -0700 "J.T. Conklin" 
<jtc@acorntoolworks.com> wrote:

> In the process of moving from the gcc-3_2-branch to the CVS head, I found
> code with string literals in struct initializers no longer compiles using
> g++.  I'm not sure whether this is just a tightening of the parser or a
> bug.
>

A bug; I'm testing a fix as we speak.

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com