"Joseph S. Myers" <[EMAIL PROTECTED]> writes:
| On Mon, 6 Feb 2006, Gabriel Dos Reis wrote:
|
| > Hi,
| >
| > One of the thing I've learnt over the week-end is that the option
| > code confusion
| >
| > http://gcc.gnu.org/ml/gcc-patches/2006-02/msg00326.html
| >
| > is partly caused by a conflicting interests between -fconst-strings
| > and -Wwrite-strings. The former has been deprecated for a long time
| > now. I would like to remove it. Objections?
|
| Note that the variable flag_const_strings (mentioned in your Subject:) and
| the option -fconst-strings (mentioned in your message body) are two
| different things - which of course is part of the confusion.
One aspect of the confusion is the following, from c-opts.c:
c-opts.c:228: flag_const_strings = c_dialect_cxx ();
and
case OPT_Wwrite_strings:
if (!c_dialect_cxx ())
flag_const_strings = value;
else
warn_write_strings = value;
break;
| The variable flag_const_strings is set by -Wwrite-strings for C. If you
| remove the variable then the c-common.c logic for building string
| constants will need to change
|
| if (flag_const_strings)
|
| to
|
| if (c_dialect_cxx () || warn_write_strings)
Yes, though I'll also reiterate my objection that a warning flag
should not be changing types. The C front-end should be able to warn
about potentially overwritten string literals without having to change
the type of the object -- the C++ front-end does that, by watching out
for implicit conversion from literal string to either char* ro wchar_t.
| The option -fno-const-strings (C++ only) could go away without removing
| the variable, if we no longer want to support that nonstandard dialect of
| C++.
Yes.
-- Gaby
