Replacing char* with const char*

[ext]

BPO	1772673
Nosy	@tim-one, @brettcannon, @amauryfa, @ncoghlan, @abalkin, @pitrou, @duggelz, @ned-deily, @ericsnowcurrently, @serhiy-storchaka
Dependencies	bpo-9369: const char* for PyObject_CallMethod and PyObject_CallFunction bpo-1699259: replacing char* with const char* in sysmodule.c/.h
Files	const_char.patch: Patch to replace char* with const char* in a few functions const_char_2.patch const_char_3.patch

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/serhiy-storchaka'
closed_at = <Date 2013-10-19.18:05:11.709>
created_at = <Date 2007-08-12.16:43:07.000>
labels = ['interpreter-core', 'type-feature']
title = 'Replacing char* with const char*'
updated_at = <Date 2014-09-28.08:29:45.063>
user = 'https://bugs.python.org/ext'

bugs.python.org fields:

activity = <Date 2014-09-28.08:29:45.063>
actor = 'python-dev'
assignee = 'serhiy.storchaka'
closed = True
closed_date = <Date 2013-10-19.18:05:11.709>
closer = 'serhiy.storchaka'
components = ['Interpreter Core']
creation = <Date 2007-08-12.16:43:07.000>
creator = 'ext-'
dependencies = ['9369', '1699259']
files = ['8163', '30440', '32124']
hgrepos = []
issue_num = 1772673
keywords = ['patch']
message_count = 16.0
messages = ['53019', '58709', '58712', '62954', '189130', '189178', '189181', '189182', '189210', '190239', '190445', '199948', '200458', '200469', '200473', '227753']
nosy_count = 13.0
nosy_names = ['tim.peters', 'brett.cannon', 'amaury.forgeotdarc', 'ncoghlan', 'belopolsky', 'gustavo', 'pitrou', 'dgreiman', 'ext-', 'ned.deily', 'python-dev', 'eric.snow', 'serhiy.storchaka']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue1772673'
versions = ['Python 3.4']

[ext]

Many functions use char* where const char* should be used. I've changed this in a few function prototypes which I use while embedding python due to numerous warnings about deprecated casting from string constants to char*.

[gustavo]

I was about to submit the very same patch :-)
I missed PyErr_NewException, but you missed the PyMapping_* methods ;)

Please look into this matter. GCC 4.2 has arrived and it has a new
warning. The code:

     char* kwlist[] = { "target", "encoding", NULL };

now gives a warning like:

"warning: deprecated conversion from string constant to ‘char*’"

at least when compiling in C++ mode...

This means that people have to declare it as "const char* kwlist", which
will then trigger another warning when calling
PyArg_ParseTupleAndKeywords with such a variable, because the prototype is:

PyAPI_FUNC(int) PyArg_ParseTupleAndKeywords(PyObject *, PyObject *,
                                                  const char *, char **,
...);

The "char **" should be "const char **". ext-, any chance you could fix
that too?

[amauryfa]

There was once a rather long discussion about "const" in
PyArg_ParseTupleAndKeywords:
http://mail.python.org/pipermail/python-dev/2006-February/060689.html

If you don't read it to the end, here is the conclusion:
"""
It sounds like the right answer for Python is to change the signature
of PyArg_ParseTupleAndKeywords() back. We'll fix it when C fixes its
const rules <wink>.
"""

"const char*" was accepted without opposition, though.

[abalkin]

Douglas Greiman submitted many similar changes with his bpo-2135 patch.
His patch also amends documentation, which is missing here.

I am adding dgreiman to the nosy list here to avoid duplication of
effort.

I am -0 on the idea.

[elibendersky]

For external APIs visible to user code, this can cause some compatibility problems and users may need to at the very least re-compile this code. So -1 here.

For new external APIs, having const wherever appropriate should be considered on a case by case basis.

For internal code this is a good idea, but I wouldn't go mass-slashing all the code base now replacing char* by const char*. Presenting some examples where this makes sense in particular would be interesting though. +0

[pitrou]

For external APIs visible to user code, this can cause some
compatibility problems and users may need to at the very least
re-compile this code.

Can you explain exactly which compatibility problems this would cause?

Actually, the point is precisely to make life easier for third-party code, since "const char *" is more accepting than "char *".

[elibendersky]

Antoine, I was referring to the discussion linked earlier (http://mail.python.org/pipermail/python-dev/2006-February/060689.html). Users of the C API needed to recompile their code and also add preprocessor hacks to make things compatible with C and C++, AFAIU.

[pitrou]

Antoine, I was referring to the discussion linked earlier
(http://mail.python.org/pipermail/python-dev/2006-February/060689.html).

Ok, I've read it through. The problem is specifically with
pointers-to-pointers:
http://mail.python.org/pipermail/python-dev/2006-February/060849.html

And indeed, PyArg_ParseTupleAndKeywords() got its "const char **"
argument changed back to "char **". But the other consts have stayed
(such as the "const char *" argument to the same
PyArg_ParseTupleAndKeywords()).

So it looks like the patch, in essence, is legit.
(of course, in practice, it probably won't apply cleanly, since it's
very old; and perhaps it's incomplete too)

[serhiy-storchaka]

In C++ we may overload functions for backward compatibility:

PyAPI_FUNC(int) PyArg_ParseTupleAndKeywords(PyObject *, PyObject *,
                        const char *, const char * const *, ...);
PyAPI_FUNC(int) PyArg_VaParseTupleAndKeywords(PyObject *, PyObject *,
                        const char *, const char * const *, va_list va);
...
#ifdef __cplusplus
inline int
PyArg_ParseTupleAndKeywords(PyObject *args, PyObject *keywords,
                        const char *format, char * const *kwlist, ...)
{
    int retval;
    va_list va;
    va_start(va, kwlist);
    retval = PyArg_ParseTupleAndKeywords(args, keywords, format,
                        (const char * const *)kwlist, va_start);
    va_end(va);
    return retval;
}
#endif

[serhiy-storchaka]

See also bpo-9369.

[serhiy-storchaka]

The patch is outdated. Some function prototypes were changed in bpo-9369, some were changed before.

Here is a new patch. It adds const qualifier to following public functions:
PyObject_DelAttrString, PyObject_HasAttrString, PyObject_GetAttrString, PyObject_SetAttrString, PyObject_DelItemString, PyMapping_DelItemString, PyMapping_HasKeyString, PyMapping_GetItemString, PyMapping_SetItemString, PyFile_FromFd, PyImport_ExecCodeModule, PyImport_ExecCodeModuleEx, PyImport_ExecCodeModuleWithPathnames, PyImport_ImportFrozenModule, PyLong_FromString, PyOS_strtoul, PyOS_strtol, PyOS_Readline, PyMarshal_ReadObjectFromString, PyParser_ParseFile, PyParser_ParseFileFlags, PyParser_ParseFileFlagsEx, PyTokenizer_FromFile. It also changes prototypes of some internal functions and structures and fixes documentation for PyDict_DelItemString. Some of functions already were documented with const qualifier.

[serhiy-storchaka]

The patch synchronized with tip.

[python-dev]

New changeset 3055e61f1e66 by Serhiy Storchaka in branch 'default':
Issue bpo-1772673: The type of char* arguments now changed to const char*.
http://hg.python.org/cpython/rev/3055e61f1e66

[ned-deily]

Compile errors on OS X 10.8 with clang:

cc -Wno-unused-result -Werror=declaration-after-statement -g -O0 -Wall -Wstrict-prototypes     -I. -IInclude -I../../source/Include    -DPy_BUILD_CORE  -c ../../source/Modules/posixmodule.c -o Modules/posixmodule.o
In file included from ../../source/Modules/posixmodule.c:5959:
/usr/include/util.h:94:5: error: conflicting types for 'openpty'
int     openpty(int *, int *, char *, struct termios *,
        ^
../../source/Include/pyport.h:676:12: note: previous declaration is here
extern int openpty(int *, int *, char *,
           ^
In file included from ../../source/Modules/posixmodule.c:5959:
/usr/include/util.h:97:7: error: conflicting types for 'forkpty'
pid_t   forkpty(int *, char *, struct termios *, struct winsize *);
        ^
../../source/Include/pyport.h:678:14: note: previous declaration is here
extern pid_t forkpty(int *, char *,
             ^
2 errors generated.
make: *** [Modules/posixmodule.o] Error 1

[python-dev]

New changeset 8bd69bd6e129 by Serhiy Storchaka in branch 'default':
Restore prototypes for the 'openpty' and 'forkpty' on BSDI (broken in issue bpo-1772673).
http://hg.python.org/cpython/rev/8bd69bd6e129

[python-dev]

New changeset 599a957038fa by Serhiy Storchaka in branch 'default':
Removed redundant casts to char *.
https://hg.python.org/cpython/rev/599a957038fa