homepage

Issue 5793 rationalized all usage of C "character" handling to use "Py_"-prefixed locale-unaware macros, at the "char" level.
In particular, this comment was added in two places to Include/bytes_methods.h in r72044:
http://svn.python.org/view/python/branches/py3k/Include/bytes_methods.h?view=diff&r1=72043&r2=72044
 /* These are left in for backward compatibility and will be removed
 in 2.8/3.2 */
Given that 3.2 is coming soon, is it time to remove these? (also, the reference to "2.8" caught my eye)
Attached is a patch to py3k which removes them, and fixes up various users that were still in the source tree.
Am I right in thinking that the undef and redefinition of the various lower-case macros from <ctype.h> was already intended to be removed? (given that this messes about with a standard C library)

Dave Malcolm wrote:
> 
> New submission from Dave Malcolm <dmalcolm@redhat.com>:
> 
> Issue 5793 rationalized all usage of C "character" handling to use "Py_"-prefixed locale-unaware macros, at the "char" level.
> 
> In particular, this comment was added in two places to Include/bytes_methods.h in r72044:
> http://svn.python.org/view/python/branches/py3k/Include/bytes_methods.h?view=diff&r1=72043&r2=72044
> 
> /* These are left in for backward compatibility and will be removed
> in 2.8/3.2 */
> 
> Given that 3.2 is coming soon, is it time to remove these? (also, the reference to "2.8" caught my eye)
+1
> Attached is a patch to py3k which removes them, and fixes up various users that were still in the source tree.
> 
> Am I right in thinking that the undef and redefinition of the various lower-case macros from <ctype.h> was already intended to be removed? (given that this messes about with a standard C library)
Please remove those as well. I suppose those were mainly used to find
instances of islower() et al. in the source code.

+1
I agree on removing the lowercase versions. It's sort of handy having them around to prevent accidental uses. But the same could be said for other similar macros/functions. Plus they wouldn't work if the isXXX symbols were really functions, not macros (not that I know if that's even allowed by the standard).
The patch looks okay to me, although I didn't actually compile and run the tests. You'll find out soon enough! The issue number in the NEWS entry needs to be fixed, of course.
Thanks for doing this, Dave.

Thanks for reviewing.
If I'm reading things correctly, the ISUPPER et al macros were added in 2.6 and 3.0, and deprecated in 2.7 and 3.1.
Tested with a full run of "-m test.regrtest -uall" here (x86_64 Fedora 13), with both 2-byte and 4-byte unicode; no failures:
 342 tests OK.
 7 tests skipped:
 test_gdb test_kqueue test_ossaudiodev test_startfile test_winreg
 test_winsound test_zipfile64
 Those skips are all expected on linux2.
I've fixed up the issue number in Misc/NEWS in my local tree.
The patch as-is doesn't make sense for 3.1 or 2.7 (the macros should remain within the maintenance branches, in deprecated form) - do I need to explicitly mark this revision (e.g. using "svnmerge.py block"?)

I'd 'svnmerge block' them, just in case anyone decides to manually merge (which I doubt will happen, but you never know).

Committed to py3k in r86210