> From: Bruno Haible <br...@clisp.org> > Cc: bug-gnulib@gnu.org, roucaries.bast...@gmail.com, ebl...@redhat.com, > bonz...@gnu.org, bug-g...@gnu.org > Date: Tue, 03 Jan 2012 14:03:11 +0100 > > Eli Zaretskii wrote: > > > HANDLE h = (HANDLE) _get_osfhandle (fd); > > > > Why not use intptr_t instead of HANDLE > > Because the common way to write code for the Windows API is to use the > Windows types, not the ISO C 89 types.
But the MS headers and MSDN explicitly document _get_osfhandle return type as intptr_t. I guess its finally 89 in Redmond as well ;-) > > and get rid of the cast, both > > here and in IsConsoleHandle above? > > Because that would yield warnings when you compile on mingw with "-Wall". I compiled with -Wall (using intptr_t) and didn't have any warnings. Maybe you are using a newer (= more pedantic) version of GCC. > > Also, I'd replace "NUL device" with a more accurate > > "all character devices such as the null device", similar to what you > > wrote in the manual. > > Well, I tested also aux, com, com1, lpt, lpt1, prn, and none of these > resulted in a successful open() call. Only nul and con could be open()ed, > and only for nul was the value of _isatty() wrong. Tested on Windows XP > (with no printer connected). YMMV. You don't need to open them, you can use redirection: _isatty < nul > prn 2>aux I tested this on XP, and I see the value 64 returned by _isatty for every character device. Btw, it also works for me to open them, but I tried `fopen', not `open'. > Thanks all for your comments. I'm applying this additional patch, > - to fix a test failure on MSVC 9, > - to fix the two wrong comments. Thanks.
