Mail Archives: cygwin-developers/2003/01/21/12:55:36
Gerald S. Williams wrote:
> Earnie Boyd wrote:
>
>>IMNSHO, the problem is with the project that uses files that differ only
>>in case. It's not portable, and if the project wishes portability, then
>>the practice must stop. I urge you to raise an argument with the
>>package maintainers accompanied with a patch. Changing Cygwin to handle
>>the problem isn't going to cause these problems to disappear as not all
>>environments will support it.
>
>
> Taking the high road, huh? :-)
>
> Of course, many of the people you'd have to convince
> would argue that the problem is with Windows.
>
The argument has nothing to do with Windows vs Unix.
> I'm not taking sides on that debate. But I have run
> into these naming issues enough times to say that,
> for me, it would be really nice to have POSIX naming
> as an option.
>
Nice, but it's still not portable.
> Another problem I've run into recently is that the
> people who ARE interested in portability are already
> taking the trouble to port their package to Windows.
> More often than not, this actually interferes with a
> Cygwin build, since they start adding WIN32 checks
> throughout the code, taking into consideration not
> only platform issues but also compiler weirdness. It
> would be nice if we can "beat them to the punch" and
> have a Cygwin port first. To this end, being able to
> deal with such files may give us a head start.
>
Adding filters for WIN32 shouldn't affect a Cygwin build at all. WIN32
is only defined if you use -mno-cygwin, -mwin32 or include windows.h.
As for compiler specificness, I agree, I've just run into #ifdef
__MSVC__ in a package instead of _WIN32. But again the argument is
portability and nothing else.
Earnie.
- Raw text -