Re: [Mingw-users] Confusion between directory name and header name?

[Greg C. <chi...@mi...>]

Norman Vine wrote:
> 
> Greg Chicares
> >
> >In the following example, creating a directory named 'new' on the
> >include path makes the system header <new> inaccessible. Is this
> >expected behavior?
> 
> Ahh..  Deja Vu.
> 
> See thread starting at
> http://sources.redhat.com/ml/cygwin/1999-12/msg00286.html

Thanks. I thought I might have misunderstood some arcane rule for
gcc directory searches.

In this message
  http://sources.redhat.com/ml/cygwin/1999-12/msg00300.html
Mumit wrote:
> 
> Can you tell us exactly what happened so that we'll know next
> time somebody runs into something similar (and you know that's
> bound to happen!)?
> 
> Is it because you had directory named `new' in the include
> search list? If so, then the bug is deep inside of win32 API
> which refuses to "open" directories like on Unix and I'd
> like to provide a way around it.

That was my problem (test case in original posting). It's not
confined to <new>; it looks like it happens whenever there's
a directory with the same name as a header. It doesn't matter
whether it's included with <>
  #include <header>
or ""
  #include "header"

For example, after setting up three such directories, I get:

C:\testing>type oops.cpp
#include <cmath>
#include "vector"
#include <float.h>

C:\testing>g++ -I. -c oops.cpp
oops.cpp:1: cmath: Permission denied
oops.cpp:2: vector: Permission denied
oops.cpp:3: float.h: Permission denied

So, at least with the default 2.95.2-1 setup, it's a very bad
idea to name a header <objc>, since that's a subdirectory of
a system header directory.