[Bug: 21.4.20] Network Drives

[Steve Higham]

Stephen,

My machine contains several perl scripts (*.pl) which are certainly
executable. Can the "executable extension list" be modified as a standard
lisp variable?

-----Original Message-----
From: Stephen J. Turnbull [mailto:stephen at xemacs.org] 
Sent: 28 April 2007 13:46
To: Steve Higham
Cc: 'Benson Margulies'; xemacs-beta at xemacs.org
Subject: RE: [Bug: 21.4.20] Network Drives

Thanks for diagnosing and testing!

Steve Higham writes:

 > /* All files are eXecutable under WIN32. Therefore we simply test for the
 >    existence of the file. I'm not sure what should happen under Cygwin */

Cygwin will need its own workarounds for this; I don't think it's
worth worrying about it now, as we may need to require a recent
version of Cygwin for Vista in any case.

 > #ifndef X_OK
 > # ifdef WIN32_NATIVE
 > #  define X_OK 0
 > # else
 > #  define X_OK 1
 > # endif
 > #endif

I don't think this is appropriate.  There are only a small number of
uses of X_OK, including at least one platform-independent function
that returns it.  You'd need to trace such returned values to be sure
they weren't used outside of Windows-specific calls, and of course you
couldn't be sure they wouldn't be in the future.

What 21.5 does in mswindows_access() in nt.c is to check if (mode &
X_OK) is nonzero, and if so calls a function
mswindows_is_executable().  I think this is the preferred idiom.

BTW, mswindows_is_executable() itself looks kinda problematic: it
checks for the extension .bat, .cmd, .com, or .exe, and decides the
file is executable iff it's one of them.  (Not a problem in 21.4,
since that function doesn't exist in 21.4.)

 > If you're comfortable with this change, could you check it into the
 > codebase. I haven't updated the xemacs source code before and would
 > rather not start now.

Note that all commits to 21.4 must be approved by Vin Shelton.  Please
don't commit to the stable branch without his explicit approval.