Bad result from stat/fstat on W95

Bad result from stat/fstat on W95

[Juanma Barranquero]

I´ve been trying to bootstrap Emacs (from CVS) on Windows 95 and I get
errors about C:/emacs/lisp/xxx.elc (for any xxx.el) not being writable,
and C:/emacs/lisp/ not being a directory.

I think the problem is related to one of these changes:

2002-04-28  Jason Rumney  <jasonr@gnu.org>

        * w32proc.c (syms_of_w32proc): Get true file attributes by default.

        * w32.c (stat, fstat): Use file index information to generate
        inodes for directories where available.

I´ve reverted the changes and after some fiddling I´ve got a working
Emacs.

AFAICS, the problem does not happen on W2K/XP.

-- 
Juanma Barranquero <lektu@terra.es>

Re: Bad result from stat/fstat on W95

[Jason Rumney]

Juanma Barranquero <lektu@terra.es> writes:

> I´ve been trying to bootstrap Emacs (from CVS) on Windows 95 and I get
> errors about C:/emacs/lisp/xxx.elc (for any xxx.el) not being writable,
> and C:/emacs/lisp/ not being a directory.
> 
> I think the problem is related to one of these changes: 
> 
> 2002-04-28  Jason Rumney  <jasonr@gnu.org>
> 
>         * w32proc.c (syms_of_w32proc): Get true file attributes by default.
> 
>         * w32.c (stat, fstat): Use file index information to generate
>         inodes for directories where available.

I don't have access to a W9x system.  Please try to debug what is
going wrong in stat or fstat.  I was careful to catch errors returned
by the system calls, as support on W9x/ME is known to be lacking.
But perhaps they are returning successfully, but with invalid data.
Can we easily tell if the file index returned is invalid? (eg is it 0
in these cases; probably not or the code would work as before I think).

-- 
Jason Rumney

Re: Bad result from stat/fstat on W95

[Juanma Barranquero]

On 06 May 2002 14:45:43 +0100, Jason Rumney <jasonr@gnu.org> wrote:

> I don't have access to a W9x system.  Please try to debug what is
> going wrong in stat or fstat.

I'll try to debug it, but it'll take a while; I only have W95 on my home
machine. BTW, that's a true W95, no 95a, 98 or Me, so perhaps it works
differently (i.e. "right" :) on those systems. 


                                                           /L/e/k/t/u

Re: Bad result from stat/fstat on W95

[Juanma Barranquero]

Jason Rumney <jasonr@gnu.org> writes:

> I don't have access to a W9x system.  Please try to debug what is
> going wrong in stat or fstat.  I was careful to catch errors returned
> by the system calls, as support on W9x/ME is known to be lacking.
> But perhaps they are returning successfully, but with invalid data.
> Can we easily tell if the file index returned is invalid? (eg is it 0
> in these cases; probably not or the code would work as before I think).

Finally I´ve found the time to debug it and I´ve found that the problem
happens exclusively when trying to write a non-existent file, because
the check to see the "writable status" of the directory on which it
would be written is failing.

I think the problem is the following line in stat() (w32.c ~ line 2209):

      buf->st_mode = (wfd.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) ?
        _S_IFREG : _S_IFDIR;

because the condition seems to be reversed. If you agree I´ll commit the
patch to fix it.


                                                     Juanma