Re: [Condor-devel] changes to condor_submit

[Matthew Farrellee <matt@xxxxxxxxxx>]

On 08/05/2010 08:09 PM, Peter Keller wrote:
> On Thu, Aug 05, 2010 at 04:58:06PM -0700, Rob wrote:
> > On Thu, 05 Aug 2010 14:56:09 Dan Bradley wrote:
> > > The problem appears to be from this line in file_transfer.cpp:
> > >
> > > +                       if ( ! access(ExecFile,F_OK | X_OK)>= 0 ) {
> > > +                               free(ExecFile); ExecFile = NULL;
> >
> > Is "access(ExecFile,F_OK | X_OK)" correct?
> > I think there's at least some redundancy there...
> >
> >
> > On my Fedora system, 'man access' says:
> >
> >
> > int access(const char *pathname, int mode);
> >    [...]
> >    The mode specifies the accessibility check(s) to be
> >    performed, and is either the value F_OK, or a mask
> >    consisting of the bitwise OR of one or more of R_OK,
> >    W_OK, and X_OK.
> >    [...]
> >    On  success  (all requested permissions granted),
> > zero is returned.
> >
> >
> > So, "F_OK | X_OK" seems to be redundant!
> >
> > The ZERO return on SUCCESS I did not realized in the past
> > and occasionally wrote buggy code....
> >
> >
> > Please double check that this access() if-conditional-clause
> > really does what it's suppose to do!
>
> Um, is that actually access()? Or access_euid()? If the latter, then
> you know why I have "obscure knowledge maintainer" on my business card. :)
>
> Later,
> -pete

http://condor-git.cs.wisc.edu/?p=condor.git;a=blob;f=src/condor_includes/condor_fix_access.h

"obscure knowledge maintainer" =?= "You think you know something about 
POSIX, well I bet I can trick you."

8oP

This one has been on my hit list for a very long time. Patches welcome 
to eliminate cognitive debt.

$ git grep "access\W*(" | grep -v _access | wc -l
59

Including, from src/condor_syscall_lib/switches.special.cpp, this gem,

/* get rid of the remapping of access() to access_euid() since system calls
   generally should be what they actually should be, and not what we think
   they should be. */
#ifdef access
# undef access
#endif


Best,


matt