Re: tmpnam() not unique

[Alan Campin <alan0307d@xxxxxxxxx>]

Note that the temp name is unique per activation group(Per the
documentation) which is why I made my temp name service program run in a
separate activation group. That you would have collision between jobs seems
very unlikely.

On Tue, Oct 14, 2014 at 10:29 AM, Scott Hanson <sphanson@xxxxxxxxxx> wrote:

> The filename is generated using a hash function based on the current job
> number and the system clock.  Because the IFS filename which is generated
> contains only six characters for uniqueness, the hash function is not a
> perfect hash and thus has potential collisions (more than one set of input
> values that maps to the same hash value and thus the same filename).  The
> collisions should be very, very unlikely but are possible and may be within
> the same job or a different job.
>
> Scott Hanson
>
> > message: 4
> > date: Tue, 14 Oct 2014 14:52:21 +0000
> > from: "McGovern, Sean" <Sean.McGovern@xxxxxxxxxxxx>
> > subject: RE: tmpnam() not unique
> >
> > Thanks Scott.
> >
> > You say tmpnam() already uses current job number but it is still
> > possible for a collision to occur (with another job) ?
> >
> >
> > -----Original Message-----
> > From: RPG400-L [mailto:rpg400-l-bounces@xxxxxxxxxxxx] On Behalf Of
> > Scott Hanson
> > Sent: 14 October 2014 15:40
> > To: rpg400-l@xxxxxxxxxxxx
> > Subject: Re: tmpnam() not unique
> >
> >
> > The code for the IFS version of tmpnam() uses both the current job
> > number and the current clock (timestamp) to generate a unique
> > filename.  Once the name is generated, the code uses stat() to make
> > sure a filename does not already exist with that name.  If the file
> > already exists, it will generate another name.  The name generated
> > by tmpnam() is unique to the system at time of generation.
> >
> > It is possible for a collision to occur with the filename such that
> > the same filename is generated twice.  The caller of the tmpnam()
> > routine should test that some other process has not created an
> > equivalently named file in the lag time between name generation and
> > file creation.  Some applications can use the tmpfile() function to
> > avoid this issue.
> >
> > Your solution of adding the job number as a prefix/suffix should
> > work well and avoid any potential issues.
> >
> > Scott Hanson
> >
> > > message: 5
> > > date: Tue, 14 Oct 2014 13:19:57 +0000
> > > from: "McGovern, Sean" <Sean.McGovern@xxxxxxxxxxxx>
> > > subject: tmpnam() not unique
> > >
> > > I've believe I have encountered an issue with my use of tmpnam().
> > >
> > > If 2 jobs use tmpnam() at the same timestamp moment, I *think* the
> > > same filename is returned (though I don't know how to test this
> > scenario).
> > > I haven't coded to handle this possibility, and end up with two jobs
> > > trying to create the same temporary IFS file.
> > >
> > > I was thinking that a solution could be just prefix/suffix the job
> > > number to the filename returned by tmpnam().
> > >
> > > Does that seem like a feasible solution ?
> > --
> --
> This is the RPG programming on the IBM i (AS/400 and iSeries) (RPG400-L)
> mailing list
> To post a message email: RPG400-L@xxxxxxxxxxxx
> To subscribe, unsubscribe, or change list options,
> visit: http://lists.midrange.com/mailman/listinfo/rpg400-l
> or email: RPG400-L-request@xxxxxxxxxxxx
> Before posting, please take a moment to review the archives
> at http://archive.midrange.com/rpg400-l.