Re: [PATCH] Libiberty for VMS - mkstemps.c don't mix case

[Douglas B Rupp <rupp at gnat dot com>]

DJ Delorie wrote:
> As a target maintainer, it's up to you to decide what the security
> risks are, and how relevent they are to your platform.  My job is to
> make sure you've thought of them :-)
>
> If we can come up with a generic solution that works right for all
> host environments, that's great.  If not, you get to decide how much
> hacking you want to do for yours.  I think you'll get arguments if you
> expect everyone else to assume their environment is not posix, though.

Right, maybe you can help me understand the risks on VMS.

Going back to the existing behavior make_temp_file will create a new 
file since the test for existence is not version specific, it will ways 
be version 1.  Toplev.c in w+b mode with create a new file with the same 
name, but with version 2 and output the assembly.  Gas, when called will 
open the newest version file for it's input.

What is the risk?  Well seems like a hacker could insert a version 3 of 
the .s file during the time cc1 is working and gas would not be the 
wiser.  Cc1 or Gnat1 could potentially run for a long time so there's 
significant window of opportunity.

Am I thinking about this correctly?

--Doug