Re: New xpdf vulnerabilities related to CAN-2004-0888

[Stefan Ulrich <stefanulrich@users.sourceforge.net>]

Frank Küster <frank@kuesterei.ch> writes:

> I also wonder why it seems that the question of compilier-optimization
> (whether the published fixes are really dangerous, because gcc might
> optimize the checks away) is not discussed?


It seems to me that this version:

   if (newSize >= INT_MAX/sizeof(XRefEntry))
      error(...);

is much more readable, more efficient (comparison to a
compile-time constant instead of run-time computation) and,
due to the optimization issue, probably also safer than:

   if (newSize*sizeof(XRefEntry)/sizeof(XRefEntry) != newSize)
       error(...);

Best,
Stefan