Re: bug found (in libc)

[Wolfgang Jaehrling]

On Sun, Apr 27, 2003 at 03:27:45PM +0200, Niels Möller wrote:
> Well, until I'm told otherwise, I believe binary compatibility *is* a
> goal, and for the not very distant future (in particular as we now (?)
> have pthreads on the Hurd).

In my understanding, binary compatibility with GNU/Linux is dangerous.
Consider the following:

#ifdef PATH_MAX
  [use PATH_MAX]
#else
  [allocate dynamically]
#endif

If this has been compiled on GNU/Linux and we run it on GNU/Hurd, this
not only introduces an arbitrary limit (which one could probably live
with), it might even result in a buffer overflow if e.g. the PATH_MAX
specific code assumes that it can pass an array of size PATH_MAX to
realpath().

Additionally, I fail to see a good reason why we *would* need binary
compatiblity.  All it gains us is that a) we do not need to recompile
the software and b) one might (at least unless there are other
problems) use the same system with either Linux or the Hurd and Mach.
Is this really that useful?

Cheers,
GNU/Wolfgang