Re: TR library extensions

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

Martin Sebor <sebor@roguewave.com> writes:

| Gabriel Dos Reis wrote:
| 
| ...
| > If they are interchangeable then that is a guarantee not made by
| > us, but by your system. Full stop.  That is not FUD.  That is a fact
| > you can verify.
| 
| No, but suggesting that using something perfectly safe and in wide

There is NO evidence, absolutely NONE, either from our documentation
that several translation units could be compiled with different thread
models, linked together and the whole manipulation would be safe. 
You reported yourself that you found no such guarantee.
If *you* think your manipulation is safe, it is something you have to
take with yourself.  The supposed "safety" you're talking about is NOT
a guarantee we make; it is NOT a behaviour we docuement.  It is
something you get from your system. It is NO FUD. It is simple evidence.

| use might actually lead to bugs is, especially in the absence of
| reliable documentation on the subject and even more so when coming
| from a maintainer of the project.

As a maintainer of the project, I feel a moral obligation, contrary to
you, NOT to make unfounded, irresponsible claim that the
compiler/library could be configured with one thread model and later
linked against a translation unit compiled with a different thread
model, and the whole thing is safe.
If you get a safety guarantee somewhere about your manipulation, it
is not we making it.  

| > The FUD is *you* making people into believing that we support use  a
| > different thread model than the one use at configuration time.
| 
| I made no claims about anything being guaranteed or supported -- I
| used the two options as an example of a feature that has been known
| to work. Coming from a user rather than a maintainer of the compiler
| it can hardly be described as FUD -- it was just an observation.

The "FUD" I was talking about refers to your previous comment.

| Incidentally, though, this patch and thread seem to suggest that the
| goal at one point was to provide support for both options with the
| same compiler: http://gcc.gnu.org/ml/gcc/1997-09/msg01087.html

-pthreads is documented as obsolete and possibly removed from future
versions of GCC.  Apparantly people don't read doc and prefer claim
that maintrainers spread FUD. 

-- Gaby