Re: merge-with-binutils documentation is wrong/incomplete

[Markus Werle <markus at lufmech dot rwth-aachen dot de>]

"Joseph S. Myers" wrote:

> > Please fill in the following form:
> > ------------------------------------------------
>
> I take it you're volunteering to update the online documentation using the
> answers to this?

If, and only if, I really get clear answers to my questions, and
if those answers are correct, yes. contact me.

Are You able to fill in the form?
If yes: (yelling) PLEASE!!!! NOW!!!

> [...] Try using /cvs/uberbaum (I haven't tried this; see discussion of
> different
> philosophies below).

I do not understand what You mean with "Try using /cvs/uberbaum",
sorry, cvs is not my favourite program.
Could You give the whole command line for the stupid?
Does "Try" mean it could end with another 3 hours wasted?

> [ ... snipped a comprehensive description ...]

Ah! looky there! It is as I expected: chaos, unclear, not converged,
different opinions.
I aim at REMOVING documentation, which tells people about stuff
that is not decided yet. I aim at telling people that gcc (or better: g++)
is in experimental stage with tons of bugs. I aim at being honest
with what is written.

This is not the first time I complain about the docs.
I once proposed a patch to the hpux-specific notes:
My proposal was to clearly say that gcc is less than alpha software
on hpux (better: alpha-version-0.01: I use it every day, it fails to build,
if it does not: my own project that works fine on linux crashes on hpux,
etc.)
Guess whether my proposal was accepted. Well - at least there was a
mini mini small temporary note that hpux-11 is not supported yet,
but cvs might work,which never was true from at least May 2000 until
now. People who reported it worked for them could never remember how
they got it built.

Sorry, I disagree with the way gcc is described. It may be acceptable
for those visiting this page every day and who read all mailings lists,
but the first-time visitor is lost, really lost.
I am sad to read that gcc-2.95.3 was tested with success on a platform
where I cannot build a working gnu-c++-compiler with the info provided
on gcc.gnu.org. I am sad, because I *know* some people did not have
problems, but I cannot figure out why.
(see http://gcc.gnu.org/ml/gcc-bugs/2001-03/msg00606.html)

I propose a new mailing list:
build-success@gcc.gnu.org

Everyone who manages to bootstrap the whole stuff
is asked to describe the whole procedure that led to success.
messages should include answers to:
- from where and how did You download
    binutils and gcc.

- describe Your build procedure with all
   comand line options etc

- describe Your environment:
   which initial compiler was used?
   gmake --version = ?
   flex available?
   etc....


Markus