Re: Suggest ``Quality of Release'' Statement in FAQ

[mrs at wrs dot com (Mike Stump)]

> To: egcs@cygnus.com
> Date: Wed, 19 Aug 1998 16:57:18 -0400
> From: "Frederick W. Wheeler" <wheeler@ipl.rpi.edu>

> A useful addition to the EGCS FAQ would be an informed statement
> about the quality of the current release.

This is a hard problem.  And, people don't usually donate much time to
work on it, so it usually it is rather partially solved.

> I'd like to understand better how well EGCS adheres to the standard
> and where the bugs might be (and probably aren't) in the compiler.

See ftp://ftp.cygnus.com/pub/g++/g++-bugs/ for a list of what was
broken that people wanted, that g++ didn't provide.  That is mostly
old now, and no longer maintained.  Whole sections (like templates)
and seen much work, but you can deduce that the code cannot be too old
and might not be as well tested as other parts.

>From there you can check out the g++ testsuite, and see what results
you get.  0 unexpected failures is the number we should for, more than
that, and there are problems that we didn't `expect'.

And finally, you can read -bugs and get a feel for what people report.

> * If a C program compiles, how sure can I be that it will behave as
    it should.

Extensive testing.

> * Same for a C++ program using various features like templates and
    exceptions

Same answer.

> There are many bug reports sent to egcs-bugs@cygnus.com and posted
> to gnu.gcc.bug.  Unfortunately, I can't sort out which complaints
> are legitimate and which are due to misinterpretation of the
> language on the users part.

Yup.  But you can get a feel.


In the end, it is personal experience that does it.