Re: GCC testing questions

[Mike Stump <mrs at windriver dot com>]

> From: gary@Intrepid.Com (Gary Funck)
> Date: Mon, 24 Apr 2000 21:09:48 -0700
> To: gcc@gcc.gnu.org

> but not exactly what standard is used by the GCC community to
> determine when the base compiler is "done" and is ready for
> deployment on the major platforms (what _are_ those major platforms?
> <g>).

It is reasonable for a developer to ask that he be allowed to finish
up current work on a port before gcc ships.  It is even reasonable
that we don't know just what the port is, hence it reasonable for us
to have no clue what ports, other than what we can gleen from the
ChangeLogs, work.  It is reasonable that people test gcc with things
we never see, it is reasonable for us to fix these bugs, hense it is
reasonable that we have no clue _how_ gcc was in fact tested before
release.  So, we wind up with no clue what was tested, how it was
tested, or what all works and doesn't work.  :-) That's one
perspective.  Another perspective, is that all the major ports work to
the extent they have always worked, and that gcc is tested against all
popular free software (with some exceptions) on the most popular
platforms, as well as more formal testing like paranoia, Plum Hall,
Perennial and so on on quite a few of the targets, as well as being
tested inhouse on real live customer code bases.

So, if you want to duplicate testing, just compile up and test all
popular software you can get via ftp, and your own code, and all the
commercial testsuites, and you should be almost on par with what is
done for a gcc release.  apache, perl, xlisp, emacs, binutils, a linux
distribution....  you get the idea.

This is _why_ I didn't originally answer your second question.  I said
it was a good question, not that it had an easy answer.