Reporting bugs in GNAT (was: [Ada] made me hate programming)

[Ludovic Brenta <ludovic@ludovic-brenta.org>]

Warren writes on comp.lang.ada:
> It can be a fun exercise to see what is provoking the bug box. While
> it is best to report those, I do find that the present reporting
> "process" is too time consuming.  I find it is quicker and eaiser to
> find a work-around and hope that the bug has already been reported.

While I understand your point of view, your attitude is exactly the one
that ensures your bug cannot possibly be fixed, ever.  There are too few
Adaists to rely on someone else to report bugs for you.  The "many pairs
of eyes" theorem does not apply to our small community, unfortunately.

> Only as a last resort, I'll report it.

OK but that's another mistake.  If reporting a bug is your last line of
defense, this means you're pretty much cornered and have become
dependent on someone else understanding your problem, implementing a fix
and delivering the fix to you.  This leaves bad feelings on both sides:
you feel neglected, and they feel your're pressuring them to work for
you in emergency mode, for free.

> They need to make that process easier somehow. For example, allow the
> user to just send all the *.adb and *.ads files used.

To paraphrase John F. Kennedy: "ask not what GCC can do for you; ask
what you can do for GCC".

I suggest that if you invest a couple of hours of your precious hobby
time on just *one* compiler bug to properly isolate it in a reproducer,
you will not only gain a lot of insight into the process but you will
also deserve, and obtain, a lot of goodwill from the GCC maintainers
(not only those at AdaCore, but also enlightened enthusiasts like Sam
Tardieu and Laurent Guerby, to name but two).  This might lead to a fix
sooner than you think.

> Instead they insist on *ada files and don't want anything that is not
> involved.  Gnat tries to name the involved sources, but I have found
> that this isn't reliable - resulting in more futzing around.

That's a small issue with file naming.  Once you've done it once, you
can do it again quite easily.

> I understand the reasons for these things but this creates a lot of
> fussing for the bug submitter.  The harder it is to submit a bug, the
> more it discourages submissions. There were some other things but I
> don't recall them now.  By the time you've submitted the report, 2-3
> hours have wizzed by.

Correct.  Those 2-3 hours are time that you give back to the people who
gave you millions of hours of expert work to produce GCC in the first
place.  They are a small price to pay, and you can also see them as an
investment.  Those 2-3 hours of your time will pay back handsomely in
the form of fixes and goodwill.

If you really cannot spend those 2-3 hours, then I suggest you spend
just 1 hour and submit the bug nevertheless.  Ignore comments from Arno
saying that the reproducer is too big.  Ask for advice here on
comp.lang.ada.  Someone might be interested enough to reduce the
reproducer for you.  I've done that in the past myself, resulting in a
quick fix after that; see http://gcc.gnu.org/PR29015 to see one example.

-- 
Ludovic Brenta.