I finally found the bug: its name is "g++"

[Roberto Bagnara]

I know the messages of Markus Werle (`I finally found the bug: its
name is "g++"') will be criticized by many.  To the contrary, I
believe his posting is useful and exposes several important problems.
I really hope the tone of his messages will not be exploited as an
excuse not to face the issues they raise.  Please, let me comment on
a couple of things.


Reporting bugs
--------------

I am personally fed up about submitting bug reports.  In these last
years I almost never received an appropriate answer.  I once got a
message telling me, 6 months after I noticed it myself, that the
current compiler no longer exhibited the bug I reported.  I do not
remember having received any help, even when I explicitly asked for
advice on possible workarounds.  I have now stopped submitting bug
reports, even though, for 3.0.1, I could exhibit ICEs, plain
segmentation faults, miscompilations, 13x performance decreases from
g++ 2.96, ... whatever.  Once upon a time a bug report was followed,
within a couple of days, by a "thank you for your bug report" and very
often by real answers, workarounds and, later, by a timely message
confirming that the bug was fixed.

Preparing good quality bug reports costs time and I am now under the
impressions that bug reports are almost completely disregarded.  They
seem also to be irrelevant in deciding when it is time for a release
(look at the number of open bug reports in GNATS, for instance).  A
message I posted on one of the mailing lists before the release of 3.0
remained unanswered.


"No time, we are volunteers"
----------------------------

I am also fed up about this argument, especially because it is used to
justify everything.  The GCC project needs its developers and its
users: no developers, no GCC; no users, no GCC.  The GCC project, as
any project that wishes to be successful, needs to pay attention to
the problems reported by its users.  This has nothing to do with the
status of "volunteer" or "paid a fortune" of the developers.  In other
words, all developers are naturally free to disregard bug reports but
this is an extremely poor decision that will frustrate the users,
especially those doing the service of submitting bug reports.

On a related issue, from time to time messages appear on the mailing
lists that treat a user like an idiot.  Sometimes the user is
reporting a problem, sometimes he is requesting a feature.  Sometimes
the answer has the tone "I am a volunteer, who are you to ask?", "why
don't you write the code yourself instead of asking?"  I will not go
here into the detail of what a community is (supposed to be), but let
me stress that this attitude of treating users like parasites is both
stupid and counterproductive.  Many, many people on the GCC mailing
lists are actively contributing to the free software movements: the
mere act of using GCC is doing so; writing to the mailing lists even
more; several are also volunteers in other free software projects.
Treating them with disrespect is just silly.


Conclusion
----------

I am following the development of GCC since at least 15 years.  I may
be mistaken on several aspects but I am sure of one thing: during the
last years the GCC project has not treated its user base the way it
deserved.

I will not leave g++, but I am now forced to use also another C++
compiler... something I could hardly imagine a few years ago.
All the best,

     Roberto

-- 
Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara@cs.unipr.it