Re: Gfortran 2006 Year End Status Report

In article <1167934430.059742.14210@xxxxxxxxxxxxxxxxxxxxxxxxxx>,
ejko123@xxxxxxxxx writes:

Keith Refson wrote:

(* Something some compiler developers (I don't mean specifically gfortran
developers) do not appear to give enough weight to is just how hard
it is and how much effort is required to distill and submit a useful
bug report.

Thank you very much for pointing this out. In my project, which is a
small code (~10,000 lines), I have found bugs in every compiler that I
have ever used, which includes every major vendor. I consider bug
reports a favor to the vendor. When I tried to submit a report to Sun,
for example, I was told to submit a example of 100 lines or less that
triggers the problem.While I understand the strong preference for small
examples, I wasn't able in 20 minutes' time to find a 100-line example,
so I gave up.

I have personally reduced other people's code from the 5K to 10K LOC
range to as few as 5 lines to find bugs in gfortran. Yes, the gfortran
developers prefer small testcase, but if that doesn't happen we will
accept large files and do the reductions ourselves.

The compiler dies with a bus error on very simple F95 codes.

In the context of the rest of your post it is possible to read that
statement as a general comment, which means that it is demonstrably
false (as all over-general statements usually are).

You are absolutely right. I meant to have written *some* very simple
F95 codes, as I can readily demonstrate. In any case, it is good to
see that gfortran is making progress.

Send a bug report to fortran@gcc dot gnu dot org. We can't investigate
nor fix a problem if it isn't reported. There are on the order of 50
open bug reports. You may be hitting one on these, but we rather have
a duplicate report than none at all.

--
Steve
http://troutmask.apl.washington.edu/~kargl/
.

Relevant Pages

  • [RFC][PATCH] Update REPORTING-BUGS (rev. 2)
    ... want to report a kernel bug for the first time. ... Having a standardized bug report form makes it easier for you not to ... +three mailing lists simultaneously, if you think that it is necessry. ...
    (Linux-Kernel)
  • [patch 2/5] Spelling and whitespace fixes for REPORTING-BUGS
    ... Send the output to the maintainer of the kernel area that seems to ... If you are totally stumped as to whom to send the report, ... -This is a suggested format for a bug report sent to the Linux kernel mailing ... and easier for the developers to find the pieces of ...
    (Linux-Kernel)
  • [RFC][PATCH] Update REPORTING-BUGS
    ... want to report a kernel bug for the first time. ... Having a standardized bug report form makes it easier for you not to ... +three mailing lists simultaneously, if you think that it is necessry. ...
    (Linux-Kernel)
  • Re: Component palette option to get rid of junk
    ... and most certainly not the multi interface approach ... that's now the current system (Usenet + QC). ... report, and also options to mark the theme (although you could use the ... sort for items marked as a bug report, thus giving you nearly the same ...
    (borland.public.delphi.non-technical)
  • Re: OpenSource documentation problems
    ... >> There is just one giant roadblock to that suggestion - Sourceforge ... > either in the content or the presentation, please report the bug at the ... up $VISUAL on a bug report template that it's already partially filled ...
    (comp.lang.python)