|
|
 |
 |
 |
 |
Fortran Programming Language
|
 |
 |
 |
 |
|
|
|
|
ICEs "normal" in Gfortran
I thought is was sort of humorous that when I reported a Gfortran ICE
(internal compiler error) as "Critical/Severe", I got an e-mail back
changing its status to "Normal"!!! If I can't even compile my
standard-conforming application, I would call that "critical".
I suspect because I gave them a tiny test case that reproduced the
problem and also mentioned how you could alter it slightly to make it
compile, they assumed I had found a workaround. Unfortunately, I
cannot make the same simple alteration in my full application.
Al
On May 25, 6:32 pm, awgreyno@earthlink.net wrote:
> I thought is was sort of humorous that when I reported a Gfortran ICE > (internal compiler error) as "Critical/Severe", I got an e-mail back > changing its status to "Normal"!!! If I can't even compile my > standard-conforming application, I would call that "critical". > I suspect because I gave them a tiny test case that reproduced the
> problem and also mentioned how you could alter it slightly to make it
> compile, they assumed I had found a workaround. Unfortunately, I
> cannot make the same simple alteration in my full application.
consider it to be a BUG of normal severity. Most bugs I have reported
have been marked normal and later been fixed.
Instead of complaining in shocked tones to comp.lang.fortran, it would
have been better to ASK on the gfortran mailing list about the
handling of your bug.
In article <1180141282.611050.73@u30g2000hsc.googlegroups.com>,
Beliavsky <beliav@aol.com> writes:
> On May 25, 6:32 pm, awgreyno @earthlink.net wrote: >> I thought is was sort of humorous that when I reported a Gfortran ICE >> (internal compiler error) as "Critical/Severe", I got an e-mail back >> changing its status to "Normal"!!! If I can't even compile my >> standard-conforming application, I would call that "critical". >> I suspect because I gave them a tiny test case that reproduced the
>> problem and also mentioned how you could alter it slightly to make it
>> compile, they assumed I had found a workaround. Unfortunately, I
>> cannot make the same simple alteration in my full application.
> When the gfortran developers mark a bug as "normal", that means they
> consider it to be a BUG of normal severity. Most bugs I have reported
> have been marked normal and later been fixed.
> Instead of complaining in shocked tones to comp.lang.fortran, it would
> have been better to ASK on the gfortran mailing list about the
> handling of your bug.
reset the severity to normal. (They also reset the priority
from p1 or p2 to at least p3 or lower). gfortran is simply
not considered to be an important component of GCC by some
GCC developers. In fact, the gfortran developers sometimes
find bugs elsewhere in compiler because we're the ones who
routinely run the gfortran testsuite.
--
Steve
http://troutmask.apl.washington.edu/~kargl/
In article <1180135923.014125.180@i38g2000prf.googlegroups.com>,
awgreyno@earthlink.net writes:
> I thought is was sort of humorous that when I reported a Gfortran ICE > (internal compiler error) as "Critical/Severe", I got an e-mail back > changing its status to "Normal"!!! If I can't even compile my > standard-conforming application, I would call that "critical". > I suspect because I gave them a tiny test case that reproduced the
> problem and also mentioned how you could alter it slightly to make it
> compile, they assumed I had found a workaround. Unfortunately, I
> cannot make the same simple alteration in my full application.
prevents GCC from bootstrapping itself or produces a C compiler
that is so screwed up that the compile is useless. It applies
almost exclusively to the C frontend, middle end, and backend.
If you search bugzilla, you'll find 5 bugs marked as critical.
gfortran bugs fall into one of two categories: normal or
enhancement. The only ones that get elevated to a higher
level are those that can be rewritten into a C testcase
(which normally means a middle end or bacvend bug).
PS: Guess how many people submit a bug report and believe that
their bug is critical? :-)
--
Steve
http://troutmask.apl.washington.edu/~kargl/
Others have commented on the GCC bug classification system, I'll just
underline a few things...
> I suspect because I gave them a tiny test case that reproduced the
> problem
And this is very cool! Self-contained testcases make for easier bug
fixing.
> and also mentioned how you could alter it slightly to make it compile,
> they assumed I had found a workaround. Unfortunately, I cannot make
> the same simple alteration in my full application.
This is not the reasoning that was followed, as was explained. However,
you will see that there are two other workarounds to get your code compile
until this is fixed (which should be done quick, because it's a recent
regression and the problem appears easy to find): you can either
(a) give an explicit type to your function result, or
(b) don't use a result variable (see
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32088#c4)
Thanks for the bug report, and happy coding!
--
FX
As a follow-up, you'll be glad to know that the bug you reported was
fixed yesterday. The nightly binaries available today probably have this
fix (and if they don't, tomorrow's binaries will!).
--
FX
On May 26, 1:32 am, awgreyno@earthlink.net wrote:
> I thought is was sort of humorous that when I reported a Gfortran ICE
> (internal compiler error) as "Critical/Severe", I got an e-mail back
> changing its status to "Normal"!!! If I can't even compile my
> standard-conforming application, I would call that "critical".
First of all, thank you for reporting your bug.
Secondly, as has already been explained, the 'severity' applies to gcc
as a whole. Thus, very bad stuff in gfortran is not severe.
The gfortran maintainers use the keywords as a Richter scale - see our
wiki page:
Bug Bashing (status 25st May 2007; incl. some double counting)
ICE-ON-VALID-CODE, REJECTS-VALID & WRONG-CODE 46 bugs (22
assigned; 4 only in <= 4.2)
ICE-ON-INVALID-CODE & ACCEPTS-INVALID 38 bugs (5 assigned; 3
only in <= 4.2)
DIAGNOSTIC 72 bugs (4 assigned)
All reports (bug reports, feature requests etc.): 277 reports
For obvious reasons, we are concentrating on the first line:)
Please use the keywords in your Bugzilla reports; we do take note of
them.
Paul
It should also be noted that this bug was posted on 2007-05-25 and
fixed on 2007-05-27. Just imagine the turn-around that you would get
if we could mark gfortran PRs as "critical/severe":)
Paul
|
|
|
|
|
|
Add to del.icio.us |
Digg this |
Stumble it |
Powered by Megasolutions Inc