* GNATS policy question
@ 2002-12-11 8:01 Volker Reichelt
2002-12-11 22:33 ` Mark Mitchell
0 siblings, 1 reply; 6+ messages in thread
From: Volker Reichelt @ 2002-12-11 8:01 UTC (permalink / raw)
To: gcc
What's the GNATS policy about PRs that issue an ICE in development versions
and a "confused by earlier errors, bailing out" message in release
versions?
We do have some reports with bogus code snippets that cause the
parser to choke and show the above behaviour (e.g. PR 8895, PR 8850).
Are there any chances that such failures will be fixed? Should we mark
them somehow since they are probably of very low priority? Or shall we
close such PRs unconditionally? In the latter case it would be nice if
we could point the submitters to some explanation about ICEs and
"confused ..." messages and why we close such PRs although there's an ICE.
Regards,
Volker
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: GNATS policy question
2002-12-11 8:01 GNATS policy question Volker Reichelt
@ 2002-12-11 22:33 ` Mark Mitchell
2002-12-12 3:35 ` Volker Reichelt
0 siblings, 1 reply; 6+ messages in thread
From: Mark Mitchell @ 2002-12-11 22:33 UTC (permalink / raw)
To: Volker Reichelt, gcc
--On Wednesday, December 11, 2002 05:55:04 PM +0100 Volker Reichelt
<reichelt@igpm.rwth-aachen.de> wrote:
> What's the GNATS policy about PRs that issue an ICE in development
> versions and a "confused by earlier errors, bailing out" message in
> release versions?
We should mark those PRs medium-priority.
GCC simply doesn't have a coherent design for error-recovery, and until
it does the set of places where we crash after error messages will move
around from time to time, hopefully shrinking over time. I have ideas
about this, but not enough time to implement them ...
--
Mark Mitchell mark@codesourcery.com
CodeSourcery, LLC http://www.codesourcery.com
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: GNATS policy question
2002-12-11 22:33 ` Mark Mitchell
@ 2002-12-12 3:35 ` Volker Reichelt
2002-12-12 10:08 ` Mark Mitchell
0 siblings, 1 reply; 6+ messages in thread
From: Volker Reichelt @ 2002-12-12 3:35 UTC (permalink / raw)
To: mark; +Cc: gcc
On 11 Dec, Mark Mitchell wrote:
>> What's the GNATS policy about PRs that issue an ICE in development
>> versions and a "confused by earlier errors, bailing out" message in
>> release versions?
>
> We should mark those PRs medium-priority.
>
> GCC simply doesn't have a coherent design for error-recovery, and until
> it does the set of places where we crash after error messages will move
> around from time to time, hopefully shrinking over time. I have ideas
> about this, but not enough time to implement them ...
Shouldn't we suspend those PRs then? We could add "[error-recovery]" to
the synopsis line. Maybe some of the PRs will be fixed with the new
parser, so IMHO we should suspend them till the new parser arrives
anyway before spending too much effort on them.
Regards,
Volker
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: GNATS policy question
2002-12-12 3:35 ` Volker Reichelt
@ 2002-12-12 10:08 ` Mark Mitchell
2002-12-13 1:31 ` Volker Reichelt
0 siblings, 1 reply; 6+ messages in thread
From: Mark Mitchell @ 2002-12-12 10:08 UTC (permalink / raw)
To: Volker Reichelt; +Cc: gcc
--On Thursday, December 12, 2002 10:48:03 AM +0100 Volker Reichelt
<reichelt@igpm.rwth-aachen.de> wrote:
> On 11 Dec, Mark Mitchell wrote:
>
>>> What's the GNATS policy about PRs that issue an ICE in development
>>> versions and a "confused by earlier errors, bailing out" message in
>>> release versions?
>>
>> We should mark those PRs medium-priority.
>>
>> GCC simply doesn't have a coherent design for error-recovery, and until
>> it does the set of places where we crash after error messages will move
>> around from time to time, hopefully shrinking over time. I have ideas
>> about this, but not enough time to implement them ...
>
> Shouldn't we suspend those PRs then? We could add "[error-recovery]" to
> the synopsis line. Maybe some of the PRs will be fixed with the new
> parser, so IMHO we should suspend them till the new parser arrives
> anyway before spending too much effort on them.
It's very hard to say where the problems are. Some might be fixed by the
new parser. Others need error recovery code in the front ends. I think
it's OK to keep them open -- they *are* defects. Just not very important
ones. Maybe mark them low priority?
--
Mark Mitchell mark@codesourcery.com
CodeSourcery, LLC http://www.codesourcery.com
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: GNATS policy question
2002-12-12 10:08 ` Mark Mitchell
@ 2002-12-13 1:31 ` Volker Reichelt
2002-12-13 8:39 ` Mark Mitchell
0 siblings, 1 reply; 6+ messages in thread
From: Volker Reichelt @ 2002-12-13 1:31 UTC (permalink / raw)
To: mark; +Cc: gcc
On 12 Dec, Mark Mitchell wrote:
>
>
> --On Thursday, December 12, 2002 10:48:03 AM +0100 Volker Reichelt
> <reichelt@igpm.rwth-aachen.de> wrote:
>
>> On 11 Dec, Mark Mitchell wrote:
>>
>>>> What's the GNATS policy about PRs that issue an ICE in development
>>>> versions and a "confused by earlier errors, bailing out" message in
>>>> release versions?
>>>
>>> We should mark those PRs medium-priority.
>>>
>>> GCC simply doesn't have a coherent design for error-recovery, and until
>>> it does the set of places where we crash after error messages will move
>>> around from time to time, hopefully shrinking over time. I have ideas
>>> about this, but not enough time to implement them ...
>>
>> Shouldn't we suspend those PRs then? We could add "[error-recovery]" to
>> the synopsis line. Maybe some of the PRs will be fixed with the new
>> parser, so IMHO we should suspend them till the new parser arrives
>> anyway before spending too much effort on them.
>
> It's very hard to say where the problems are. Some might be fixed by the
> new parser. Others need error recovery code in the front ends. I think
> it's OK to keep them open -- they *are* defects. Just not very important
> ones. Maybe mark them low priority?
Low priority sounds good to me.
Three final questions:
* Should we add "[error-recovery]" to the synopsis line so that one does
not stumble over those bugs over and over again?
* Shall I prepare a patch for the policies section of gnatswrite.html
to document the procedure?
* There are some bugs in GNATS that were closed because a "confused by
earlier ..." message was issued. Shall I reopen them?
Regards,
Volker
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: GNATS policy question
2002-12-13 1:31 ` Volker Reichelt
@ 2002-12-13 8:39 ` Mark Mitchell
0 siblings, 0 replies; 6+ messages in thread
From: Mark Mitchell @ 2002-12-13 8:39 UTC (permalink / raw)
To: Volker Reichelt; +Cc: gcc
> Low priority sounds good to me.
>
> Three final questions:
> * Should we add "[error-recovery]" to the synopsis line so that one does
> not stumble over those bugs over and over again?
No; I think that will get too crowded. But a note in the audit trail
would be nice.
> * Shall I prepare a patch for the policies section of gnatswrite.html
> to document the procedure?
Yes, please.
> * There are some bugs in GNATS that were closed because a "confused by
> earlier ..." message was issued. Shall I reopen them?
No, it's really not worth spending that much effort on this class of
bugs. Once we never, ever silently generate bad code we can worry much
more about these kinds of things... :-)
--
Mark Mitchell mark@codesourcery.com
CodeSourcery, LLC http://www.codesourcery.com
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2002-12-13 16:26 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-12-11 8:01 GNATS policy question Volker Reichelt
2002-12-11 22:33 ` Mark Mitchell
2002-12-12 3:35 ` Volker Reichelt
2002-12-12 10:08 ` Mark Mitchell
2002-12-13 1:31 ` Volker Reichelt
2002-12-13 8:39 ` Mark Mitchell
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).