[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: release critical gcc bugs



On 99-11-08 Matthias Klose wrote:
> Christian Kurz writes:
>  > On 99-11-07 Daniel Jacobowitz wrote:
>  > > Marking every g++ failure as release-critical is not OK, simply because
>  > > the bugs have workarounds and are very, very difficult to fix.  If you
>  > 
>  > I never talked about marking every bug against g+= as RCB, but I would
>  > say that in internal compiler error is an RCB-bug.

> ICE's are not so critical. You see that something is going wrong and
> can workaround or not. Worse are these bugs which generate wrong code
> ...

And what if I can't work around? Is it then also not release-critical
and just a normal bug?

>  > > meanwhile, there is nothing to be done.  We forward the bugs to the
>  > > upstream authors, and we do what we can to figure out the problem, but
>  > > in a piece of software as complex as gcc it can take a long time to
>  > > uncover the bugs.
>  > 
>  > I never said, that this behaviour is bad or that I dislike it, but such
>  > thinkg please be documented a bit more in the BTS?

> that becomes difficult without any feedback from the maintainers of
> these packages or the people who compile them on alpha. The gcc

Then you should maybe close bugs when the submitter hasn't provided you
with the requested information after 1 month. We can't let bugs stay
open even if the submitter hasn't  provided new information for one
year. We need to make a cut at some point and stop this behaviour.

> maintainers currently lack someone very familiar with the C++ part of
> gcc ... so if somebody listens and would comment the open bugs of g++
> and libstdc++*, please do!

This would be a really good help. You can also ask for help on debian-qa
and see if there are some people around which can help.

> "a bit more documented"? It's a time problem. What I would like to do
> is to reassign these incomplete bug reports to a "debian-alpha"
> package and waiting for the back assignment with the completed
> information.

Interesting idea. What about establishing some virtual packages for
those ports and assign arch-specific bugs to those packages?

Ciao
     Christian
-- 
********************************************************************
* Christian Kurz                          Debian Developer/QA-Team *
*               Use Debian - a free Operating System               *
********************************************************************


Reply to: