Re: interpretation of wontfix

[Adrian Bunk <bunk@debian.org>]

On Thu, Mar 29, 2018 at 04:21:58PM +0200, Wouter Verhelst wrote:
> On Thu, Mar 29, 2018 at 02:16:31PM +0100, Ian Campbell wrote:
> > On Thu, 2018-03-29 at 14:02 +0100, Ian Jackson wrote:
> > > Don Armstrong writes ("Re: interpretation of wontfix"):
> > > > 2) wontfix+help: this bug requires too much effort to fix, so I won't be
> > > >    working on it, but patches will be accepted.
> > > 
> > > I dislike the use of "help" in this context.  If the maintainers think
> > > the bug is not worth their *own* time, it seems perverse for the
> > > maintainers to set a tag which suggests to other contributors that
> > > they should spend *their* time on it.
> > 
> > The case which caused this thread was[0] "maintainer does not have the
> > time/inclination to investigate/fix bugs on this non-release
> > architecture", the implication being that "the porters of that arch
> > should deal with this bug and provide a patch which the maintainer will
> > evaluate".
> > 
> > To that end perhaps this is a special enough case of "help" that a
> > specific "porter" tag is warranted? (or perhaps a set of "porter-ARCH"
> > tags or a combination of "porter" and "ARCH" tags, or whatever). In
> > fact I don't see why we would limit it to non-release arches, it seems
> > useful for release arches too.
> > 
> > Or perhaps this just needs a consensus on the appropriate use of some
> > `port@debian.org` usertags?
> 
> You mean like https://lists.debian.org/debian-devel-announce/2009/03/msg00015.html ?
> 
> (that never really caught on though)
> 
> I still think that having a per-port tag (which is *not* a usertag)
> would be a good idea, but Don wasn't convinced (at the time; perhaps he
> could be convinced today).

When looking at FTBFS on the buildds common patterns are e.g.:
- FTBFS on big endian
- FTBFS on 32bit
- FTBFS on architectures where char is unsigned

I would not see the benefits of tagging 8 ports when the pattern of 
red/green makes it clear that the problem is due to char being unsigned.

There are special cases like config.{guess,sub} on brand-new archs,
but otherwise it is very rare that a package fails on exactly one 
architecture only.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed