[dmd-internals] Oldest five bugs

[Andrei Alexandrescu]

On 1/18/12 4:24 PM, Don Clugston wrote:
> On 18 January 2012 22:48, Andrei Alexandrescu<andrei at erdani.com>  wrote:
>> We must start somewhere, and with time the "de-aging" process will
>> accelerate. Without giving old bugs due process, we undermine people's
>> confidence that we're thorough and that every bug will be ultimately looked
>> at.
>
> Do you have any evidence for that?

I do. Every so often people in forums mention "... and that issue is 
opened since 2007 at critical level and with a patch too!"

One person who submitted a bunch of pertinent bug reports and patches 
that stayed ignored for years became the most vocal D naysayer on reddit 
and in IRC. The declared reason was precisely this - that there's 
there's no due process for bugs that are important to some, even when 
accompanied by patches and discussed repeatedly. That person alone did 
very, very much damage to D's image in those forums.

>> I agree regressions are important. But we shouldn't frame things as "either
>> we look at older bugs or fix regressions". I think each release should pay
>> attention to both.
>
> I do not agree. We CANNOT do everything. Anything we do diverts
> attention from something else.
> I don't see any reason for old bugs to be a high priority. Sure it's
> nice, but so are a lot of other things.

Unless what I wrote above and below changes your opinion I think we've 
reached irreducible positions in the matter.

> It would be true if we'd reached a state where all the open bugs are
> of roughly equal importance, (and roughly equal difficulty), but we're
> still not there yet.

I think oldest bugs are of higher importance than randomly placed bugs.

> More importantly, I think this attention is misplaced.
> Organizationally, the attention needs to be on Phobos, which, in stark
> contrast to DMD, has not noticeably benefited from the move to git.
> Look at the changelogs. There are zillions of compiler fixes, and
> practically nothing for Phobos.
> Phobos is not moving. What can we do to improve the situation?

Working on Phobos (adding new features, libraries) has quite a different 
dynamics from working on the compiler (fixing issues, bringing features 
up to predefined spec). This makes it difficult to compare the two and 
tenuous to drive conclusions from the comparison.

To answer the question at face value, I think we need more contributions 
from people who have the time, talent, and willingness. Probably time is 
the major lack in our ranks. We have a few things in the works - the new 
json library, serialization, allocators, containers. Most importantly, 
there are a few good examples of submissions and a good process in place.


Andrei