[dmd-internals] Oldest five bugs

[Don Clugston]

On 19 January 2012 02:10, Andrei Alexandrescu <andrei at erdani.com> wrote:
> On 1/18/12 5:34 PM, Leandro Lucarella wrote:
>>
>> If people don't "experience" bugs they will not even look at the age of
>> bugs, they will not even look at the bug tracker! How many times did you
>> checked the bug from other languages?
>
>
> I actually do, and also there's plenty of press that can be seen about it.
> "The flash focus bug in firefox has been opened for five years!" I saw
> similar issues with gcc, php, and others.
>
>
>> I use GCC and Python on a daily
>> basis for years now, and I think I might had a look in their bug
>> trackers about 2 or 3 times in that time frame. With D I look at it this
>> many times per month maybe. That's not because I like to see how old
>> bugs are, but because I get bit by bugs. That are the bugs that should
>> be fixed first.
>
>
> Consider what's now the oldest opened bug:
> http://d.puremagic.com/issues/show_bug.cgi?id=143. Are there more frequently
> encountered bugs? Sure. Are you likely to hit this in a small personal
> project? Probably not. But if someone sees it they may think, "so these
> punks designed and advertised a feature in the language and didn't implement
> it in five years? Good riddance!"

Very bad example. That is one of the highest voted bugs. It's
important for reasons other than age.

> Again, our process of fixing bugs is stochastic enough to not make old bugs
> any less important than any others. So all other things being equal, old
> bugs are more important because they mean bad PR.

Yes, but all other things are far from equal. There are many things
which are bad for PR.
Note that we have an excellent way of finding which bugs are actually
being hit - the voting system in Bugzilla.
We don't have to guess this based on age.


Please note that the process of fixing bugs is not random.
I documented the priority list a very long time ago, it was originally
based on a post by Walter.

http://prowiki.org/wiki4d/wiki.cgi?LanguageDevel#DMDCompilerStability

I've been following that for several years now. It's pretty accurate,
it covers about 70% of the bugs in the changelog.
(For the past eight months, I've focused on CTFE, but within the CTFE
bugs I still follow that priority list).
The notable problem with the priority list, and with the bug fixing,
is that voted bugs haven't historically had a high priority.

One thing which is definitely true is that old bugs are more
difficult. Generally speaking, easy bugs get fixed first.
Those bugs are still open not because they've been forgotten, but
because they're hard.