https://www.bountysource.com/issues/1325905-shared-phobos-library-doesn-t-work-on-all-linux-distributions

Over $2000 in open bounties left:

https://www.bountysource.com/trackers/383571-d-programming-language

https://www.bountysource.com/trackers/455080-gdc

https://www.bountysource.com/trackers/283332-ldc



Andrei

Nick recently did a patch for this one:

https://www.bountysource.com/issues/1327154-dmd-never-inlines-functions-that-could-throw+

we've had a lot of movement on this one

https://www.bountysource.com/issues/1326911-dtoh-utility-convert-d-files-to-c-header-files

and it pretty well works now waiting on the OK to merge:

https://github.com/D-Programming-Language/tools/pull/39



Generally though, I don't think the bounties are going to change much behavior; the only issues that will be addressed are the ones that we were going to do anyway, since the dollar amount is just too small to change a business decision.

Take multiple alias this for example. I've looked at the dmd source for that before and judged it would be a pretty big job - probably a week devoted to it, if not more. The bounty is $100.... so that's, what, $2/hour? There's very little practical difference between that and doing it just because I felt like it; financially, I'd be better off flipping burgers. No real change to the incentive.

On Thursday, 13 March 2014 at 18:20:16 UTC, Andrei Alexandrescu wrote:
> https://www.bountysource.com/issues/1325905-shared-phobos-library-doesn-t-work-on-all-linux-distributions
>
> Over $2000 in open bounties left:
>
> https://www.bountysource.com/trackers/383571-d-programming-language

Looks like most of these are on compiler bugs.

The only Phobos one is the std.getopt one, however its situation is two abandoned patches and no clear goal as to what constitutes a change worthy of marking the issue as "fixed" and paying out the bounty.

As for the compiler bugs... well, all of these are HARD, at least from the perspective of someone inexperienced with DMD's codebase. If they were easy, they'd have been solved already. DMD is not something you can easily dive into and start moving code around to fix big problems. Speaking from experience, it wasn't once that a 50-line patch would take me days to author and debug. And even if I were to manage through, the chances are high that the patch ends up crap because you need a lot of knowledge about how the compiler works to understand what's a good idea, and what isn't. Not even Kenji's pulls are always approved.

However, the biggest problem is the open pull request count. What good is authoring a patch if no one wants to take time to review it?

IMHO, we don't need more bug bounties - we need REVIEWER bounties. Some way to convince more experienced D developers to review others' contributions. I have a number of open DMD pull requests myself, and sometimes I consider bribing someone to just look at them.

On Thursday, 13 March 2014 at 18:38:45 UTC, Adam D. Ruppe wrote:
> Nick recently did a patch for this one:
>
> https://www.bountysource.com/issues/1327154-dmd-never-inlines-functions-that-could-throw+
>
> we've had a lot of movement on this one
>
> https://www.bountysource.com/issues/1326911-dtoh-utility-convert-d-files-to-c-header-files
>
> and it pretty well works now waiting on the OK to merge:
>
> https://github.com/D-Programming-Language/tools/pull/39
>
>
>
> Generally though, I don't think the bounties are going to change much behavior; the only issues that will be addressed are the ones that we were going to do anyway, since the dollar amount is just too small to change a business decision.
>
> Take multiple alias this for example. I've looked at the dmd source for that before and judged it would be a pretty big job - probably a week devoted to it, if not more. The bounty is $100.... so that's, what, $2/hour? There's very little practical difference between that and doing it just because I felt like it; financially, I'd be better off flipping burgers. No real change to the incentive.

The difference is though that people are working on problems, hard problems even, with the compiler already for free. I for one wouldn't mind getting a $100 for fixing something I already have an interest in fixing. In other words, I don't think they do any harm.

On 3/13/14, Vladimir Panteleev <vladimir@thecybershadow.net> wrote:
> However, the biggest problem is the open pull request count. What good is authoring a patch if no one wants to take time to review it?
>
> IMHO, we don't need more bug bounties - we need REVIEWER bounties. Some way to convince more experienced D developers to review others' contributions. I have a number of open DMD pull requests myself, and sometimes I consider bribing someone to just look at them.

I think I know what's causing this stalling. As more bugs get fixed the bugs that are left opened are the ones which are the most complicated to fix (essentially most of the easy ones are fixed already).

E.g. Kenji tends to tackle more and more complex and intricate bugs. So these complicated DMD changesets can only really be reviewed by people with intricate knowledge of the compiler. I don't think there's that many people around with that kind of knowledge. Personally, I've only ever tackled smaller bugs, most of Kenji's changesets go way over my head and I wouldn't be able to merge something like that with confidence, regardless of the autotester.

You could easily substitute "PHD" with "DMD" on this page: http://matt.might.net/articles/phd-school-in-pictures/

On Thursday, 13 March 2014 at 18:38:45 UTC, Adam D. Ruppe wrote:
> Generally though, I don't think the bounties are going to change much behavior; the only issues that will be addressed are the ones that we were going to do anyway, since the dollar amount is just too small to change a business decision.
>
> Take multiple alias this for example. I've looked at the dmd source for that before and judged it would be a pretty big job - probably a week devoted to it, if not more. The bounty is $100.... so that's, what, $2/hour? There's very little practical difference between that and doing it just because I felt like it; financially, I'd be better off flipping burgers. No real change to the incentive.

I agree with Adam.  While I think it's great that you're taking the bounty approach, which I suggested sometime back, this isn't enough money to incentivize people much.  It might get someone like Brad, who had these issues on his todo list anyway, to reorganize his list to put the bountied items on top, but it's not going to do much more than that.

I still think D needs a paid compiler to drive real change, hopefully someone will use ldc with the closed patches model I laid out in this forum earlier and push that through.

Adam D. Ruppe, el 13 de March a las 18:38 me escribiste:
> Nick recently did a patch for this one:
> 
> https://www.bountysource.com/issues/1327154-dmd-never-inlines-functions-that-could-throw+
> 
> we've had a lot of movement on this one
> 
> https://www.bountysource.com/issues/1326911-dtoh-utility-convert-d-files-to-c-header-files
> 
> and it pretty well works now waiting on the OK to merge:
> 
> https://github.com/D-Programming-Language/tools/pull/39
> 
> 
> 
> Generally though, I don't think the bounties are going to change much behavior; the only issues that will be addressed are the ones that we were going to do anyway, since the dollar amount is just too small to change a business decision.
> 
> Take multiple alias this for example. I've looked at the dmd source for that before and judged it would be a pretty big job - probably a week devoted to it, if not more. The bounty is $100.... so that's, what, $2/hour? There's very little practical difference between that and doing it just because I felt like it; financially, I'd be better off flipping burgers. No real change to the incentive.

Yeah, I see the bounties more like a reward ("OK, you fixed something I wanted to be fixed, there you go") than a motivator to fix issues (I don't think anybody will say "Oh! I want to work in this issue because I can make $100!").

Is still better than nothing, and at least a nice gesture to the community, but definitely not bounty-driven development :D

-- 
Leandro Lucarella (AKA luca)                     http://llucax.com.ar/
----------------------------------------------------------------------
Como un rinoceronte que lleva un pájaro en el lomo,
yo te alimento, no te veo ni te toco.

On 3/13/2014 2:38 PM, Adam D. Ruppe wrote:
>
>  financially, I'd be better off flipping
> burgers. No real change to the incentive.

I flipped burgers for a month when I was 16. Money being the same, I'd rather patch a compiler ;)

But yea, if you do the math it does tend to work out less than minimum wage. Still, that's more than the $0/hr for most of the D contributions we're happy to do, so I'm not complaining. At the very least it can help support the coffee/snacks fund (ie, code fuel) for the next D contribution :)

On 3/13/14, 11:38 AM, Adam D. Ruppe wrote:
> Nick recently did a patch for this one:
>
> https://www.bountysource.com/issues/1327154-dmd-never-inlines-functions-that-could-throw+
>
>
> we've had a lot of movement on this one
>
> https://www.bountysource.com/issues/1326911-dtoh-utility-convert-d-files-to-c-header-files
>
>
> and it pretty well works now waiting on the OK to merge:
>
> https://github.com/D-Programming-Language/tools/pull/39

Thanks.

> Generally though, I don't think the bounties are going to change much
> behavior; the only issues that will be addressed are the ones that we
> were going to do anyway, since the dollar amount is just too small to
> change a business decision.
>
> Take multiple alias this for example. I've looked at the dmd source for
> that before and judged it would be a pretty big job - probably a week
> devoted to it, if not more. The bounty is $100.... so that's, what,
> $2/hour? There's very little practical difference between that and doing
> it just because I felt like it; financially, I'd be better off flipping
> burgers. No real change to the incentive.

What would make the amounts interesting?

Andrei

On 3/13/2014 4:48 PM, Andrej Mitrovic wrote:
>
> So these complicated DMD changesets can only really be reviewed by
> people with intricate knowledge of the compiler. I don't think there's
> that many people around with that kind of knowledge.

I think that's a good, and important, point. And I think that's one thing that's good about the bounties: They can help encourage more people to become more knowledgeable about DMD's internals. It can help seed future DMD gurus.