www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Dconf AGM draft agenda

reply Nicholas Wilson <iamthewilsonator hotmail.com> writes:
The Dconf AGM draft agenda  is up at 
https://github.com/thewilsonator/Dlang-AGM

If you feel that something is missing or I've messed something up 
please do open a PR.
Apr 02 2019
next sibling parent RazvanN <razvan.nitu1305 gmail.com> writes:
On Wednesday, 3 April 2019 at 00:10:53 UTC, Nicholas Wilson wrote:

 The Dconf AGM draft agenda  is up at 
 https://github.com/thewilsonator/Dlang-AGM

 If you feel that something is missing or I've messed something 
 up please do open a PR.


I would throw in the possibility of deprecating `inout`.
Apr 03 2019
prev sibling next sibling parent reply John Colvin <john.loughran.colvin gmail.com> writes:
On Wednesday, 3 April 2019 at 00:10:53 UTC, Nicholas Wilson wrote:

 The Dconf AGM draft agenda  is up at 
 https://github.com/thewilsonator/Dlang-AGM

 If you feel that something is missing or I've messed something 
 up please do open a PR.


For the sake of transparency, seeing as there is a big initial 
commit and no other attributions:

1) Who wrote this agenda?

2) Where did they get the content from?
Apr 03 2019
parent Nicholas Wilson <iamthewilsonator hotmail.com> writes:
On Wednesday, 3 April 2019 at 08:54:41 UTC, John Colvin wrote:

 For the sake of transparency, seeing as there is a big initial 
 commit and no other attributions:

 1) Who wrote this agenda?


Me.


 2) Where did they get the content from?


Built the list up over the course of the last ~6 months from 
forum, PR and slack discussions.
Apr 03 2019
prev sibling next sibling parent reply Dennis <dkorpel gmail.com> writes:
This is a great initiative.

On Wednesday, 3 April 2019 at 00:10:53 UTC, Nicholas Wilson wrote:

 If you feel that something is missing or I've messed something 
 up please do open a PR.


I think it's the opposite of missing things. Looking at: 
http://dconf.org/2019/talks/agm.html

Duration: 4 hours
"expected to last an-hour-and-a-half, but it could go longer"

It's a great backlog of things, but no way you can properly 
discuss that all in a couple of hours. I suggest to make 
something like a top 5 big issue points (things like DIP1000 or 
DMD as a library), and for each point a description of what you 
want out of the discussion / what the goal is.

Thinks like:
int[$] arr = [1,2,3,4];
Are cool, but not worth precious Dconf time compared to larger 
issues I think. I'd really focus on things that were already 
discussed online but did not reach a satisfying conclusion yet.

In any case, I'm looking forward to the outcome of this meeting.
Apr 03 2019
next sibling parent Nicholas Wilson <iamthewilsonator hotmail.com> writes:
On Wednesday, 3 April 2019 at 09:52:19 UTC, Dennis wrote:

 This is a great initiative.

 I think it's the opposite of missing things. Looking at: 
 http://dconf.org/2019/talks/agm.html

 Duration: 4 hours
 "expected to last an-hour-and-a-half, but it could go longer"

 It's a great backlog of things, but no way you can properly 
 discuss that all in a couple of hours. I suggest to make 
 something like a top 5 big issue points (things like DIP1000 or 
 DMD as a library), and for each point a description of what you 
 want out of the discussion / what the goal is.


Well I'm hoping that some/most of the points will have already 
been discussed (possibly even resolved) by the time the AGM takes 
place (it's on the last day).


 Thinks like:
 int[$] arr = [1,2,3,4];
 Are cool, but not worth precious Dconf time compared to larger 
 issues I think. I'd really focus on things that were already 
 discussed online but did not reach a satisfying conclusion yet.


That's one case already ;)


 In any case, I'm looking forward to the outcome of this meeting.


  Me too.
Apr 03 2019
prev sibling next sibling parent Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:
On 4/3/19 5:52 AM, Dennis wrote:

 This is a great initiative.
 
 On Wednesday, 3 April 2019 at 00:10:53 UTC, Nicholas Wilson wrote:

 If you feel that something is missing or I've messed something up 
 please do open a PR.


 
 I think it's the opposite of missing things. Looking at: 
 http://dconf.org/2019/talks/agm.html
 
 Duration: 4 hours
 "expected to last an-hour-and-a-half, but it could go longer"


My flight to Boston takes off from LGW at 4:55 on Saturday so I'll be a 
bit timeboxed.
Apr 03 2019
prev sibling parent Mike Parker <aldacron gmail.com> writes:
On Wednesday, 3 April 2019 at 09:52:19 UTC, Dennis wrote:


 Duration: 4 hours
 "expected to last an-hour-and-a-half, but it could go longer"


That duration of 4 hours is an overlooked error. When editing the 
Ddoc for that page, I started with one of the hackathon files and 
simply missed it.
Apr 03 2019
prev sibling next sibling parent reply Seb <seb wilzba.ch> writes:
On Wednesday, 3 April 2019 at 00:10:53 UTC, Nicholas Wilson wrote:

 The Dconf AGM draft agenda  is up at 
 https://github.com/thewilsonator/Dlang-AGM

 If you feel that something is missing or I've messed something 
 up please do open a PR.


Thanks a lot for this!

It would be super nice to have handy links to the relevant GH 
PRs, Bugzilla issue or forum discussion.
Also each point should have a short description, because I don't 
know what you want to discuss with e.g. "Stuff from 
github.com/truedat101/dlang-sv-community".

Anyhow, a few points:


 Deprecate autodecode – plan of action


tl;dr: the only way we can get rid of it is std.v2


 Dub tester to regression scan for dmd releases - Neia


How is this different to the existing project tester?


 64 compiler release for windows


It's listed twice.
In short: the release setup is pretty complicated and only Martin 
has all the VMs required for the build setup.
Also, you might want to add to your list that the binaries for 
Windows are no longer signed.


 int[$] arr = [1,2,3,4];


I wouldn't get my hopes up for this one.
A DMD PR has been reverted and then closed.

https://github.com/dlang/dmd/pull/3615
https://github.com/dlang/dmd/pull/4377

Since then `staticArray` has been added to Phobos:





the subpoints seem unrelated.


 Quality of documentation


It's pretty bad. DMD documentation is almost unusable, for 
druntime many modules don't have long descriptions or examples. 
Phobos is a lot better, but the ddoc <-> ddox duplication is 
still around.
The tour.dlang.io should have been revisited and reworked a long 
time ago, but no one got around (in short: we want to have a more 
natural intro for readers and show off D features in the 
beginning, not end).


 Transition from DMC+OPTLINK to msvc/mingw + lld


I have tried this before. Walter will definitely object to this. 
Other people also gave it a shot, see e.g.

https://github.com/dlang/dmd/pull/8347
https://github.com/dlang/dmd/pull/9085

With mingwg + lld there is also this problem: 
https://issues.dlang.org/show_bug.cgi?id=19760

At least, dub now does the sensible thing:

https://github.com/dlang/dub/pull/1661


 FreeBSD 11/12


It's a mess and as almost no one uses D on FreeBSD, this doesn't 
move forward:

https://github.com/dlang/dmd/pull/8567

FWIW toolchain triplet selection would have been nice too:

https://github.com/dlang/dmd/pull/8020


 State of std.experimental


The entire idea of std.experimental kinda failed, because 
breaking changes can't be SemVer-ed,
If you're asking whether anything can be stabilized, we're far 
from it:

- stdx.logger: people want a  nogc logger (and IIRC there were 
some API issues)
- stdx.checkedint: it's not even  safe by default (see e.g. 
https://github.com/dlang/phobos/pull/6369, 
https://github.com/dlang/phobos/pull/5928, 
https://github.com/dlang/phobos/pull/6550)
- stdx.allocator: still changes happening very often. The 
community fork diverged and is usable with betterC 
(https://github.com/dlang-community/stdx-allocator). There are 
also quite a few bugzilla issues on it
- stdx.typecons: probably the only module that could be 
stabilized (though it only contains two functions)

I'm not sure what to do with wrap as the existing wrap would need 
to be deprecated and removed, but we could stabilize Final (D's 
head-const):

https://dlang.org/phobos/std_experimental_typecons.html#.Final


 DMD as a library


With stuff like https://github.com/dlang/dmd/pull/9507, it's a 
constant battle :/


 Get bugzilla under control (prioritisation)


We tried quite a few things over the last years:

- #dbugfix campaign
- voting on Bugzilla issues

Currently, we're back to people complaining loudly on the NG 
about very important issues as this is the only way to grab 
attention.

About a year ago, I started an internal discussion about moving 
the issues to GitHub. I tested a potential migration and it 
looked pretty promising. The two big downsides were:
- issue numbers would be different (though a redirect on 
issues.dlang.org and mentioning the original issue number should 
help greatly with that. Furthermore, we could create empty 
"dummy" issues to align the issue numbers if deemed necessary)
- no longer in one global issue database (though since GitHub 
supports issue transfers, probably not a big problem anymore)

In the end it wasn't done because of the fear of loosing control 
over the issue data as the only way to programmatically export 
them is the GitHub API which is severely rate-limited. Though to 
be honest, at the moment Brad is the only with access to 
issues.dlang.org (and no one else to my knowledge is keeping 
backups of issues.dlang.org).

The migration script is here: 
https://github.com/wilzbach/bugzilla-migration
Apr 03 2019
parent reply Nicholas Wilson <iamthewilsonator hotmail.com> writes:
On Wednesday, 3 April 2019 at 11:53:11 UTC, Seb wrote:

 Thanks a lot for this!


No problem.


 It would be super nice to have handy links to the relevant GH 
 PRs, Bugzilla issue or forum discussion.


True, I'm not sure how much I'll be able to get done, I fly out 
tomorrow evening and will be pretty busy until dconf.


 Also each point should have a short description, because I 
 don't know what you want to discuss with e.g. "Stuff from 
 github.com/truedat101/dlang-sv-community".


Whoops I thought I'd fixed that.


 Anyhow, a few points:


 Deprecate autodecode – plan of action


 tl;dr: the only way we can get rid of it is std.v2


Grumble, grumble.


 Dub tester to regression scan for dmd releases - Neia


 How is this different to the existing project tester?


IIRC it was run against all dub projects, as opposed to a select 
group.


 64 compiler release for windows


 It's listed twice.


Its very important ;) (whoops)


 In short: the release setup is pretty complicated and only 
 Martin has all the VMs required for the build setup.


Hmm, that screams single point of failure to me.


 Also, you might want to add to your list that the binaries for 
 Windows are no longer signed.


Noted, thanks.


 int[$] arr = [1,2,3,4];


 I wouldn't get my hopes up for this one.
 A DMD PR has been reverted and then closed.

 https://github.com/dlang/dmd/pull/3615
 https://github.com/dlang/dmd/pull/4377

 Since then `staticArray` has been added to Phobos:


Hmm:


 [Andrei]: I am definitely thinking [$] should not be in D. We 
 need to achieve most of what it does with libraries.




I don't like that reasoning: just because we can doesn't mean we 
should (unnecessary template instances). I note that the same 
reasoning seems to have been used to remove hex strings which 
have come to the end of their deprecation period.





 the subpoints seem unrelated.


That was the PR that took forever to get approved after _I_ 
attempted to reconcile the lack of documentation about dip1000. 
(yes I'm still annoyed)


 Quality of documentation


 It's pretty bad. DMD documentation is almost unusable, for 
 druntime many modules don't have long descriptions or examples. 
 Phobos is a lot better, but the ddoc <-> ddox duplication is 
 still around.
 The tour.dlang.io should have been revisited and reworked a 
 long time ago, but no one got around (in short: we want to have 
 a more natural intro for readers and show off D features in the 
 beginning, not end).


Yup, I'm hoping that having discussed it in the morning some 
otherwise idle people might take the charge at the hackathon (man 
those CIs are going to get hammered).


 Transition from DMC+OPTLINK to msvc/mingw + lld


 I have tried this before. Walter will definitely object to 
 this. Other people also gave it a shot, see e.g.

 https://github.com/dlang/dmd/pull/8347
 https://github.com/dlang/dmd/pull/9085


Well he's going to have to seriously justify that position 
(modulo below).


 With mingwg + lld there is also this problem: 
 https://issues.dlang.org/show_bug.cgi?id=19760


Bummer.


 At least, dub now does the sensible thing:

 https://github.com/dlang/dub/pull/1661


Good.


 FreeBSD 11/12


 It's a mess and as almost no one uses D on FreeBSD, this 
 doesn't move forward:

 https://github.com/dlang/dmd/pull/8567

 FWIW toolchain triplet selection would have been nice too:

 https://github.com/dlang/dmd/pull/8020


Indeed.


 State of std.experimental


 The entire idea of std.experimental kinda failed, because 
 breaking changes can't be SemVer-ed,
 If you're asking whether anything can be stabilized, we're far 
 from it:

 - stdx.logger: people want a  nogc logger (and IIRC there were 
 some API issues)
 - stdx.checkedint: it's not even  safe by default (see e.g. 
 https://github.com/dlang/phobos/pull/6369, 
 https://github.com/dlang/phobos/pull/5928, 
 https://github.com/dlang/phobos/pull/6550)
 - stdx.allocator: still changes happening very often. The 
 community fork diverged and is usable with betterC 
 (https://github.com/dlang-community/stdx-allocator). There are 
 also quite a few bugzilla issues on it
 - stdx.typecons: probably the only module that could be 
 stabilized (though it only contains two functions)

 I'm not sure what to do with wrap as the existing wrap would 
 need to be deprecated and removed, but we could stabilize Final 
 (D's head-const):

 https://dlang.org/phobos/std_experimental_typecons.html#.Final


 DMD as a library


 With stuff like https://github.com/dlang/dmd/pull/9507, it's a 
 constant battle :/


I know.


 Get bugzilla under control (prioritisation)


 We tried quite a few things over the last years:

 - #dbugfix campaign
 - voting on Bugzilla issues

 Currently, we're back to people complaining loudly on the NG 
 about very important issues as this is the only way to grab 
 attention.


Even a list of bounties would be useful.
Apr 03 2019
parent Nicholas Wilson <iamthewilsonator hotmail.com> writes:
On Wednesday, 3 April 2019 at 12:29:25 UTC, Nicholas Wilson wrote:

 On Wednesday, 3 April 2019 at 11:53:11 UTC, Seb wrote:

 the subpoints seem unrelated.




Ah I see, the indentation was off, that was supposed to be a 
subpoint of the point above.
Apr 03 2019
prev sibling next sibling parent reply Mike Franklin <slavo5150 yahoo.com> writes:
On Wednesday, 3 April 2019 at 00:10:53 UTC, Nicholas Wilson wrote:


 If you feel that something is missing or I've messed something 
 up please do open a PR.


I would like to request you please defer judgment on  property 
unless someone who's done significant work with it is present to 
advocate on its behalf.  I'm afraid I won't be there, and I fear 
a rash judgment.

I am well aware that Walter and Andrei are disappointed in 
 property, but I have not heard a single objective argument 
explaining why.  It is mostly "It sucks" and "I don't like it" 
and arguments that optional parens fulfills the need (It doesn't 
and I can make that case).

 property is not fundamentally flawed (at least I haven't heard a 
good argument proving so).  Its implementation is slightly broken 
and very incomplete.  I believe that with a few fixes, a few 
years of adjustment, and the necessary work to complete the 
implementation,  property could become an asset to the language.

What I'm asking for is that unless the right person is there to 
articulate the issue (there are a few), please allow the DIP 
process to work.  We don't need *a* decision, we need the right 
decision and we have a process for reaching that decision.  This 
actually goes for any issue on the AGM agenda.  I think it's OK 
to discuss it, but please defer judgment, and let the DIP process 
work, unless there is someone there that can make the full case 
for it.

 property needs 3 separate DIPs to be a success.  I had 
originally abandoned the Binary Assignment Operators DIP, but 
Andrei's recent "We've been worrying too much about changing 
things" and "We need one standard library that is entirely 
pay-as-you-go" posts has re-kindled my hope that maybe, just 
*maybe*, there still might be some hope for D to be more than an 
obscure language with a cult following.

It won't be easy to rescue  property, but I believe it can be 
done, and will ultimately earn the support of the community.  I 
need help.  If you would like to help, please get in touch with 
me (JinShil on GitHub and Slack).  We need someone with good 
writing skills and the ability to be thorough to write the DIPs.  
Either that or it's on me, and that will take a long while.

Mike
Apr 04 2019
parent Mike Franklin <slavo5150 yahoo.com> writes:
On Thursday, 4 April 2019 at 23:55:16 UTC, Mike Franklin wrote:


  property is not fundamentally flawed (at least I haven't heard 
 a good argument proving so).


BTW, I'm a big advocate of fail early.  If you can make the case 
against  property I'd love to hear it so I don't waste my time.

Mike
Apr 04 2019
prev sibling parent Paolo Invernizzi <paolo.invernizzi gmail.com> writes:
On Wednesday, 3 April 2019 at 00:10:53 UTC, Nicholas Wilson wrote:

 The Dconf AGM draft agenda  is up at 
 https://github.com/thewilsonator/Dlang-AGM

 If you feel that something is missing or I've messed something 
 up please do open a PR.


It would be interesting to understand the plan around having a 
better CTFE performance.

The decision is to wait for Stefan's work to land into master?

If yes, Walter, or someone else, has given a fast look to the 
current state-of-work?
If no, there's a plan B?

- P
Apr 05 2019