Re: dip1000 state

[carblue via Digitalmars-d]

On Saturday, 3 March 2018 at 02:12:28 UTC, Walter Bright wrote:

On 3/2/2018 10:07 AM, carblue wrote:
I generally already used -dip1000 since DConf2017 and it 
served me well, until about 2 month ago, "by accident" code 
was committed to std.uni that broke my builds, see issue 
#17961. I invested a lot of time to fix this by PR 6041.
The current state is: I don't know any reason why it doesn't 
get (can be?) merged and now it languishes on page 2 of 3 of 
PRs and Walter started a new PR 6212 recently after my PR is 
ready for weeks. In total 4 of my -dip1000 related PRs are 
stuck and mostly for unknown or arguable reasons.


For reference, I submitted:

  https://github.com/dlang/phobos/pull/6212

which I remarked was puzzlingly different from your PR:

  https://github.com/dlang/phobos/pull/6041

I would appreciate your advice on that as well.

I was unaware of your PR. Sometimes, it's worth while to make 
some noise if you're feeling overlooked.


I just want to get a problem solved, as soon as possible, and 
std.uni breaking -dip1000 builds happens since at least 
2017-11-02 (https://issues.dlang.org/show_bug.cgi?id=17961), 
seemingly sleeping for 2.5 month until I assigned myself in 
bugzilla willing to fix this issue.
If there is duplication/overlapping in PRs from several people, 
so what, it just happens - while not as effective as I would like 
D processes to be.


The point I wanted to make here is, that the longer a PR is 
languishing (farer from PR stack's top), the more likely dup-PRs 
will be and in this case even PRs marked "Blocking Other Work, 
Bug Fix" being overlooked.
In my situation of growing frustration about 4 stuck, espec. 
https://github.com/dlang/phobos/pull/6204 and 
https://github.com/dlang/phobos/pull/6041 (out of 8(9) submitted 
-dip1000 PRs; #6195 is known to possibly not being wise to merge 
unchanged), PR 6212 came along, You analyzed puzzling 
differences, I commented on that 8 days ago and since then 
nothing changed from my perspective:
Now, 4 months after opening 17961, it's still unfixed and 
possibly not fixed with the next release.


It's as simple as: I want my time invested being of any value, 
and if another PR solves a problem better than my solution and 
gets merged: Fine and I'll learn from that.


Andrei said at DConf2017 sys. panel talk: He intends to be "nice" 
when NOT saying: This xy-PR won't make it into dlang. To the 
contrary I explicitly want "negativ/refusal" feedback as well and 
I feel, retaining this is not "nice". We learn from our faults 
mostly.



Both Andrei and I are way overloaded, and I generally defer
to Andrei to watch the Phobos PRs.


I know and like Your unobtrusive way to point to overload. I 
consider delving into dmd code. Perhaps that might once help 
taking some load from Your shoulders.

Sorry for the missing linkages.


For reference, here are your open PRs:
  
https://github.com/dlang/phobos/pulls?q=author%3Acarblue+is%3Aopen


You've had 4 others that were pulled in the last month; you 
haven't been totally ignored:


  
https://github.com/dlang/phobos/pulls?q=author%3Acarblue+is%3Aclosed


I do appreciate the work you're doing to get -dip1000 working 
with Phobos. It's important!


That's pleasant to hear from the language author, and frankly, I 
didn't cast doubts on that concerning You, once You take notice, 
but mostly others are dealing with the PR processing.


I can say, I very much appreciate the D language, Andrei's and 
Your work and pushing DIP1000 even against a wall of 
indifference. I believe the latter is changing (now) and the 
sooner - howsoever - #17961 get's fixed, people can use/try 
-dip1000 again.


-dip1000 compilabe phobos is pretty close - as You say - and 
finalizing helps taking more bricks out of the wall.

Okay, I'll give it another try to push that.

carblue, Carsten Blüggel

Re: dip1000 state

[Radu via Digitalmars-d]

On Saturday, 3 March 2018 at 09:04:04 UTC, Walter Bright wrote:

On 3/2/2018 10:48 PM, Manu wrote:
I've been arguing for DIP1000 (or something very much like it) 
for
almost 10 years. It's one of the small handful of issues that 
lead me

to the forum in the first place.
'Patience' might not be the right word at this point ;)


It is nice to see some interest in it. I've felt like I've been 
pushing against a wall of indifference for a couple years on it.


The good news is:

1. it's implemented
2. it works
3. it's effective

The bad news is:

1. it's viral like const is. Ya gotta commit to it!
2. Phobos is not completely dip1000 compatible, but it's pretty 
close. Druntime is fully dip1000.


dip1000 is pivotal to Dlang's succes, but I don't need to tell 
you about this.


It needs those bugs sorted out and some vigouros leadership to 
shepperd it out of the DIP process and make it default.


I want to use it in a betterC lib I'm developing, but the slow 
progress on bug front at least made me worry about its future. 
Good news is that for some parts it worked nicely, but it is not 
there yet!


Thanks.

Re: dip1000 state

[Radu via Digitalmars-d]

On Friday, 2 March 2018 at 18:07:34 UTC, carblue wrote:

On Thursday, 22 February 2018 at 14:36:10 UTC, Radu wrote:

[...]


The fact that it takes 8 days for any reply, doesn't that say 
something?
@safe is a high ranked technical issue in vision papers (in 
german we say something like "paper is patient"), but when 
trying to turn to reality and push forward dip1000 for phobos 
(as I did in the past ~1.5 month), only a few members like Seb 
really seem to care a lot.


[...]


Sorry to hear about your lousy experience with your PRs.

I hope now that you got Walter's attention things will improve. 
Please remember that for better or worse D is a community project 
so things might not be as smooth as one would expect.
That is not an excuse for not making them better btw, looking at 
you Dlang BDFL.


Nice to see that things are rolling, forget about those ego 
bruises and keep those PR's rolling!

Re: dip1000 state

[ixid via Digitalmars-d]

On Saturday, 3 March 2018 at 09:04:04 UTC, Walter Bright wrote:

On 3/2/2018 10:48 PM, Manu wrote:
I've been arguing for DIP1000 (or something very much like it) 
for
almost 10 years. It's one of the small handful of issues that 
lead me

to the forum in the first place.
'Patience' might not be the right word at this point ;)


It is nice to see some interest in it. I've felt like I've been 
pushing against a wall of indifference for a couple years on it.


The good news is:

1. it's implemented
2. it works
3. it's effective

The bad news is:

1. it's viral like const is. Ya gotta commit to it!
2. Phobos is not completely dip1000 compatible, but it's pretty 
close. Druntime is fully dip1000.


Might it be worth you or Andrei trying to get Kenji back?

Re: dip1000 state

[Walter Bright via Digitalmars-d]

On 3/2/2018 10:48 PM, Manu wrote:

I've been arguing for DIP1000 (or something very much like it) for
almost 10 years. It's one of the small handful of issues that lead me
to the forum in the first place.
'Patience' might not be the right word at this point ;)


It is nice to see some interest in it. I've felt like I've been pushing against 
a wall of indifference for a couple years on it.


The good news is:

1. it's implemented
2. it works
3. it's effective

The bad news is:

1. it's viral like const is. Ya gotta commit to it!
2. Phobos is not completely dip1000 compatible, but it's pretty close. Druntime 
is fully dip1000.

Re: dip1000 state

[Manu via Digitalmars-d]

On 2 March 2018 at 10:07, carblue via Digitalmars-d
 wrote:
> On Thursday, 22 February 2018 at 14:36:10 UTC, Radu wrote:
>>
>> Whould like to know what's the state of dip1000?
>
>
> The fact that it takes 8 days for any reply, doesn't that say something?
> @safe is a high ranked technical issue in vision papers (in german we say
> something like "paper is patient"), but when trying to turn to reality and
> push forward dip1000 for phobos (as I did in the past ~1.5 month), only a
> few members like Seb really seem to care a lot.

I've been arguing for DIP1000 (or something very much like it) for
almost 10 years. It's one of the small handful of issues that lead me
to the forum in the first place.
'Patience' might not be the right word at this point ;)

Re: dip1000 state

[Walter Bright via Digitalmars-d]

On 3/2/2018 10:07 AM, carblue wrote:
I generally already used -dip1000 since DConf2017 and it served me well, until 
about 2 month ago, "by accident" code was committed to std.uni that broke my 
builds, see issue #17961. I invested a lot of time to fix this by PR 6041.
The current state is: I don't know any reason why it doesn't get (can be?) 
merged and now it languishes on page 2 of 3 of PRs and Walter started a new PR 
6212 recently after my PR is ready for weeks. In total 4 of my -dip1000 related 
PRs are stuck and mostly for unknown or arguable reasons.


For reference, I submitted:

  https://github.com/dlang/phobos/pull/6212

which I remarked was puzzlingly different from your PR:

  https://github.com/dlang/phobos/pull/6041

I would appreciate your advice on that as well.

I was unaware of your PR. Sometimes, it's worth while to make some noise if 
you're feeling overlooked. Both Andrei and I are way overloaded, and I generally 
defer to Andrei to watch the Phobos PRs.


For reference, here are your open PRs:

  https://github.com/dlang/phobos/pulls?q=author%3Acarblue+is%3Aopen

You've had 4 others that were pulled in the last month; you haven't been totally 
ignored:


  https://github.com/dlang/phobos/pulls?q=author%3Acarblue+is%3Aclosed

I do appreciate the work you're doing to get -dip1000 working with Phobos. It's 
important!

Re: dip1000 state

[carblue via Digitalmars-d]

On Friday, 2 March 2018 at 18:17:02 UTC, Paolo Invernizzi wrote:

On Friday, 2 March 2018 at 18:07:34 UTC, carblue wrote:

(It may be absolutely unrelated, but there once was a very 
productive and knowledgeable compiler et. al. contributor, 
9rnsr, Hara Kenji; though not contributing to dmd since ~ 1.5 
years any more, he's still ranked #1 in number of 
contributions; I think, he's a busy professor at a Tokyo 
university, but I'm really curious to know why he stopped 
coding; I guess, dmd improvement speed still suffers from his 
decision).


https://github.com/dlang/dmd/pull/5239
https://github.com/dlang/dmd/pull/5304


Many thanks for the pointers, Paolo.
Just from what I grasped from a short read: It happened that 
"high-value" contributor Hara Kenji got pissed off in a 
discussion about code formating style (and probably more 
background I didn't read about): Incredible, that there was no 
way for de-escalation. It seems to support my (3.), but I won't 
appoint myself as a judge and we don't want to lose our BDFL.

At least I understand my (1.) better now.

Re: dip1000 state

[Paolo Invernizzi via Digitalmars-d]

On Friday, 2 March 2018 at 18:07:34 UTC, carblue wrote:

(It may be absolutely unrelated, but there once was a very 
productive and knowledgeable compiler et. al. contributor, 
9rnsr, Hara Kenji; though not contributing to dmd since ~ 1.5 
years any more, he's still ranked #1 in number of 
contributions; I think, he's a busy professor at a Tokyo 
university, but I'm really curious to know why he stopped 
coding; I guess, dmd improvement speed still suffers from his 
decision).


https://github.com/dlang/dmd/pull/5239
https://github.com/dlang/dmd/pull/5304

Re: dip1000 state

[carblue via Digitalmars-d]

On Thursday, 22 February 2018 at 14:36:10 UTC, Radu wrote:

Whould like to know what's the state of dip1000?


The fact that it takes 8 days for any reply, doesn't that say 
something?
@safe is a high ranked technical issue in vision papers (in 
german we say something like "paper is patient"), but when trying 
to turn to reality and push forward dip1000 for phobos (as I did 
in the past ~1.5 month), only a few members like Seb really seem 
to care a lot.


Lately I haven't noticed much activity on it, and at least on 
the bug front there are about 28 entries opened:


https://issues.dlang.org/buglist.cgi?quicksearch=dip1000%20OR%20%5Bscope%5D_id=219758


Look at my #18444 and it's "Depends on" and exclude 18444 from 
counting: It's just a tracking/organizational issue intended to 
i.a. perhaps add help for "dmd coder's" priority. No observable 
reaction there since Feb 15 so far. And there are others that I 
wouldn't take into account as bugs. My first reaction looking at 
this list again now is: Most issuers expect "somethíng to happen" 
on the dmd part.


I'm asking this as I am eagerly waiting for it for about 1 
year, wanting to use it for a new project that is in pipeline, 
and no major progress has been made since last year on both 
finalizing the spec and using the implementation in 
Phobos/druntime.


IIRC, druntime is compiled -dip1000 already and for phobos, have 
a look into the aa settings in PR 6195: Only "a few" phobos 
modules are not -dip1000 compilable, currently.


I'm also awaiting -dip1000 compilable phobos, and we are now 10 
month past Walter Bright's DConf2017 talk "pointers gone wild", 
where he said "-dip1000 is implemented already" (omitting what's 
missing). From an eMail conv. with a member I know, Walter is 
aware of missing pieces in DIP1000.md and implementation and that 
it's on his TODO-list, probably as many other things are as well.
But, except for special cases, there is no reason not to use 
-dip1000 today:

-dip1000 implementation is far better than it's reputation.

I generally already used -dip1000 since DConf2017 and it served 
me well, until about 2 month ago, "by accident" code was 
committed to std.uni that broke my builds, see issue #17961. I 
invested a lot of time to fix this by PR 6041.
The current state is: I don't know any reason why it doesn't get 
(can be?) merged and now it languishes on page 2 of 3 of PRs and 
Walter started a new PR 6212 recently after my PR is ready for 
weeks. In total 4 of my -dip1000 related PRs are stuck and mostly 
for unknown or arguable reasons.
Knowing my experience now, You won't expect me to contribute for 
-dip1000 any longer, but of course, You may do so and perhaps 
have more luck.
There's a nice german children's song that comes to my mind here: 
"Ein Loch ist im Eimer".

known in english as: "There's A Hole In My Bucket".
It just leaves me stunned, shaking the head.

Are there major roadblocks ahead? Or is just a matter of 
prioritization/budget?


I would say (biased as I am), these are roadblocks:
1. Lack of (sufficient/knowledgeable) compiler contributors.
2. Some lack of coordination/enforcement/management: vision <-> 
make it happen.
3. A tendency to scare away (impatient,) possibly long-time 
contributors, i.a. - as I perceive that - by lack of 
appreciation, instead taking for granted volunteer contribution 
(which shows up in complete unresponsiveness sometimes: When I 
spend e.g. > 1 day on a non-trivial D-Programming-Deimos issue 
and there is no reaction to my PR for 2 month, not even a 1 
minute response like "sorry no time" or "doesn't LGTM 
because...": That's at least impoliteness (I take it as lack of 
appreciation) that I won't accept any more. On the other hand, 
for "trivial" PRs, You may get very quick replies from 2 members.
From DConf2017 talk "Systems Programming Panel" I know, I'm not 
alone being discouraged.


(It may be absolutely unrelated, but there once was a very 
productive and knowledgeable compiler et. al. contributor, 9rnsr, 
Hara Kenji; though not contributing to dmd since ~ 1.5 years any 
more, he's still ranked #1 in number of contributions; I think, 
he's a busy professor at a Tokyo university, but I'm really 
curious to know why he stopped coding; I guess, dmd improvement 
speed still suffers from his decision).


Though this might have raised more questions than giving answers 
(to the time schedule): Was it helpful?