www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.announce - D Language Foundation December 2025 Monthly Meeting Summary

reply Mike Parker <aldacron gmail.com> writes:
The D Language Foundation's December 2025 monthly meeting took 
place on Friday the 13th and lasted about an hour and thirty-five 
minutes.

Unfortunately, I had the wrong configuration selected in my 
screen recording software. My system audio was muted and I only 
have the recording for my microphone. As such, I can't put 
together the kind of detailed summary I like to write up. The 
best I can do is provide some highlights reconstructed from the 
chat log, my audio, and memory.



The following people attended:

* Walter Bright
* Iain Buclaw
* Rikki Cattermole
* Jonathan M. Davis
* Dennis Korpel
* Mathias Lang
* Mike Parker
* Razvan Nitu
* Átila Neves
* Robert Schadek
* Nicholas Wilson






Before we got started on the agenda items, I gave everyone a 
quick update on DConf '26. Since our first time at CodeNode in 
2022, which was during a period when they were ramping back up 
post-pandemic, they had offered us the same significant discount 
for any dates in August every year, and a smaller discount for 
September. This time, they had extended the August discount rate 
into the first week of September. I had already confirmed with 
Symmetry that they were okay with that week, so CodeNode was 
holding the dates for us. The next step was for Symmetry to sign 
the contract with Brightspace, our event planners, so they could 
in turn sign the venue contract.

(__UPDATE__: In case you missed the subsequent announcement, 
[DConf '26 is happening September 2 - 
4](https://dconf.org/2026/index.html) at CodeNode in London.)



The first agenda item was from Rikki regarding his idea for Type 
System Analysis Algorithms. [He posted in the forums about 
it](https://forum.dlang.org/post/waerqsesrohtokjctnyj forum.dlang.org) a couple
of weeks before the meeting, so you can read about the the concept there.

This took up over 30 minutes of the meeting. Rikki posted the 
following links in the chat log:

__Fast and sound static analysis__: 
https://www.absint.com/astree/index.htm
__Principles of Abstract Interpretation__: 
https://mitpress.mit.edu/9780262044905/principles-of-abstract-interpretation/

I recall there was a lot of skepticism about this. At one point, 
we were talking about the refactoring going on aimed at preparing 
to enable running the compiler as a daemon. I warned that we 
should keep that in mind when discussing major structural changes 
to the compiler source. Beyond that, I have no details of the 
discussion. I can say the proposal was neither accepted nor 
rejected.

I ultimately asked Rikki if we could put a pin in it and move on 
to the other agenda items, and he said yes.



Walter wanted to know about the status of the editions 
implementation. The conversatoin veered into what the first 
edition was supposed to encompass. I reminded everyone we had 
agreed some time ago that the first edition would be about fixing 
incomplete and broken features.

We decided to have a separate meeting to finalize the details 
about the first edition, e.g., which incomplete or broken 
features we wanted to include. Given the holiday season in 
December an a quarterly meeting in January, we opted to hold the 
editions meeting in late January. I'll summarize that in a 
separate writeup.



Nicholas had [a PR to implement with 
expressions](https://github.com/dlang/dmd/pull/22196), e.g., 
`with(auto x = whatever())`, a feature that had been requested in 
the forums. The only issue with it was [a BuildKite 
failure](https://github.com/dlang/dmd/pull/22196#issuecomment-3627316856) with
one of Átila's projects where he used `with(immutable S())`.

I recall someone mentioned that `if(immutable S())` failed to 
parse, as did `for`. The discussion focused on which of three 
possible options to pursue:

1. merge the PR as is, with the breakage;
2. change the PR so that the breakage doesn't occur;
3. go with Option 2 and also fix the parsing errors in `if` and 
`for`.

As a first step, we opted for Option 2. You'll find this [in the 
changelog for the next DMD 
release](https://dlang.org/changelog/pending.html#dmd.with-assign):


 For backwards compatibility, `with` still also accepts a 
 qualified expression without assignment to a variable as in 
 `with(immutable expression())`.


We also decided someone should look at the parser and see what 
sort of changes are required to get `if` and `for` to behave the 
same way. If it's not too gnarly, we can make `if` and `for` 
conform with `with`. Otherwise, we might modify `with` in a 
future edition to conform with `if` and `for`.



Walter wanted to know when the next DMD release was coming. This 
led to a discussion about the release process in general. As part 
of that, we touched on code signing certificates. Thanks to Luna, 
we should be okay on Mac releases. The problem was Windows.

We haven't had a certificate for our Windows releases for quite a 
while. Cost was part of the problem, but more importantly was the 
requirment for a dongle. There was a conflict there with our 
desire to automate the release process. It was doable, but a bit 
of a headache.

We had discussed this before and decided we needed to get the new 
release process in place first. Then we could look at options we 
could integrate and their costs.

(__UPDATE__: [DMD 2.112.0 was released on January 7, 
2026](https://dlang.org/changelog/2.112.0.html).)



Before we left, I let everyone know that Fei, one of Ben Jones's 
students, was interested in working with us as an unpaid intern 
for a few months. He had participated in GSoC 2025 working on 
Steve's jsoniopipe. He was interested in whatever work we could 
throw his way.

(__UPDATE__: There was some red tape to sort out first, but Fei 
officially started working with us a few weeks later. He has 
since joined a couple of monthly meetings, and Nicholas Wilson 
has been keeping him occupied.)



Rikki had one other agenda item we discussed briefly, a 
suggestion for a GSoC project to implement cent/ucent integers, 
but I have no context for, or memory of, that conversation.

The first Friday of January is usually when we have a quarterly 
meeting, but given that this year the first Friday was on January 
2nd, I suggested we push it back to the 9th, and our January 
monthly in turn to the 15th. Then we could have our editions 
meeting on the 23rd. Everyone agreed.

If you have anything you'd like to bring to us in a monthly 
meeting, please let me know.
Apr 30
next sibling parent Sergey <kornburn yandex.ru> writes:
On Thursday, 30 April 2026 at 12:02:52 UTC, Mike Parker wrote:

 The D Language Foundation's December 2025 monthly meeting took 
 place on Friday the 13th and lasted about an hour and 
 thirty-five minutes.


Happy New Year everybody!
Congratulations with the milestone :)
Apr 30
prev sibling parent reply jmh530 <john.michael.hall gmail.com> writes:
On Thursday, 30 April 2026 at 12:02:52 UTC, Mike Parker wrote:

 The D Language Foundation's December 2025 monthly meeting took 
 place on Friday the 13th and lasted about an hour and 
 thirty-five minutes.

 Unfortunately, I had the wrong configuration selected in my 
 screen recording software. My system audio was muted and I only 
 have the recording for my microphone. As such, I can't put 
 together the kind of detailed summary I like to write up. The 
 best I can do is provide some highlights reconstructed from the 
 chat log, my audio, and memory.


Thanks as always for these, regardless of technical difficulties.

My recollection is that you use transcribing tools to help. Do 
you make any use of the LLMs for summarizing? Might help reduce 
the time commitment.
Apr 30
parent Mike Parker <aldacron gmail.com> writes:
On Thursday, 30 April 2026 at 16:25:57 UTC, jmh530 wrote:


 Thanks as always for these, regardless of technical 
 difficulties.

 My recollection is that you use transcribing tools to help. Do 
 you make any use of the LLMs for summarizing? Might help reduce 
 the time commitment.


Yes, for a while now. The output has become usable only recently. 
Before, it required so much revision that it didn't help much.
Apr 30