On Wed, 16 Oct 2024, 01:51 Meta via Digitalmars-d, <digitalmars-d@puremagic.com> wrote:
On Tuesday, 15 October 2024 at 09:07:12 UTC, Manu wrote:
> So, since I've been off the wagon for a couple of years;
> VisualD, which
> used to be ROCK SOLID seems to have suffered major regressions
> in almost
> every aspect of its functionality.
> I think it's largely related to dmd-as-a-lib now being the
> foundation for a
> lot of tooling, and it's just criminally unstable...
>
> Rainer used to maintain his own semantic analyser used for
> formatting,
> auto-complete and suggestions, code navigation, and debugging;
> it worked
> beautifully! But a couple years back, VisualD was switched to
> use DMD
> frontend for those duties, and it barely works anymore.
> The old bespoke code is still available, but it's so out of
> date with the
> modern language that it's not usable anymore.
>
> ...to make this worse; Rainer has effectively checked-out too.
> We've lost another one of our finest.
>
> This is a general category of problem that's been an issue for
> a long time; having unfunded one-man efforts maintain essential
> infrastructure. I wonder if there are any opportunities
> available to do a lot better here. Does the dlang foundation
> have any budget for critical infrastructure? And/or anyone that
> would even consider working on the boring but essential stuff?
>
> Is there actually anyone here who develops on Windows? I don't
> understand how it could have regressed so far, unless it's just
> that nobody is using it.
>
> I now recognise a really major conundrum; I've recently
> returned to D to
> start a company with a greenfields project. VisualD failing is
> essentially
> terminal. I'm not sure what to do.
> I don't have time available to try and pick up the project and
> work it
> myself, but the current state is really pushing at the border
> of forcing me
> to completely rewrite all my code in C++ on account of ecosystem
> reliability.
>
> Ideally, we really need to be properly funding development for
> critical infrastructure... but I'm not sure we've ever had a
> sufficient budget to maintain that sort of commitment.

What specific problems are you hitting? I don't write much D
anymore (or do really *any* hobby programming, really), but I did
recently dust off one of my old projects and get it compiling
again with the latest dmd. I'm on Windows 11 and using Visual
Studio with Visual D, and besides it not recognizing some stuff
like the shortened method syntax, I haven't really noticed any
issues.

Syntax highlighting takes 10s of seconds, sometimes minutes, and may never complete. 
Go-to definition for most symbols don't work. 
Press '.' after an aggregate or an enum or whatever, the list that pops up rarely if ever shows any appropriate members or completion suggestions. Ctrl-space (complete what I'm typing) doesn't work anymore.

I think it's all related to the new DMD-based semantic analysis bailing out in various situations that I don't understand.

The debugger has significant problems to; important debug features that do essential stuff like show the string for a custom string type have been locking up under unpredictable circumstances.

It's all just very unstable and unreliable in general. Again I don't mean to criticise Rainer; it's been rock solid for the past decade, but the enormous scale of regressions since I was last actively using it really demonstrates we have extreme key-person risk with this stuff... and this is deal-breaker stuff. If Visual D doesn't work well, I will migrate my company to C++; there is no practical alternative.

I've spent 15+ years trying to see D move beyond a hobby/curiosity/toy, but this stuff shows we're not in good shape even after such a long time :/