00:59:26 -!- indigaz24 is now known as indigaz2
01:10:39 <Discord|> <12a​sciiphilia> I thought we would maintain compatibility by completely removing RequireJS. I haven’t had a chance to properly read the PR yet. DWEM works by hooking into RequireJS’s loading process, dynamically parsing scripts through AST analysis, and injecting specific code snippets into the desired patching context. This is because it was designed to create a plugin structure that can continuously adapt to code that updates in real time,
01:10:40 <Discord|> like the trunk. Would it be possible for the mainline code to support this kind of functionality? It feels rather hacky to me.  In any case, once the new WebTile JavaScript part stops using RequireJS, hooking will no longer be possible, and with things like bundling being introduced, it seems that the increased encapsulation will make it difficult to rebuild this kind of mechanism.  I’ll likely receive a lot of questions about why the broken DWEM
01:10:40 <Discord|> features can’t be used anymore, but I don’t have enough time to handle all those issues and migrate everything to the new code structure. Porting my code structure to the mainline would mean losing some of the custom server functionality I need, and I won’t be able to port everything anyway. I also don’t have enough time for that work. At least for me, this is shaping up to be a maintenance nightmare.
01:14:17 <Discord|> <12a​sciiphilia> Since DWEM is a third-party project, I have no intention of pressuring the main development direction with this. I just wanted to explain that this is the situation on my end.
01:18:27 <Discord|> <06m​umra> > I thought we would maintain compatibility by completely removing RequireJS.  Not quite sure what you mean here -- the old versions still expect requirejs to be present in order for them to both load their own scripts as well as injecting the client/comm/key_conversion modules which are necessary for interaction with the server.  This means the API of those three components (which are what it seems you are largely hooking into with DWEM)
01:18:27 <Discord|> will remain unchanged. (And even if we wanted to significantly restructure them, we'd still have to provide a compatibility wrapper for legacy versions)
01:20:40 <Discord|> <06m​umra> It definitely sounds like the AST method will potentially fail once it tries to look at minified code; although, given that none of the public exports of the core modules have actually changed, I am wondering how necessary this was in the first place? And what additional hooks we might have to expose so you can do what you need to do without having to do something so hacky? 🙂
01:24:47 <Discord|> <06m​umra> I've just been having a bit of a look through most of the modules in your DWEM repository. It seems like for the most part there's nothing terribly complicated going on in most of them (unless I've missed some really thorny detail that necessitates doing things that a public interface can't provide)
01:26:07 <Discord|> <06m​umra> In some cases (in particular, the Advanced RC Editor, and the advanced chat) I'm wondering if you would perhaps consider raising a PR to get these into master? Something that will be a LOT easier with the modernised js stack.
01:26:24 <Discord|> <06m​umra> The RC editor particularly just seems like a welcome improvement
01:27:18 <Discord|> <12a​sciiphilia>  _define = define define(moduleArray, moduleFunc) {     ast = ast_library.parse(moduleFunc.toString())     location = ast.findCodeLocation(condition)     addCustomCode(ast, location, customCode.toString())     _define(moduleArray, eval(ast.toCode())) }  This is sudo-code of DWEM core patch logic
01:29:38 <Discord|> <12a​sciiphilia> Once a module is loaded using the import statement, it cannot be interfered with.
01:34:46 <Discord|> <12a​sciiphilia> I thought the translation was incorrect.  I understood it as if you were trying to implement an entirely new loader by removing requirejs altogether. (In the previous version, I thought it wasn’t about selectively loading with requirejs, but about loading everything using something completely new.)
01:38:01 <Discord|> <12a​sciiphilia> Currently, the object exported (returned) by the webtile module only includes what’s necessary for the webtile itself. However, the internal variables used by DWEM go far beyond that scope. We need access to many internal variables that are not exposed outside the module. But I’m not sure if exposing variables that aren’t otherwise used, just for the sake of external hooks, is the right approach from a design pattern
01:38:02 <Discord|> perspective.
01:39:20 <Discord|> <06m​umra> Yes that's correct
01:41:28 <Discord|> <06m​umra> Basically new versions will be loaded just by inserting a single <script> tag on the page that loads the bundle. The DCSS interface is made available on window.  Old versions will still use requirejs and get hold of the DCSS interface that way.
02:00:14 <Discord|> <06m​umra> I feel like "providing hooks to allow server admins to customise the experience" isn't by itself a bad design, they just need considering on a case by case basis
02:04:33 <Discord|> <12a​sciiphilia> In conclusion, fact is that DWEM is technically dead now. Because of my graduate studies, I no longer have the time or motivation to maintain it — it will probably just remain broken and abandoned. I can’t really say anything about the development direction aiming to improve the codebase. And the suggestion to contribute my code to the mainline doesn’t really mean much to me at this point. Yes… it’s just unfortunate.
02:06:36 <Discord|> <12a​sciiphilia> Efforts to translate on the client side also ended in failure.
02:07:11 <Discord|> <12a​sciiphilia> wtrec will no longer work either.
02:15:19 <Discord|> <12a​sciiphilia> It feels a bit bittersweet — it was my mistake to assume that RequireJS would remain unchanged for the sake of backward and forward compatibility.
02:34:48 <Discord|> <06m​umra> I feel like the "correct" way to support this kind of extensibility, in a way compliant with modern JS design, would be for us to actually publish the lobby package to npm, so a server admin can create their own custom environment by importing the library and wrapping around that with their own functionality and building it themselves. Rolling their own client basically. Then providing additional hooks in the game modules which can be
02:34:48 <Discord|> plugged into as needed. Of course this only then works for current and future versions, and I can see the appeal of DWEM that it worked for all past versions as well.  Of course, the AST method of injection was always very fragile (with or without requirejs) and any slight internal change could have broken things too. If public APIs were actually exposed for some of these use cases, you'd have an actual guarantee that they will stay supported.  I do
02:34:49 <Discord|> think it was a really novel solution and really impressive. And it's a shame you don't have more time 🙂 but it's how life is. But I don't know if it's as impossible as you claim (and really, I'm looking at the modules, and many of them can easily work without the AST injection). Maybe such is the nature of open source, someone from the CNC community might step up if there is enough motivation to update these features?
02:43:19 <Discord|> <06m​umra> But yeah ... some kind of rewrite of this nature has been discussed on and off for a long time. There was already edlothiol's nearly complete rewrite in React/TypeScript. Plus my experiment with emscripten more recently. We can't reasonably be expected to keep webtiles development stuck 10+ years in the past to guarantee compatibility with requirejs unforunately 🙁
02:51:12 <Discord|> <06r​egret-⸸nde※> With our severely limited active dev pool and that long history of people clearly not deigning to rewrite the ever-terrible webtiles implementation, I'm fairly more inclined towards getting to use what does exist now (DWEM's new features attached to the current webtiles implementation), rather than what does not and continues to not exist (re-implemented features made by Somebody Else after a full rewrite after this
02:51:13 <Discord|> Javascript modernizing), especially when it can address ongoing current tangible issues (like providing much better rc file editing UI for newcomers).
02:53:08 <Discord|> <06m​umra> (Honestly, we could skip the minification stage of the bundler. It wouldn't add a super big weight to the bundle. Then the AST technique might work still, you would just have to intercept the entire game.js script instead of hooking into requirejs)
03:00:09 <Discord|> <04d​racoomega> Like, I am not even slightly qualified to evaluate the technical end of any of this, but it seems a very meaningful downside if whatever changes here would break compatibility with webtiles extensions that were popular enough to get repeated mention in #dcss as things mainline would benefit from having.
03:01:58 <Discord|> <04d​racoomega> Especially given that it's unclear to me what the short-term benefit of these changes are besides 'moderately improved load times'. I am sure there are other things (as I've said, my webdev knowledge is basically zero), but I'm not sure how many of these are 'theoretical future benefit' that may or may not materialize
03:08:50 <Discord|> <06m​umra> Well the primary benefit is that there is actually some error checking on the JavaScript. There are many further benefits to any developer actually wanting to modify the scripts.
03:14:10 <Discord|> <06m​umra> (Specifically regarding the rc file editor: having looked at that module, it'd be pretty straightforward to bring that into master under the new build stack in place. Which would roll it out to everyone, where if server admins want to support it via DWEM they have to go to a lot of extra steps to install a 3rd party modification system that runs on some impressive black magic but is almost impossible to understand and very fragile...)
03:18:31 <Discord|> <06m​umra> And yeah potential future benefits in terms of possibly more developers actually being willing to contribute to webtiles are certainly theoretical, but it is certainly the case that right now the state of the existing JavaScript would make any developer versed in modern syntax run a mile away...
03:20:18 <Discord|> <06m​umra> (I wonder had the codebase been more amenable in the first place, would ASCIIPhilia have maybe thought about contributing some further improvements directly rather than going down this whole injection route, which honestly was not easy?)
03:23:41 <Discord|> <12g​e0ff> with linting, it should be possible to easily catch errors like the recent TILE_WALL_STONE_CRACKLE_1 problem, which took you two commits to fix
03:25:50 <Discord|> <04d​racoomega> Oh?
03:26:08 <Discord|> <12g​e0ff> %git 22c0543b7eb
03:26:09 <Discord|> <04C​erebot> DracoOmega * 0.34-a0-933-g22c0543b7e: Actually fix the aforementioned issue properly (ASCIIPhilia) (9 days ago, 1 file, 1+ 1-) https://github.com/crawl/crawl/commit/22c0543b7eb1
03:27:18 <Discord|> <04d​racoomega> No, no, I recall what you're talking about. I just didn't realize this would change things on that front.
03:28:47 <Discord|> <04d​racoomega> Like, when I said that I don't understand the consequences of mumra's branch, I really mean that I don't have a clue. The only JS or web code of any sort that I've written in my life has been the small bits of webtiles code I've had to touch.
03:30:48 <Discord|> <12g​e0ff> fr: a short overview of immediate and future benefits of leaping forward 12-15 years in the web tech for https://github.com/crawl/crawl/pull/4839
03:32:41 <Discord|> <12g​e0ff> (mumra has mentioned some of them here and on the other Discord, but some things might not be obvious to non-web devs, so it would be mega useful to have all that in one place)
03:33:14 <Lantell> Experimental (bcrawl) branch on underhound.eu updated to: 0.23-a0-5261-gd9800d219b
03:35:33 <Discord|> <12a​sciiphilia> I think the developer pool at CNC is smaller than that of other servers. I think this probably stems from structural issues in culture and the way games became popular.
03:38:22 <Discord|> <06m​umra> Yeah this is really just a starting point of "get things moved onto a stack where we can use new features".  I stopped short of more thorough changes that would go a lot further in terms of ensuring code correctness (for instance, in the current form, I don't think it would catch the TILE_WALL_STONE_CRACKLE_1  , because the enums export is composed in such a way that the js interpreter still has no knowledge about the structure of
03:38:22 <Discord|> that export). I really wanted to get the build working, and passing a "recommended" standard lint ruleset (which, already added a lot in terms of code quality, e.g. enforcing const vs let, and it did fix a number of iffy issues such as variables being used uninitialised in a few places, which is now an error in strict mode, and elsewhere something like using isNaN to actually test if something was an array, which was obscure enough that it took me a
03:38:22 <Discord|> good 15 minutes to figure out what that code was supposed to be doing)
03:38:57 <Discord|> <06m​umra> Unfortunately some of these fixes aren't really visible in the PR as so much was changed at once by the automated pass, and I ended up squashing that together with some manual fixes done after
03:39:08 <Discord|> <12a​sciiphilia> Well, my motivation for developing DWEM wasn’t only about that. The features in DWEM (such as SoundSupport and AdvancedRCEditor) were created not just for Crawl users in general, but also specifically for CNC. At a manageable level of deployment and maintenance, I needed a tool that could quickly and non-invasively modify code without directly altering the vanilla core.
03:42:44 <Discord|> <12a​sciiphilia> And since JavaScript can be modified on the client side anyway, I wanted users to be able to customize their game as much as they wanted. (In fact, DWEM supports user module loading, not just modules installed on the server.)
03:43:58 <Discord|> <06m​umra> > I don't think it would catch the TILE_WALL_STONE_CRACKLE_1  (Although I should note, I was very close to rewriting enums,js specifically to catch this type of thing, not being aware of that specific one)  The next big step would be converting the code to typescript at which point you'd have full typesafety across the whole js codebase
03:44:59 <Discord|> <12a​sciiphilia> Actually, I think things like using a linter or TypeScript would still be possible without abandoning RequireJS.
03:45:40 <Discord|> <06m​umra> Sure but why wouldn't you want to use import / export syntax? You used it yourself on DWEM modules 🙂
03:46:51 <Discord|> <06m​umra> (Although, I don't think TypeScript would work nicely with RequireJS. There would be no way of knowing the type of the injected object)
03:49:09 <Discord|> <12a​sciiphilia> That’s only natural — each module in my code isn’t tied to RequireJS anyway.
03:49:31 <Discord|> <06m​umra> You would have to manually cast the type of every object you import, which would make things very error prone and a lot of extra boilerplate just to import something
03:50:12 <Discord|> <06m​umra> Nothing was ever "tied to requireJS", it was just that was the only module loader system available at the time webtiles was originally written
03:50:28 <Discord|> <06m​umra> My point is if you write something from scratch nowadays, you would never consider using requirejs
03:52:56 <Discord|> <12a​sciiphilia> That feels like little nitpicking. I already know that.
03:53:33 <Discord|> <12a​sciiphilia> I agree with this opinion.
03:54:31 <Discord|> <06m​umra> Sorry I didn't mean it to sound like that. I was just saying that it was natural for you to use import/export syntax to modularise your code. Which is exactly the same decision I would make myself for any new project, and what any JavaScript developer would expect to see these days when looking at a project's source code.
03:55:16 <Discord|> <06m​umra> So that is the desire to move crawl to the standardised module syntax, it's just what I would expect to see if I was to see the project for the first time
04:22:58 <NotCheibriados> 03mumra02 07https://github.com/crawl/crawl/pull/4839 * 0.34-a0-992-g00bca500c8: Advanced rcfile editor (ASCIIPhilia) 10(24 seconds ago, 7 files, 2735+ 3-) 13https://github.com/crawl/crawl/commit/00bca500c8da
04:25:12 <Discord|> <06m​umra> Thought I'd just have a go at integrating the rc editor myself, to demonstrate how much easier it is to actually extend things after this upgrade
04:27:01 <Discord|> <06m​umra> https://cdn.discordapp.com/attachments/747522859361894521/1429793022660317294/image.png?ex=68f76e04&is=68f61c84&hm=930cf31cb0fd6919176c1f332e8b21e551d922b95805f035c6050e6cf4fcca36&
04:27:19 <Discord|> <12a​sciiphilia> Well, AdvancedRCEditor is one of the DWEM modules that has the fewest connections to the main code.
04:28:00 <Discord|> <12a​sciiphilia> Still, it's good that the expansion speed is much better than before.
04:28:32 <Discord|> <06m​umra> Certainly, that's why it was easy to drop in (it is also not that much code as most of the lifting is done by ace editor)
04:29:59 <Discord|> <06m​umra> (It's not quite finished, and I need to autogenerate the completions.json from options txt; it will also be great that it stays in sync with trunk)
04:30:26 <Discord|> <06m​umra> Actually I would argue that a DWEM module that has a lot of connections to code is a lot easier to implement in trunk, if it's a good enough module that users generally want it
04:32:08 <Discord|> <06m​umra> Thanks to the bundler I could also use a dynamic import, so it automatically gets bundled separately to the main bundle and won't actually load until the rc editor is opened
04:33:03 <Discord|> <12a​sciiphilia> I’m not sure what it would look like to separate the lobby into an npm package,   but here are a few of my ideas and suggestions.   When I’m able to work on it in the future, I’d like it to take this kind of form:  0. Coexist with the current requirejs method until the plugin structure is complete and the rewrite is finished (if we want to accommodate CNC’s convenience).   1. Define the plugin structure.   2. Officially
04:33:04 <Discord|> support the ability to dynamically redefine functions or variables within each class without using AST-based modification methods.   3. Convert all requirejs modules into class-based structures.   4. Turn as many local variables as possible into class member variables.   5. Allow users (not just server admins) to inject plugins — though server admins should be able to configure whether this is allowed. (This is a relatively weak suggestion.)   6.
04:33:04 <Discord|> Provide code support similar to IOHook, RCManager, ModuleManager, and CommandManager.  If the WebTile code evolves in this direction, I could manage the CNC module part as a plugin in my own repository,   while contributing the general-purpose feature modules to the mainstream as plugins — and eventually merge them into the main codebase.   (Alternatively, it could just be developed as part of the main code from the start.)
04:44:21 <Discord|> <12a​sciiphilia> If the codebase can accommodate detachable, plugin-style code that allows for easy contribution even at a moderate level of completeness, it will likely attract much more participation than it does now.
05:19:29 <Discord|> <06m​umra> There is something interesting I missed: vite does have an option for amd output (i.e. requirejs-compatible modules). So maybe it's possible to have the cake and eat it.
05:21:36 <Discord|> <06m​umra> But regarding the above, I would firstly say: I definitely would not have the time or energy for such a major rewrite. And if I did, I feel it would be better to spend that time really rewriting things into a functional component-based architecture (i.e. React), where you could then use those components to compose your own lobby etc. if needed, rather than trying to make everything a class and everthing overridable. Software design
05:21:37 <Discord|> hides implementation details for a reason; otherwise you're tying yourself (and the core project) in ever worse knots as it becomes impossible to change anything without breaking someone's module.  Much better to have well-defined integration points and as few of them as possible to effectively achieve what you need.
05:23:23 <Discord|> <06m​umra> How I imagine an ideal extension system looking would be various event handlers you can subscribe to to receive internal state and (sometimes) to be able to modify it. IO is an obvious example; comm.js could very easily expose an event handler so any extension can easily listen in on any messages being sent/received (and we can add some extra security considerations because of course we want to filter out any messages that might
05:23:24 <Discord|> container a user's password)
05:24:20 <Discord|> <06m​umra> Actually you could already do this (in the new framework) just by monkeypatching window.DCSS.comm; no need for any AST rewrite
05:28:07 <Discord|> <06m​umra> i think "ability to dynamically redefine functions or variables within each class" is a VERY dangerous road to go down honestly and not an extension architecture i would even want to encourage 🙂
05:32:55 <Henzell> Unstable branch on crawl.akrasiac.org updated to: 0.34-a0-976-g34f68e3 (34)
05:32:59 <Discord|> <12a​sciiphilia> I don't know how much new comm framework you are talking about is capable of modification, but at least in the current situation, it is a story that can be said to achieve the requirements for implementation.
05:33:44 <Discord|> <12a​sciiphilia> If comm is well defined, the situation might be different.
05:41:56 <Discord|> <06m​umra> comm hasn't changed, the implementation hasn't changed, its public interface hasn't changed
05:42:08 <Discord|> <06m​umra> there's no "new comm framework", it's the same comm object as before
05:42:26 <Discord|> <06m​umra> comm.send_message is called when we send a message over websockets
05:43:10 <Discord|> <06m​umra> comm.handle_message_immediately and then also comm.handle_message (if the message wasn't handled immediately) are called when a message is received over websockets
05:43:32 <Discord|> <06m​umra> so you can patch those three methods with your own implementation to intercept and transform messages in any way you like
05:44:27 <Discord|> <12a​sciiphilia> Then, it seems that most modules, except for RCEditor, would be difficult to use on the new framework. There are things that require changing the behavior of comm or client, as well as the behavior of the scripts in the game/*.js directory.
05:44:32 <Discord|> <06m​umra> of course if we had typescript definitions for module exports (with tsdoc comments), things would be even better defined, but this has always been the case
05:45:01 <Discord|> <06m​umra> I don't see how that is true, I am literally telling you how you can COMPLETELY change the behaviour of comm
05:46:21 <Discord|> <06m​umra> then, most of client's behaviour is implemented by calling comm.register_handlers or comm.register_immediate_handlers
05:46:46 <Discord|> <06m​umra> so you can actually just rewrite client's behaviour by hijacking those methods on comm
05:48:33 <Discord|> <06m​umra> this could be slightly improved by providing a better event-based model, but this works right now and still works on the new build, without ever having to deal with ASTs
05:49:26 <Discord|> <12a​sciiphilia> https://github.com/refracta/dcss-webtiles-extension-module/blob/main/modules/cnc-public-chat/index.js
05:49:51 <Discord|> <12a​sciiphilia> Can this structure be implemented using only comm hijacking?
05:52:33 <Discord|> <06m​umra> It seems like the main integration point you use here is IOHook.handle_message ?
05:53:41 <Discord|> <12a​sciiphilia> No, I have to overwrite a lot of the original code's behavior with code like chatInjector.
05:54:41 <Discord|> <12a​sciiphilia> With comm alone, you can't modify the behavior that the webtile was originally intended to do.
05:55:42 <Discord|> <12a​sciiphilia> The injector is much more important than the input/output hooks.
06:14:30 <Discord|> <06m​umra> I mean ... Firstly i think this (along with a bunch of your other chat improvements) would be great stuff to have in trunk, where it'd be way easier to achieve full stop (and wouldn't require an additional websockets server to handle the public room)
06:36:04 <Discord|> <06m​umra> Ok I've had more of look at what your injector is doing and I'm pretty confident it can also be achieved by a combination of both intercepting the comm handlers as before, plus patching the chat export (e.g. to override the clear function)
06:37:19 <Discord|> <06m​umra> I didn't expose the chat export but this can be trivially done (it can be made available via require, as well as putting it on the global DCSS namespace)
06:38:15 <Discord|> <12a​sciiphilia> I'm not just talking about that one module, that type of code won't work with just comm.
06:38:45 <Discord|> <12a​sciiphilia> And I know that the implementation you're talking about is more ideal.
06:40:59 <Discord|> <12a​sciiphilia> Ultimately, comm alone can't make all my modules work properly, so why do you think it can? Even with just the chat module, you have to expose multiple parts of the chat, right? What if it's another module? We'll have to expose another part, right? I don't know how many modules you're referring to, but this problem isn't limited to just one or two.
06:43:36 <Discord|> <12a​sciiphilia> So, should I just make a PR request to the official repo every time I want to expose some handlers to the public, just to create a feature for my server? This is a bit weird.
06:44:10 <Discord|> <12a​sciiphilia> That's why I said it would be nice to have a feature like a plugin.
06:44:42 <Discord|> <06m​umra> I'm not at my laptop right now but when i get back I'll whip up an example. I dont think you realise how much is already exposed and indirectly hijackable
06:45:46 <Discord|> <12a​sciiphilia> There's no need to do that, I already know that plugins can hijack user cookies and passwords and send them elsewhere.
06:46:49 <Discord|> <12a​sciiphilia> But I think we can definitely create a good structure by compromising.
06:46:54 <Discord|> <12a​sciiphilia> Anyway, something like that is possible by injecting just one TemperMonkey script.
06:49:07 <Discord|> <12a​sciiphilia> Many games support modding frameworks. I think webtiles will be possible too.
06:49:47 <Discord|> <12a​sciiphilia> In a script engine that doesn't compile anyway, encapsulation can be completely broken if you want.
06:51:24 <Discord|> <06m​umra> Yeah it's unfortunate... The only way to secure that would be by moving the login form into an iframe
06:57:51 <Discord|> <06m​umra> It's true but is it something that should be encouraged? It's super bad practise from a security and privacy point of view. And then it ties us in binds: we can't change any internal details of a module without potentially breaking some mod. We can't even rename an internal variable that isn't exposed anywhere, because someone might depend on it. Every implementation detail becomes part of the public API.
06:58:48 <Discord|> <06m​umra> Games that have modding engines have strict interfaces and boundaries around what mods can do, what they have access to, and anything new that you want to be exposed would certainly require a change request or PR.
07:11:27 <Discord|> <12a​sciiphilia> You said we could kill DWEM and just make a few adjustments under your new structure, but in reality, nothing has worked out. If you were going to say that the plugin extension structure request would be rejected due to time constraints and design considerations, you shouldn’t have given us hope in the first place.
07:12:44 <NotCheibriados> New branch created: pull/4841 (2 commits) 13https://github.com/crawl/crawl/pull/4841
07:12:45 <NotCheibriados> 03unutbu02 07https://github.com/crawl/crawl/pull/4841 * 0.34-a0-907-gf5b18652a1: Add the ability to toggle banes in wizmode. To use, type `&}` then the name of the bane. 10(22 minutes ago, 3 files, 39+ 0-) 13https://github.com/crawl/crawl/commit/f5b18652a14d
07:12:45 <NotCheibriados> 03unutbu02 {GitHub} 07https://github.com/crawl/crawl/pull/4841 * 0.34-a0-978-gc560000deb: Merge branch 'crawl:master' into wizmode-bane 10(2 minutes ago, 0 files, 0+ 0-) 13https://github.com/crawl/crawl/commit/c560000debdd
07:14:20 <Discord|> <06m​umra> Well you're take my words wayyy out of context there 🙂 Firstly, I am actively trying to find a way to support DWEM under the new build system, and suggesting many alternative strategies. I am looking at the example you've shown me and nothing there needs code injection, it could already be achieved via the exposed interfaces without having to do any parsing of ASTs or dynamic splicing of scripts.
07:15:35 <Discord|> <06m​umra> You've suggested a complete rewrite of the webtiles codebase into a class based syntax to support a specific plugin architecture you have in mind and I'm telling you that is beyond the time I have available but that a huge amount would be possible without having to do that
07:16:40 <Discord|> <12a​sciiphilia> Aren't you talking about AdvancedRCEditor?
07:16:50 <Discord|> <12a​sciiphilia> Are other modules like that too?
07:19:28 <Discord|> <06m​umra> I'm saying advanced RCE editor should be part of trunk
07:20:09 <Discord|> <06m​umra> But now I'm looking at the kinds of code injection that other modules are doing and seeing that in most cases the same result would be possible without code injection
07:20:46 <Discord|> <06m​umra> Each module is a different case really
07:20:58 <Discord|> <12a​sciiphilia> That would be limited to when you can replace the handler with export.
07:21:17 <Discord|> <12a​sciiphilia> That will take a lot of time, I think, whether it's organizing and exporting the necessary handlers through export or changing the class structure, it will take about the same amount of time.
07:21:33 <Discord|> <06m​umra> The relevant handlers are already available for chat
07:22:14 <Discord|> <06m​umra> As far as I can see
07:26:04 <Discord|> <06m​umra> Chat exports the following  comm.register_handlers({   dump: handle_dump,   chat: receive_message,   update_spectators: update_spectators,   toggle_chat: toggle_entire_chat,   super_hide_chat: super_hide_chat, });  export default { spectators, clear, focus, reset_visibility, show_in_chat };   You are already  overriding the update_spectators with your own handler (so there was no need to do this in a code injection, you only needed to
07:26:04 <Discord|> call comm.register_handlers yourself to override it)  Any of the exports like clear() or reset_visibility() can just be overwritten with your own definition
07:26:12 <Discord|> <12a​sciiphilia>  func a { b() } func b { } return {'a': a, 'b': b}  comm.b = func c {}  comm.a() => call b  You know what? Some parts of the webtile’s client or chat code don’t use exported functions — they use functions within their own context instead.
07:27:40 <Discord|> <12a​sciiphilia>  # comm func a { b() } func b { }  << INJECT (DWEM) >> a = func c {}  return {'a': a, 'b': b}
07:28:26 <Discord|> <06m​umra> Can you give a specific example?  I understand the principle but every example I've looked at so far this wasn't a problem
07:28:59 <Discord|> <12a​sciiphilia> inside chat or comm or client module
07:29:55 <Discord|> <12a​sciiphilia> you can find exported function then we can see above called directly
07:29:55 <Discord|> <06m​umra> I mean a specific example of a DWEM module
07:31:32 <Discord|> <06m​umra> There's no getting away from the fact that doing things this way means your modules can break at the slightest internal changes
07:32:14 <Discord|> <12a​sciiphilia> Yes, but that's kind of the fate of modding frameworks.
07:33:16 <Discord|> <12a​sciiphilia> I don't remember the specific code example, but that kind of code required direct code injection. That's what got me thinking about AST-based injection.
07:34:14 <Discord|> <06m​umra> But an official modding framework will in general try to be backwards compatible. We actual have a certain level of "official" modding in terms of clua and great care is taken to make sure any changes there remain backwards compatible.
07:53:12 <Discord|> <12a​sciiphilia> If it were possible to hook into AMD modules, it would be nice if we could just coexist that way. I couldn’t get any work done today because I got too emotional about this issue and spent too much time on it. Thinking that a personal project I’ve poured hundreds of hours into might suddenly end because of something like this is disheartening, and I’m honestly not in a good mood. If I said anything that came across the wrong
07:53:12 <Discord|> way, I apologize. I understand your position, and if that’s the direction you’re determined to take, there’s nothing I can do about it. Still, if you ever end up supporting functionality similar to what I mentioned, I’d appreciate it if you let me know.
07:53:25 <Discord|> <06m​umra> (Btw I am looking at the translation module now, and I can see here that at least for the action panel injector, there is certainly no way to customise the show_tooltip behaviour, since it's called from inside handle_mouse)
08:07:24 <Discord|> <06m​umra> I sincerely apologise for having caused such an issue. And please understand I have no wish to kill a community project, for my part I have just been attempting to improve the webtiles codebase with what I saw as an unequivocal upgrade that would improve things for everyone 🙂 So it's also quite emotional for me that something I put quite a bit of this weekend into and something I thought everyone would want might ultimately go nowhere.
08:08:49 <Discord|> <06m​umra> As I initially discussed with gammafunk, there would be no merging this without coordination with server admins. And when I started this I obviously didn't realise just how tightly CNC was coupled with the implementation specifics
08:14:28 <Discord|> <06m​umra> I do think it's a very unfortunate situation for webtiles to be stuck with ancient js methodologies. And I think there's a way that all the module functionalities required could be supported. It seems to be that the AST injection is only currently required for a very small number of things, where additional hooks or alternative strategies would eliminate the need for it altogether.
09:32:26 -!- wvc3 is now known as wvc
09:56:36 <NotCheibriados> 03unutbu02 {dolorous} 07* 0.34-a0-977-g65985d8095: Add the ability to toggle banes in wizmode. 10(3 hours ago, 3 files, 36+ 0-) 13https://github.com/crawl/crawl/commit/65985d809575
09:56:36 <NotCheibriados> 03dolorous02 07* 0.34-a0-978-g7b2b1f96eb: Add support for partial bane name matches. 10(47 minutes ago, 3 files, 77+ 13-) 13https://github.com/crawl/crawl/commit/7b2b1f96eb21
11:21:05 <Discord|> <09g​ammafunk> I obviously can't comment at all on the webtiles implementation and architecture, but we definitely have to figure out a solution for CAO before we roll this out. I might look into setting up a local gcc on CAO this weekend
11:23:20 <Discord|> <09g​ammafunk> I think there's a complication in that the chroot also needs to have some updates for e.g. glibc
11:23:42 <Discord|> <09g​ammafunk> which is what that chroot update project was about
12:19:28 <Discord|> <09g​ammafunk> oh I see, aidanh's cpp-17 project ported the chroot to a later ubuntu and ported the crawl-dev user's build dir to the same type of chroot
12:19:41 <Discord|> <09g​ammafunk> I was really wondering how that could possibly work, but I guess that makes sense
12:20:45 <Discord|> <09g​ammafunk> hrm, it'd even run webtiles from this dev chroot
12:20:54 <Discord|> <09g​ammafunk> I wonder what kinds of issues this might create
12:23:56 <Discord|> <09g​ammafunk> it's a clever solution to the problem of the host being out of date, though
12:25:03 <Discord|> <09g​ammafunk> from what I can see, there's awkwardness where you have to mount proc/pts filesystems every time you want to install packages into the dev chroot
12:25:23 <Discord|> <09g​ammafunk> I wonder if there's a problem with just keeping those mounted
12:25:29 <Discord|> <09g​ammafunk> much like how we do with the chroot itself
12:26:01 <Discord|> <09g​ammafunk> what I'd like to do is see if there's a more current "cloud image" we could use and then update aidanh's technique here
12:26:31 <Discord|> <09g​ammafunk> I'm not sure if it matters, but it'd be nice if the cloud image was itself a debian release rather than an ubuntu one, but maybe debian doesn't make those
12:38:26 <Henzell> horredbonez (L6 DsWn) Crash caused by signal #6: Aborted (D:4)
12:39:14 <Discord|> <09g​ammafunk> seem like there is in fact a debian cloud image we can use based on as recent a debian as we want, and the lts timeframes are ultimately the same as ubuntu's, for whatever that's worth: https://cloud.debian.org/images/cloud/trixie/latest/ Presumably the generic image is what we'd want, since I'm sure these provider-specific ones have modifications we don't need. That said, there's a current ubuntu minimal image for their latest LTS,
12:39:15 <Discord|> which would presumably require less updating in terms of aidanh's scripts/approach: https://cloud-images.ubuntu.com/minimal/releases/noble/release/
12:39:53 <Discord|> <09g​ammafunk> aidanh's scripts are here https://github.com/crawl/crawl/tree/master/crawl-ref/source/util/server with the wiki doc he wrote here: https://github.com/crawl/crawl/wiki/CPP-17-Upgrade
12:40:37 <Discord|> <09g​ammafunk> advil mentioned that this turned into a bit of a rube goldberg machine, but implied that the setup wasn't hard to get going, so I feel like it's something reasonable to attempt as a near-term solution
12:41:44 <Discord|> <09g​ammafunk> it would get us both nodejs (for any webtiles changes we ultimately do want to make) along with a modern dev toolchain. The lack of the latter has been a problem on CAO for the crawl binary itself often enough
12:44:08 <Discord|> <09g​ammafunk> If we pull that off, we can start requiring a cpp standard that's relatively recent, probably cpp-20? cao's gcc technically doesn't quite fully support even cpp-11
12:44:24 <Discord|> <09g​ammafunk> but in practice we're able to mostly target that
12:45:06 <Discord|> <09g​ammafunk> oh, hrm, I realize one very big issue that I didn't quite think of
12:45:33 <Discord|> <09g​ammafunk> which is disk...
12:46:01 <Discord|> <09g​ammafunk> this would be an all-or-nothing move of the data from old chroot to new chroot
12:46:48 <Discord|> <09g​ammafunk> since there's not enough room to hold two copies
13:05:11 <Discord|> <06m​umra> Can you copy somewhere offsite as a failsafe backup?
13:08:09 <Discord|> <09g​ammafunk> heh...I suppose that's technically possible, but it's like 1TB of data, not to mention it'd probably be pretty slow
13:08:30 <Discord|> <09g​ammafunk> I imagine a mv even within the same filesystem (which this would be) might be quite slow
13:10:01 <Discord|> <09g​ammafunk> on the one hand, in theory with the one-shot move approach, the worst that happens is that there are problems with the new chroot and more downtime is added to CAO, possibly culimanting in a move back to the old chroot
13:11:52 <Discord|> <09g​ammafunk> on the other, possibly there are issues that we'd not really foresee, since I don't think aidanh tested this from e.g. a wheezy vm or anything
13:13:10 <Discord|> <05i​coson> I was able to get that chroot thing going (at least to the first stage) but I ended up deciding it was a really non-ideal workaround to simply updating the machine
13:13:34 <Discord|> <09g​ammafunk> right, it's just that, from what I understand, rax isn't prepared to undertake said upgrade?
13:13:53 <Discord|> <09g​ammafunk> and this would at least be a near-term solution to woefully out of date dev toolchain
13:14:15 <Discord|> <05i​coson> well they've indicated that they are game at various points but haven't done it, I think it interacts in some complicated way with the virtualization setup?
13:14:58 <Discord|> <05i​coson> I htink I got a pretty modern gcc working on cao actually
13:15:49 <Discord|> <05i​coson> oh I'm thinking of python
13:16:01 <Discord|> <09g​ammafunk> yeah, your local python3 is working fine
13:16:16 <Discord|> <09g​ammafunk> this is an aside, but one problem I did just realize with the new chroot approach is that we don't really have enough disk for this on the current filesystem
13:16:30 <Discord|> <09g​ammafunk> that minimal chroot probably takes more disk than what we have free
13:17:19 <Discord|> <09g​ammafunk> we could delete files to get us down to low 90s in terms of free disk, but I think the new chroot would start its life in home, a separate filesytem
13:17:39 <Discord|> <05i​coson> I still have the chroot I build in my home directory, it's about 848M (by now the actual versions are completely outdated I'm sure)
13:18:13 <Discord|> <05i​coson> iirc there was also some weird mount stuff I had to do, wasn't sure if there were any implications to that
13:19:14 <Discord|> <05i​coson> btw not sure if this matters any more but cdo also had some quirky hand-built gcc version
13:20:07 <Discord|> <09g​ammafunk> yeah, it does, although I'm not sure if cdo's gcc/toolchain matters for this nodejs thing specifically
13:20:16 <Discord|> <09g​ammafunk> it's gcc 6.3.0
13:20:36 <Discord|> <09g​ammafunk> but I don't think any of its builds would be webtiles builds that would run node ever, at least
13:21:41 <Discord|> <05i​coson> maybe that did get updated
13:22:11 <Discord|> <09g​ammafunk> oh interesting, node 16.9 is actually in the path on CDO
13:22:39 <Discord|> <05i​coson> well cdo also doesn't run webtiles at all
13:22:44 <Discord|> <09g​ammafunk> right
13:23:57 <Discord|> <06m​umra> @asciiphilia Some good news, I've had some success getting Vite to output requirejs wrapped modules as separate files instead of bundled (and with minification disabled, it should let things work as they are for you). We can perhaps offer either build mode for servers, if the admin would rather prioritise download times. It's just a different build configuration, the js source is unchanged. I'll push this version in a bit if you want to
13:23:58 <Discord|> test if the injection still works with it (unable to test this myself since i can't run the container still)
13:25:28 <Discord|> <09g​ammafunk> looking at the cao situation again, I suppose we'd just make the new chroot be a subdir of /chroot so it could be totally on the correct filesystem (again once we made sure there was enough free disk for the new chroot). Otherwise I think you run into issues where you try to move the old chroot game data to the wrong filesystem
13:29:56 <Discord|> <09g​ammafunk> But perhaps we should start by asking rax again about an upgrade, probably along with asking for permission and describing what we'd like to attempt (in terms of this chroot upgrade) should she not be able to try to upgrade things.
13:30:18 <Discord|> <09g​ammafunk> it might be wise to wait on all of this until post tournament
13:52:25 <Discord|> <12g​e0ff> the last tournament was 5 and a half months ago, so there's still 6-7 weeks left (if the next T is in December)
13:55:12 <Discord|> <04d​racoomega> Surely it hasn't been 5 and a half months already, has it? T.T
13:56:01 <Discord|> <12g​e0ff> "Reforge Yourself" was released on 2nd May
14:00:26 <Discord|> <04d​racoomega> This causes me pain
14:01:35 <Discord|> <04d​racoomega> In any event, I am still working on weapon arts stuff for this version. (Well, I'm currently refactoring a whole lot of movement code as part of a rabbit hole that sent me down, but I will be returning to it as soon as I'm done this.)
14:02:18 <Discord|> <04d​racoomega> And that whole package will come with enough stuff that wants a decent period of balance observation before considering freezing
14:02:33 <Discord|> <04d​racoomega> (It's taken me way more time overall than expected, but forward progress continues to happen)
14:11:49 <Discord|> <09g​ammafunk> yeah, that's fine, if it takes us 8-9 months for this release, that's just how long it takes
14:12:08 <Discord|> <09g​ammafunk> (or just 7 months, if that's actually feasible)
14:15:26 <Discord|> <04d​racoomega> Well, I'm not planning any other large projects after that, for this version. Maybe a few small things that can be fit in during testing and observation, but nothing that should eat up much additional time on their own.
15:02:28 <NotCheibriados> 03mumra02 07https://github.com/crawl/crawl/pull/4839 * 0.34-a0-993-g3652d0103e: Require-compatible Vite build 10(68 seconds ago, 10 files, 633+ 83-) 13https://github.com/crawl/crawl/commit/3652d0103e9f
15:12:08 <Discord|> <06m​umra> @asciiphilia  ^This version should be ready to be tested against DWEM. Since it seems like in any case this will probably not be buildable on CAO until at least post-tournament, it seems there is plenty of time to make sure everything works before then. Still, I would strongly recommend looking at alternatives to AST injection over time; even with requirejs support remaining, it's very hard to guarantee that some future work on webtiles
15:12:09 <Discord|> won't break DWEM modules at random. Exposing specific APIs for specific extension scenarios just seems like a much more stable way to go. I think allowing server admins more hooks to customise the UI is a great idea if such interfaces are well defined.
15:14:54 <Discord|> <06m​umra> (It's unfortunate this branch will inevitably need some horrific git resolution by that time but that's what I get for mass reformatting and syntax upgrade 😂 )
15:40:31 <Lantell> Unstable branch on underhound.eu updated to: 0.34-a0-978-g7b2b1f96eb (34)
17:18:53 <Henzell> horredbonez (L19 DsWn) Crash caused by signal #6: Aborted (Abyss:2)
18:38:52 <Discord|> <12a​sciiphilia> @mumra  I tried the fast deploy (develop) command and the web tiles are installed properly. (WSL, Docker 26.1.1, Ubuntu 22.04.3 LTS)
18:38:52 <Discord|> <12a​sciiphilia> https://cdn.discordapp.com/attachments/747522859361894521/1430007396767629322/image.png?ex=68f835ab&is=68f6e42b&hm=3960bbbaaece9015b50f024b5eb50d3bb0e310be1ca85596d2c38806ed8e3b98&
18:39:54 <Discord|> <12a​sciiphilia> https://cdn.discordapp.com/attachments/747522859361894521/1430007658756182137/image.png?ex=68f835ea&is=68f6e46a&hm=e8be95c09e0ec0d0ded80a3adf4647fa4788a4e9808774d64c60f2516d481967&
19:17:27 <NotCheibriados> 03WizardIke02 07* 0.34-a0-979-g7c143c1dcc: Add an invert method to FixedBitArray 10(6 minutes ago, 2 files, 6+ 7-) 13https://github.com/crawl/crawl/commit/7c143c1dcc4b
20:03:12 <NotCheibriados> 03WizardIke02 07* 0.34-a0-980-gd7b9d59e62: refactor cloud_tile_variation 10(14 minutes ago, 3 files, 11+ 17-) 13https://github.com/crawl/crawl/commit/d7b9d59e628b
20:52:08 <NotCheibriados> 03WizardIke02 07* 0.34-a0-981-g6892638188: Change parameter type of put_colour_ch to char32_t 10(5 minutes ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/6892638188bf
21:00:40 <NotCheibriados> 03WizardIke02 07* 0.34-a0-982-ga50f8efd80: Remove unused code from discover_mimic 10(2 minutes ago, 1 file, 0+ 5-) 13https://github.com/crawl/crawl/commit/a50f8efd802f
22:20:51 <Discord|> <02D​arby> Having no knowledge of Lua, I don't yet know how to best fix this or what implications it may have regarding vaults elsewhere, but I can at least share the due_megacathedral situation: Normally, it spawns 10 bats over time across 6-11 altars or fountains, tracked by bat_count. If you leave and re-enter the floor, every altar and fountain gets its own bat_count, spawning up to 60-110 bats, assuming you did so while bat_count was 0. This
22:20:52 <Discord|> has always been true, since the vault was added 14 years ago.
22:35:28 <Gretell> Unstable branch on crawl.develz.org updated to: 0.34-a0-982-ga50f8efd80 (34)
22:58:38 <Gretell> Windows builds of master branch on crawl.develz.org updated to: 0.34-a0-982-ga50f8efd80
23:30:05 <Discord|> <02D​arby> (also grunt_lady_of_the_lake spawns 5 merfolk avatars in the same circumstance)
23:35:22 <Cbrotell> Unstable branch on cbro.berotato.org updated to: 0.34-a0-982-ga50f8efd80 (34)
23:46:34 <Discord|> <06m​umra> Oh, this happened to me recently. I thought there were an awful lot of bats...
23:47:33 <Discord|> <06m​umra> Hmm, fast deploy is the one method i didn't try for some reason ...
23:47:34 <Discord|> <02D​arby> yeah, I'm currently just thinking over how to approach (lady of the lake is straightforward and I'll push the 5-lady lake fix shortly)
23:50:09 <Discord|> <02D​arby> having only the central megacathedral altar spawn bats feels sad and like I'm leaving the fundamental problem unfixed, but it's probably better than 100-bat cathedral
23:52:22 <Discord|> <02D​arby> (on the other hand it usually seems to spawn all its bats before the player has even arrived, making all the code a bit irrelevant?)
23:55:36 <Gretell> Monster database of master branch on crawl.develz.org updated to: 0.34-a0-982-ga50f8efd80