<< 18-01-2021 >>

---

00:19:07	giaco	I'm trying to use valgrind to spot memory leak. I'm compiling with --gc:arc, --debuginfo, -d:useMalloc, --stacktrace:on, and I'm running with valgrind --xtree-leak=yes (that implies --leak-check=full and --show-leak-kinds=all), but result file in kcachegrind shows location = UnknownFile??? right where my leak is
00:21:06	saem	hmm, does debuginfo imply linedir and more importantly whether linedir will help at all in this case? Maybe give that a go if it's easy to try?
00:23:41	giaco	saem: thanks, trying now
00:26:15	giaco	still UnknownFile/UnknownFn
00:30:58	*	^Q-Master^ quit (Read error: Connection reset by peer)
00:31:04	*	Q-Master joined #nim
00:34:29	*	antranigv joined #nim
00:35:47	saem	The official docs on profile nim are old but supposedly you shouldn't need debug, stacktrace, or linedir.
00:38:00	saem	giaco: are you using any native libraries as part of your project?
00:39:06	giaco	saem: yes, and I'm trying to spot the nearest point in nim bindings that causes the leak
00:40:54	saem	I bet you do need debug symbols for this, and I wonder if perhaps you don't have those for the native libs?
00:48:17	giaco	saem: I should have them, yeah, let me double check first
00:48:35	giaco	(using -dbg packages from apt)
00:50:35	*	Gustavo6046_ joined #nim
00:51:28	giaco	yeah, I do have dbg packages installed
00:52:34	*	Gustavo6046 quit (Ping timeout: 272 seconds)
00:53:11	saem	Not sure quite what else to try but finding working examples and then extrapolating from there, sorry. Hopefully someone who's more familiar with it comes by soon.
00:54:02	giaco	saem: not related to my question, but I had to uninstall your nim plugin for vscode and fallback to the one from Zaitsev as I've figure out that it slowsly causes memory to fill up by opening many nimsuggests processes until vscode is closed
00:55:25	giaco	all I have to reproduce the issue is write code and press F6 to run, many hours later I have 16GB memory filled up my nimsuggests processes
00:55:27	saem	giaco: they both have the same issue, IIRC. In both cases you need to setup a "nim.project" in your workspace or folder settings and specify the nim files
00:56:17	*	Gustavo6046 joined #nim
00:56:18	giaco	saem: the one from Zaitsev doesn't seem to have it, checking with htop
00:56:44	saem	Oh, I wonder if the issue with the run command then.
00:57:01	saem	I basically never use that, so perhaps its related
00:57:03	*	Gustavo6046_ quit (Ping timeout: 260 seconds)
00:59:02	giaco	I've installed vscode just to learn nim, never used before, actually don't know anything else about it :D
01:00:47	*	Tanger joined #nim
01:01:19	saem	If you do get a chance to replicate it and don't mind grabbing a listing of the nimsuggest processes with their command line arguments, that would quickly give me an indication of what's up.
01:02:15	saem	It's possible that it loses track of the processes somehow, also the indexing operation is _not_ cheap.
01:03:46	giaco	saem: I will do, if that helps
01:03:53	giaco	just reinstalled it
01:04:52	saem	Mostly it tells me if it's the same thing file over and over again and maybe an idea of how far apart, a clue as to whether it's something high or low frequency that's creating them.
01:15:36	saem	hmm, I bet I know where the bug is already. :D
01:25:40	giaco	so far it seems different files, but just 20 minutes run. Damn, each nimsuggests is 692MB resident memory :D
01:26:36	saem	yeah, that's where the project setting makes a big difference, both extensions will assume each nim file is a project
01:26:57	saem	And they do an at startup scan of all nim files to build a type and file index.
01:27:47	giaco	it makes sense, actually
01:28:04	giaco	I do have many independent nim "playground" files, to experiment and learn
01:33:31	saem	Those can each be added to the project array or you can use project mapping., which uses pattern matching.
01:33:48	saem	I haven't used that at all and it's rather convoluted to be honest.
01:42:48	*	krux02 quit (Remote host closed the connection)
01:44:51	saem	giaco: what's your github handle if you don't mind me asking?
02:09:04	giaco	saem: https://github.com/arkanoid87
02:10:50	*	abm quit (Quit: Leaving)
02:19:25	*	oculux quit (Ping timeout: 264 seconds)
02:19:25	*	actuallybatman quit (Ping timeout: 264 seconds)
02:19:42	*	actuallybatman joined #nim
02:20:18	*	oculux joined #nim
02:21:20	*	rockcavera joined #nim
02:41:39	leorize	saem: nope, I haven't looked closely at sem interactions with the VM
02:42:21	leorize	disruptek: have you figured out the cps issue?
02:45:44	saem	giaco: new release will be out in a few minutes, let me know how it goes.
02:46:06	saem	I'll keep an eye on it on my end, as well.
02:47:34	giaco	saem: now I have 3 processes open on different nim files. Basically just by doing "Go To Definition" it spawns a new one and 700MB are gone
02:47:40	saem	leorize: np, I think the issue might be how sem is done while in the macro context and less about the eval there after.
02:49:03	saem	giaco: if they're different files and as long as those match your nim.project setting or that setting is empty, then that's as expected. If they're for the same file, then that's a leak.
02:52:21	giaco	They are on different files, so I think that's the expected behaviour. On the other hand it is sufficient to open a dozen of nim files while tracking a bug to fill whole system memory
02:52:38	leorize	saem: I think you should standardize your nim.project thingy into the compiler config or nimble config so other tools/editors can use it too
02:52:54	leorize	(and maybe that will also pave the ground for one nimsuggest doing everything)
02:53:43	giaco	btw I'm not using nim.project, I just clicked on install and started pressing F6 and right click "go to declaration"
02:53:47	saem	leorize: cfg isn't the right place, there needs to be a higher level concept. But I do agree that's where it should live.
02:55:01	leorize	once I look into adding "package" as a concept to the compiler we can discuss that in details I suppose
02:55:28	leorize	sometimes I hate that it seems easier to write an implementation then pitch for it than writing an RFC
02:56:18	saem	It's both backwards and not.
03:00:43	saem	Right now I'm _trying_ to focus on sem and suggest for macros in particular.
03:06:13	*	lritter quit (Ping timeout: 264 seconds)
03:06:57	*	lritter joined #nim
03:10:39	giaco	if I close the tabs with open files, processes are still alive even for closed files
03:16:14	*	muffindrake quit (Ping timeout: 264 seconds)
03:17:36	saem	There is code to close nimsuggest files when you don't have nim.project defined, but it's not good at handling errors, so I'm guessing it's flaking out. Good news is that every invocation of nimsuggest will result in it doing that check to clean things up, so eventually it succeeds.
03:17:45	*	muffindrake joined #nim
03:17:52	saem	s/nimsuggest files/nimsuggest processes
03:18:23	saem	"good news"
03:18:38	leorize	I actually skip that in nim.nvim and no one seems to care :P might be because people don't run nvim for a really long time
03:18:53	giaco	yeah I do confirm that they close after some minutes
03:19:09	giaco	so what's important is to keep small number of open tabs
03:20:21	saem	or use project, which is what I do for compiler personally and I have goto def and suggest working for that code base.
03:21:27	leorize	does your project thingy have a defined semantics? I can hook that into nim.nvim for interop if it does
03:22:10	*	Tlangir joined #nim
03:22:16	*	Tanger quit (Read error: Connection reset by peer)
03:22:42	saem	leorize: I'm using the vscode config stuff.
03:23:04	leorize	is it json or json with comments?
03:23:20	leorize	if its the former then I can read just fine
03:23:21	saem	That's an example: https://github.com/saem/vscode-nim/blob/main/.vscode/settings.json
03:23:49	saem	I believe it is json with comments, I think there is a schema for it too, but whatever.
03:24:10	leorize	I'm definitely not gonna implement a parser in vimscript
03:24:18	saem	LoL, yeah no
03:32:21	leorize	I did have to write one, regrettably
03:32:28	leorize	for parsing Nim in reverse :P
03:32:44	leorize	allows this stuff so it's pretty cool: https://asciinema.org/a/RXbwCTJXGRKEhMRiHIUEoDGd7
03:32:44	ForumUpdaterBot	New thread by 0x100: Help with Nim concurrency – data not persisted, see https://forum.nim-lang.org/t/7397
03:36:10	*	zedeus joined #nim
04:06:29	leorize	disruptek: so I pushed some walkarounds to test my eventqueue impl, but it seems to me that result lifting for cps has some issues: https://github.com/alaviss/nim-sys/runs/1719141572?check_suite_focus=true#step:7:13
04:12:30	leorize	btw, how can I get the return value from a non-cps proc?
04:12:39	leorize	or do I need a cps entry point?
04:17:52	*	spiderstew_ joined #nim
04:18:29	*	spiderstew quit (Ping timeout: 246 seconds)
04:32:14	*	vicfred quit (Quit: Leaving)
04:32:46	*	rockcavera quit (Remote host closed the connection)
05:02:24	disruptek	we haven't figured out the syntax for passing results around, but it's blocked anyway.
05:03:27	saem	wump wump
05:04:31	disruptek	i guess i still don't really have internet.
05:04:34	*	disruptek sighs.
05:06:04	saem	:(
05:07:13	leorize	disruptek: that means I will have to settle with a var param?
05:13:25	disruptek	no, but you need to track the results outside of .cps. -- there's no shim for `let x = someCpsCall()` for inside a cps proc.
05:14:55	disruptek	we have a discussion for this but disbot doesn't know how to find discussions yet.
05:15:01	disruptek	disbot: will you ever learn?
05:15:02	disbot	you may rely on it.
05:16:02	leorize	out of curiousity I added a var param, and it still doesn't work: http://ix.io/2MpS
05:16:06	disruptek	leorize: look at zevv's iterator for example, in stash
05:17:05	leorize	so the result type has to be in the continuation object
05:17:11	leorize	that's annoying
05:17:34	disruptek	var params are a separate problem with a separate discussion and zevv doesn't think we'll support them. i think we will, but maybe only for stack types, which would suck.
05:17:58	leorize	would views enable support for them?
05:18:27	disruptek	results? results are just a syntax and type eraser problem.
05:18:38	leorize	var params I mean
05:18:59	disruptek	i don't see how.
05:19:14	*	letto quit (Quit: Konversation terminated!)
05:19:59	leorize	since `var T` will then be a type that can be placed in the Env object?
05:20:01	disruptek	i think clyybber has to weigh in on this, too, because right now i believe we are stripping the ast we'd need for var params anyway.
05:20:36	disruptek	well, it's just [T] that goes in the env, because otherwise you have no idea what the return type of the result is.
05:20:48	disruptek	at least, this is how i implemented it by hand.
05:20:59	*	letto joined #nim
05:21:20	disruptek	look at my generic continuations concept in experiments/chain.nim to get a taste.
05:21:46	leorize	so you still has some of asyncdispatch problems in regards to signatures
05:22:15	disruptek	you mean, like `void`?
05:22:47	leorize	I want var params to work, that is
05:23:00	leorize	if they don't work then I only have pointers or ref T
05:24:16	disruptek	i mean, it's not a small ask.
05:24:28	disruptek	think about what it means.
05:25:16	disruptek	where exactly do we store your var params and how?
05:25:53	leorize	you can store it in Env and views get you free analysis
05:26:05	*	idf quit (Ping timeout: 240 seconds)
05:27:01	leorize	I can make do with pointers, they are just annoying when you want the experience between async and sync to be the same
05:29:11	disruptek	are you saying that even if you had results, you'd still need to use your own ptrs?
05:29:47	leorize	yes, see my `read()` proc, I read into a pre-allocated buffer and return how much I've read
05:30:17	disruptek	yeah, that's going to be pretty hard to do without views at least.
05:30:56	ForumUpdaterBot	New thread by Drkameleon: Portable binaries with Nim - what works?, see https://forum.nim-lang.org/t/7398
05:35:08	leorize	I can probably walkaround by creating a `proc read(c: Cont, f: AsyncFile, b: var openArray[byte], read: var int): Cont` that calls the `proc read(f: AsyncFile, buf: pointer, size: int, read: ptr int) {.cps: Cont.}` behind the scene
05:35:20	leorize	there will be no escape analysis so safety is literally zero
05:35:41	disruptek	well yeah, because it escapes.
05:36:06	disruptek	i mean, if you pass a mutable buffer to me, i won't have a view on it by the end of the proc.
05:36:22	leorize	but given that it has to be called from cps, assuming the user have a buffer on the "stack", it will be safe 60% of the time :P
05:36:33	disruptek	so there's no way for cps to make control flow against it, since all control flow has to goto.
05:39:07	disruptek	maybe we're overthinking this.
05:40:34	disruptek	you invoke read() with a var param and we simply ignore it; don't add it to the env.
05:41:14	disruptek	then when the file is ready and we reach the ultimate cpsMagic call, we issue it with the var param.
05:41:19	disruptek	why doesn't that work?
05:44:45	leorize	mainly because I can't imagine it xd
05:45:02	*	waleee-cl quit (Quit: Connection closed for inactivity)
05:45:06	leorize	can you describe it in a bit more details?
05:46:26	disruptek	well, your magic is a closure over the buffer.
05:48:25	leorize	but my magic is a wait() and it doesn't take a buffer?
05:49:03	disruptek	no, i mean the read itself.
05:53:20	leorize	still can't imagine it, sorry
05:56:19	saem	Why can't the env close over the var param?
05:57:13	disruptek	i think the magic invocations can but the continuations themselves are nimcalls.
05:57:39	saem	nimcalls you generate by morphing the existing AST?
05:57:48	disruptek	yeah.
05:59:00	saem	Let me noodle, it feels like it shouldnt' be a problem.
05:59:25	leorize	note that Nim's closure can't capture var params because of poor escape analysis
06:01:57	*	vicfred joined #nim
06:03:45	*	idf joined #nim
06:04:29	disruptek	imagine this code: cpsCall immutableInput: body for magic which is a closure we run to yield the result...
06:04:38	FromDiscord	<JSONBash> any tips on how to convert an object with a datetime to a string? the date field can be a string. example code: https://play.nim-lang.org/#ix=2MpZ
06:15:24	*	blueberrypie quit (Quit: leaving)
06:16:10	*	blueberrypie joined #nim
06:22:58	saem	The lack of escape analysis really butts it up. :D
06:32:55	saem	Ugh, i got a terrible idea and disruptek is going to hate it. But could var T captured by the continuation be given a colour wherein it becomes CpsWrappedVar[var T] and you use that to break type compatibility with it getting passed across an FFI boundary or something else nasty?
07:12:13	*	habamax joined #nim
07:34:38	FromDiscord	<pietroppeter> JsonBash: https://play.nim-lang.org/#ix=2Mqc
07:35:26	FromDiscord	<pietroppeter> (Object have a $ already defined, if you want a JSON you need to define % for DateTime)
07:42:46	FromGitter	<AllenDang> @pmutua Do you know how to convert a wxString to string? I've tried all the way I know but still doesn't work. I'm using ` var name = nameCtrl.getValue().cStr().asCString()` but it reports error `error: cannot initialize a variable of type 'NCSTRING' (aka 'char *') with an rvalue of type 'const char *'`
07:44:47	*	habamax_ joined #nim
07:47:32	*	habamax quit (Ping timeout: 256 seconds)
07:50:18	*	krux02 joined #nim
07:50:24	*	narimiran joined #nim
07:53:30	*	hoijui joined #nim
08:03:11	*	hoijui quit (Ping timeout: 246 seconds)
08:12:08	*	habamax_ quit (Quit: leaving)
08:12:21	*	PMunch joined #nim
08:12:28	*	habamax joined #nim
08:51:42	Araq	saem, we have "escape analysis" or rather "borrow checking" via the - er - unfinished view types feature
08:51:54	saem	Araq: debugging twithin_macro & twithin_macro_prefix (after the 1st call). twithin_macro_prefix works once, by way of processmodule:180 to processTopLevelStmts:73 & does the sem pass + suggestExprNoCheck. After it goes processmodule:189 & closes the pass, like twithin_macro. It's trying to sug on the unforgiving AST (not sugExprNoCheck)? A fix: note AST ranges as from macros and redo those, is that possible, alternatives?
08:51:59	saem	Haha, was writing that all along
08:52:08	PMunch	Hmm, this triggers a C compiler error: http://ix.io/2Mqq
08:52:29	PMunch	http://ix.io/2Mqr
08:53:22	saem	Probably could also hack some approximations with tags for a few narrow cases.
08:55:19	PMunch	It works if it is a ref object..
08:56:06	saem	Just to explain the twithin_macro issues, seems like after the first eval of the macros in the prefix variety it falls back to the non-prefix version. Where the prefix version works is that the first time it does in fact do the sem pass for the partially written proc def inside the macro at 25:10. But the non-prefix one never goes to sem during sug, only happens during the initial compile for nimsuggest startup.
09:00:16	*	leorize[m]1 quit (Quit: Idle for 30+ days)
09:00:21	*	watzon quit (Quit: Idle for 30+ days)
09:00:23	*	xicheng[m] quit (Quit: Idle for 30+ days)
09:00:24	*	wontruefree[m] quit (Quit: Idle for 30+ days)
09:01:30	saem	Well, that's not entirely true, it does go into semStmt -> semExprNoType, but it doesn't seem to hit the sugExpr path.
09:02:18	Araq	saem, well you're now in the postion where you know more than me
09:02:26	saem	D'oh!
09:02:29	saem	:D
09:02:36	Araq	esp since my brain is doing something else
09:04:42	saem	Fair point.
09:06:11	PMunch	Araq, something like this doesn't work with the default GC because Foo is copied (and it shares strings and such across threads which isn't a great idea) http://ix.io/2Mqv
09:06:33	PMunch	But with --gc:arc it works fine, but is it safe to do things like that with ARC=
09:06:58	*	Vladar joined #nim
09:13:46	Araq	PMunch, looks ok to me
09:14:11	PMunch	So sharing strings like that in a globally shared table is fine under ARC?
09:14:56	Araq	pretty much, yeah, that's why we have ARC
09:15:07	PMunch	Holy shit, that's amazing!
09:16:12	PMunch	I've just been stoked for ARC because it's faster and I knew it would help multi-threading. But I didn't know it was that good!
09:16:46	PMunch	Will this work with ORC as well?
09:16:51	saem	OK, I think this is going to be a rest of the week problem. Apologies, if a rather significant PR shows up because of this, the solutions I'm thinking of right now aren't small -- hopefully I see something better in the next few days.
09:17:40	*	Tlangir quit (Quit: Leaving)
09:28:03	FromGitter	<AllenDang> Anyone know how to convert a wxString to string? I've tried all the way I know but still doesn't work. I'm using var name = nameCtrl.getValue().cStr().asCString() but it reports error error: cannot initialize a variable of type 'NCSTRING' (aka 'char *') with an rvalue of type 'const char *'
09:28:47	FromGitter	<AllenDang> I tried `var name = $nameCtrl.getValue().cStr().asCString()`, same error.
09:29:01	FromGitter	<AllenDang> And `let name = nameCtrl.getValue().cStr().asCString()`, same error.
09:29:42	FromGitter	<AllenDang> I'm trying to use wxnim to create GUI desktop app and stuck at string conversion now.
09:31:02	FromDiscord	<ElegantBeef> Pmunch might know
09:31:36	PMunch	Haha, I was actually just looking at that forum thread and I'm trying to remember how I did this
09:32:00	FromGitter	<AllenDang> @Pmunch thank god you sees this...
09:32:17	PMunch	There is an example that tried to echo out some text from a text-box, but it doesn't work any longer. I remember that I fixed it, but I must've forgotten to push that fix..
09:32:51	FromGitter	<AllenDang> wxnim is the most beautiful wrapper for wxWidget, I'm eager to use it.
09:40:59	Araq	PMunch, for strings, yes, for cyclic data, no
09:41:23	PMunch	Aha, so don't share cyclic data across threads
09:41:51	Araq	don't say "share", say "send"
09:42:20	FromDiscord	<ElegantBeef> Also araq i showed my friend `scanTuple` and his first reaction was "why isnt i an option" 😄
09:42:22	PMunch	Ah right
09:42:25	FromDiscord	<ElegantBeef> (edit) "i" => "it"
09:42:28	Araq	our non-atomic refcounts don't support sharing, only sending data around... :-)
09:42:41	FromDiscord	<ElegantBeef> So to option or not to 😛
09:42:52	PMunch	I see, and sending a cyclic data structure causes weirdness?
09:43:34	Araq	no, that works too once some API was added
09:44:29	Araq	we need to ensure the cycle collector doesn't have registered roots, call GC_runOrc before sending
09:44:59	*	hnOsmium0001 quit (Quit: Connection closed for inactivity)
09:45:01	Araq	well, the API exists
09:45:14	PMunch	Ah, so it's not a fundamental issue
09:45:31	Araq	but in an ideal world the channel's send operation calls GC_runOrc for you
09:53:14	*	lritter quit (Quit: Leaving)
09:57:15	PMunch	Hmm, @AllenDang, I have found something that works now
09:57:41	FromGitter	<AllenDang> @PMunch Can't wait to know it!
09:57:47	PMunch	I'm using a TextCtrl for my testing
09:58:53	PMunch	And `echo myTextCtrl.getLineText(0).cStr.asCstring` doesn't work, but `var myStr = myTextCtrl.getLineText(0); echo myStr.cStr.asCstring` works just fine
10:03:17	FromGitter	<AllenDang> var name = nameCtrl.getValue()
10:03:17	FromGitter	<AllenDang> echo name.cStr().asCString()
10:03:51	FromGitter	<AllenDang> I got same error 'error: cannot initialize a variable of type 'NCSTRING' (aka 'char *') with an rvalue of type 'const char *'
10:04:11	*	jkiesian joined #nim
10:04:17	Araq	you need to tell your C++ compiler to be less picky
10:04:25	PMunch	Hmm, not sure why your rvalue things are `const char *`..
10:04:43	PMunch	@AllenDang, do you have a minimal example? And what C++ compiler do you use?
10:05:38	FromGitter	<AllenDang> Error: execution of an external compiler program 'clang++ -c -std=gnu++14 -funsigned-char `wx-config --cppflags` -I/usr/local/Cellar/nim/1.4.2/nim/lib -I/Users/allendang/Desktop/nimtest -o /Users/allendang/.cache/nim/test_d/@mtest.nim.cpp.o /Users/allendang/.
10:06:47	FromGitter	<AllenDang> Compile with `nim cpp -r test.nim`
10:07:10	PMunch	Do you have test.nim?
10:07:11	FromGitter	<AllenDang> I'm using MacOS v11.1
10:07:46	FromGitter	<AllenDang> http://ix.io/2MqJ
10:08:12	*	Tanger joined #nim
10:08:25	FromGitter	<AllenDang> Pleaes remove the `` in beginning and ending of the test code
10:08:39	PMunch	@AllenDang, that compiles and runs just fine for me..
10:08:53	PMunch	So it is a C++ compiler strictness thing as Araq said
10:09:06	FromGitter	<AllenDang> Enter some text in name, and click button
10:09:15	PMunch	By the way you don't have to put the imports in string literals
10:09:26	PMunch	It outputs that name to the terminal
10:09:31	FromGitter	<AllenDang> I got () in terminal as echo's result...
10:10:04	Araq	you can usually tweak these things with more casting inside .importcpp pattersn though
10:10:05	PMunch	That's the same issue I got before I assigned it to a variable first
10:10:33	FromGitter	<AllenDang> http://ix.io/2MqK
10:11:04	FromGitter	<AllenDang> Should I create a nim.config and write something there?
10:12:33	Oddmonger	do you know a gui lib which could be integrated in my own SDL rendering ? I mean, i have already defined window and renderer, and don't want to use a context defined by the UI lib (so, no gtk UI, for example)
10:15:20	Oddmonger	i tried and installed imgui, but even the sample provided on README.md wasn't working
10:21:49	PMunch	@AllenDang, not quite sure how to fix this TBH
10:22:22	PMunch	The issue is that GCC is a bit more relaxed than clang when it comes to const char * -> char * conversions
10:24:13	FromGitter	<AllenDang> @PMunch I should try to use GCC instead of clang as compiler.
10:24:32	PMunch	That would definitely work as it is what I'm doing
10:24:46	PMunch	But it's unfortunate that this doesn't work with clang..
10:25:49	Araq	Oddmonger, iirc 'nimx' uses SDL as the backend too so it would be a natural fit
10:27:54	FromGitter	<AllenDang> http://ix.io/2MqP
10:28:18	FromGitter	<AllenDang> I created a nim.cfg and set `cc = gcc`, seems it compiles with g++, but error remains.
10:31:12	Oddmonger	thank you Araq , it seems to fit to what i want
10:32:47	PMunch	@AllenDang, could you run `which gcc`?
10:32:58	PMunch	It might just be symlinked to clang on your system
10:33:08	PMunch	Or rather `which g++`
10:33:16	FromGitter	<AllenDang> http://ix.io/2MqU
10:33:57	FromGitter	<AllenDang> You got it. It is symlinked to clang.
10:40:03	*	abm joined #nim
10:40:16	FromDiscord	<apollo> oh wow the creator was here!
10:40:32	*	gangstacat quit (Quit: Ĝis!)
10:40:38	FromDiscord	<Rika> theyre often here
10:40:50	FromDiscord	<apollo> thats kl
10:41:59	FromGitter	<AllenDang> http://ix.io/2MqW
10:42:15	FromGitter	<AllenDang> I can confirm that gcc-10 is used, but error remains...
10:43:09	*	vicfred quit (Quit: Leaving)
10:43:12	PMunch	It uses `g++` though, not `gcc`
10:43:17	PMunch	Is that alias also updated?
10:44:36	FromGitter	<AllenDang> I will try now.
10:45:32	Araq	Rika: I'm still a single person, male, and not schizo
10:45:51	Araq	if you think "they" is the least offensive pronoun to pick... no. it's not.
10:46:01	FromGitter	<AllenDang> @PMunch ❯ which g++
10:46:01	FromGitter	<AllenDang> g++: aliased to g++-10
10:46:01	FromGitter	<AllenDang> ❯ g++ version
10:46:01	FromGitter	<AllenDang> g++-10: error: version: No such file or directory
10:46:03	FromGitter	<AllenDang> g++-10: fatal error: no input files
10:46:06	Araq	just fyi, I won't bring it up again
10:46:06	FromGitter	<AllenDang> compilation terminated.
10:46:46	FromGitter	<AllenDang> http://ix.io/2MqZ
10:46:49	PMunch	@AllenDang, `g++ --version`
10:47:00	FromGitter	<AllenDang> g++ is aliased to g++-10
10:47:10	PMunch	Hmm, and it still doesn't work?
10:47:15	PMunch	That is strange
10:47:22	FromGitter	<AllenDang> Error remains
10:47:39	PMunch	I'm also using GCC 10.2
10:48:07	PMunch	Wait, which WxWidgets version are you using
10:48:23	PMunch	`wx-config --version`
10:49:15	FromGitter	<AllenDang> 1) 1.4
10:49:48	PMunch	Huh?
10:53:25	FromGitter	<AllenDang> Am i using a wrong version?
10:53:43	PMunch	Are you using version 1.4 of WxWidgets?
10:54:20	FromGitter	<AllenDang> brew tap andantissimo/wxmac
10:54:25	FromGitter	<AllenDang> brew install [email protected] --with-static
10:54:31	FromGitter	<AllenDang> This is how i install wxwidget
10:54:46	PMunch	Ah okay, so version 3.1 which is fairly new
10:54:55	PMunch	1.4 would be from about 1997 :P
10:54:56	FromGitter	<AllenDang> Im using version 3.1.4 of wxwidgets
10:55:03	PMunch	Right :P
10:55:24	PMunch	I'm using 3.0.5, so it might be that it doesn't work on the 3.1 branch
10:55:49	FromDiscord	<Rika> araq force of habit my dude
10:57:05	FromGitter	<AllenDang> One solution comes in my mind is to create a cpp file which invoke strdup and returns a char * to do the conversion.
10:57:18	FromGitter	<AllenDang> But im not sure this is the right way to do.
10:57:53	PMunch	Oh there's definitely some hacks like that you can use
10:58:08	PMunch	I saw some references to strchr as well when I was looking around for a solution
10:59:19	FromGitter	<AllenDang> Thanks very much for your effort. Waiting for something comes from your side now. I'm new to nim.
11:09:34	PMunch	Did you try wxWidgets 3.0.5? That might be what's causing the bug to begin with..
11:16:34	FromGitter	<AllenDang> Sure, I will try it now.
11:18:36	FromGitter	<AllenDang> http://ix.io/2Mrb
11:18:43	FromGitter	<AllenDang> No luck... Error remains.
11:21:17	PMunch	Hmm, then I'm out of ideas :(
11:32:48	FromDiscord	<dom96> > I'm still a single person, male, and not schizo
11:33:05	FromDiscord	<dom96> What does being “schizo” have to do with anything?
11:33:22	PMunch	Multiple personality disorder
11:34:59	FromDiscord	<ache of head> he's assuring you that he doesn't have any more descriptions like this.
11:35:04	FromDiscord	<dom96> Is schizophrenia a super set of multiple personality disorder?
11:35:57	*	gangstacat joined #nim
11:36:41	PMunch	No idea, but it is commonly associated with it
11:50:32	Zevv	nah, Ive seen schizoprhenia from very nearby in two good friends, and its nothing MPD related (in these cases anyway)
11:51:05	Zevv	it's hearing&seeing things, losing sense with reality and tons of paranoia and distrust of others
11:54:20	*	hoijui joined #nim
11:58:41	FromDiscord	<mratsim> @leorize, can you go over my proposal https://github.com/weavers-guild/weave-io/blob/master/design/design_2_continuations.md↵and my branch and check if having nim-sys CPS aware proc be suspend or coroutines is enough to have CPS support: https://github.com/disruptek/cps/tree/mratsim-public-api-proposal/mratsim↵↵This is a "high-level" API proposal for continuations.
11:59:03	*	hmmm joined #nim
11:59:28	Tanger	Heya folks, anyone know if there's a way to create a standard PEG file from the npeg library?
12:00:49	Tanger	ie with the standard grammar syntax and not the macrofied nim version. Just for documentation
12:11:50	PMunch	@Tanger, I'm not sure, don't think so. You can generate railway diagrams for documentation though.
12:12:18	PMunch	Try asking Zevv. But I think the issue is that NPEG isn't 100% compatible with normal PEG (but I might be wrong)
12:12:24	Tanger	@PMunch, railway diagrams are the ones like "command o────┬─[stdlib]" yeah?
12:12:31	PMunch	Yes
12:13:03	Tanger	Ah, I dig
12:13:16	Tanger	Well I got the railway and a dotfile from it, so that should be fine
12:29:49	Zevv	PMunch: there is no such thing as normal PEG
12:32:38	PMunch	Oh, I thought that was a specification of some kind
12:32:47	Tanger	@Zevv, So the PEG syntax definitions are just a general way of how to implement it?
12:33:59	Zevv	There was an initial implementation by Bryan Ford, but generally it's a "way" of doin things
12:34:22	Zevv	Ford used Packrat parsing, which is pretty classical and relies on memoization the whole input string
12:34:30	Zevv	it's pretty effective CPU wise, but eats a lot of memory
12:34:57	Zevv	NPEG is build on the ideas of Roberto Ierusalimschy (Author of Lua), but better
12:35:17	Zevv	The original Lua LPEG uses a VM, while NPEG actually compiles this VM into Nim code
12:36:24	Tanger	Wow, super cool
12:37:23	hmmm	dudes how do you validate a string date? I wrote an horrible 20 line kludge to do that but it strikes me there must be a simpler way :|
12:38:10	FromDiscord	<lqdev> Zevv: so the VM compiles to efficient machine code, almost as if i wrote a parser by hand?
12:42:26	hmmm	horrible kludge -> https://play.nim-lang.org/#ix=2Mry
12:43:06	ForumUpdaterBot	New question by Alex Craft: Nim template where early binding is not possible?, see https://stackoverflow.com/questions/65774708/nim-template-where-early-binding-is-not-possible
12:43:31	Zevv	lqdev: nope, if you think performance matters, write by hand
12:43:45	Zevv	NPeg used to be faster then stock Nim json parser, but it is no longer
12:43:59	Zevv	but it *is* pretty fast.
12:44:05	Zevv	check -d:npegExpand to see what is generated
12:44:16	FromDiscord	<lqdev> k, was just curious
12:44:27	*	jkiesian quit (Ping timeout: 256 seconds)
12:46:55	*	literal quit (Remote host closed the connection)
12:47:03	*	literal joined #nim
12:49:14	FromDiscord	<exelotl> what's the deal with packedsets, is it a rewrite of intsets? What changed? just better performance?
12:49:37	PMunch	Zevv, did the JSON parser get faster, or did NPEG get slower?
12:50:30	FromDiscord	<exelotl> ah I found the answer https://github.com/nim-lang/Nim/pull/15564
12:50:31	disbot	➥ Make IntSet a generic ordinal set PackedSet[A] + new module packedsets
12:51:16	Zevv	json got faster
12:51:26	FromDiscord	<mratsim> json plural
12:51:31	*	jkiesian joined #nim
12:52:00	FromDiscord	<mratsim> somehow 4 json implementations popped out in the past 2 months
12:52:03	FromDiscord	<mratsim> or was it 5?
12:56:29	*	hoijui quit (*.net *.split)
12:56:30	*	Gustavo6046 quit (*.net *.split)
12:56:30	*	crem quit (*.net *.split)
12:56:30	*	pbb quit (*.net *.split)
12:56:30	*	npgm quit (*.net *.split)
12:56:30	*	Jesin quit (*.net *.split)
12:56:30	*	Zoom[m] quit (*.net *.split)
12:56:31	*	sknebel quit (*.net *.split)
12:56:31	*	mahlon quit (*.net *.split)
12:57:06	*	tsujp1 quit (Ping timeout: 272 seconds)
12:59:01	*	dom96 quit (Ping timeout: 246 seconds)
12:59:11	*	hoijui joined #nim
12:59:11	*	Gustavo6046 joined #nim
12:59:11	*	crem joined #nim
12:59:11	*	pbb joined #nim
12:59:11	*	npgm joined #nim
12:59:11	*	Jesin joined #nim
12:59:11	*	Zoom[m] joined #nim
12:59:11	*	sknebel joined #nim
12:59:11	*	mahlon joined #nim
12:59:45	*	tsujp1 joined #nim
13:01:13	*	cvoxel joined #nim
13:01:19	*	Torro joined #nim
13:01:22	*	dom96 joined #nim
13:01:23	*	dom96 quit (Changing host)
13:01:23	*	dom96 joined #nim
13:01:51	*	rockcavera joined #nim
13:02:41	*	cadmium[m] quit (Ping timeout: 244 seconds)
13:02:41	*	Avatarfighter[m] quit (Ping timeout: 244 seconds)
13:02:53	*	dilawar_uchiha[m quit (Ping timeout: 258 seconds)
13:02:56	*	cmc[m] quit (Ping timeout: 240 seconds)
13:02:58	*	Avahe[m] quit (Ping timeout: 240 seconds)
13:03:06	*	Juno[m] quit (Ping timeout: 240 seconds)
13:03:06	*	Clonkk[m] quit (Ping timeout: 240 seconds)
13:03:06	*	k0mpjut0r quit (Ping timeout: 240 seconds)
13:03:06	*	VijayMarupudi[m] quit (Ping timeout: 240 seconds)
13:03:08	*	matthias[m] quit (Ping timeout: 240 seconds)
13:03:08	*	GitterIntegratio quit (Ping timeout: 240 seconds)
13:03:09	*	Zoom[m] quit (Ping timeout: 246 seconds)
13:03:12	*	stisa[m] quit (Ping timeout: 244 seconds)
13:03:12	*	planetis[m] quit (Ping timeout: 244 seconds)
13:03:12	*	lnxw37d4 quit (Ping timeout: 244 seconds)
13:03:13	*	j-james[m] quit (Ping timeout: 244 seconds)
13:03:13	*	vindaar[m] quit (Ping timeout: 244 seconds)
13:03:14	*	leorize[m] quit (Ping timeout: 244 seconds)
13:03:14	*	BitPuffin quit (Ping timeout: 244 seconds)
13:03:17	*	codic quit (Ping timeout: 258 seconds)
13:03:17	*	xigoi[m] quit (Ping timeout: 258 seconds)
13:03:18	*	lytedev[m] quit (Ping timeout: 258 seconds)
13:03:18	*	unclechu quit (Ping timeout: 258 seconds)
13:03:25	*	Kurre[m] quit (Ping timeout: 240 seconds)
13:03:38	*	silvernode[m] quit (Ping timeout: 264 seconds)
13:03:39	*	BauxnaMatrix[m] quit (Ping timeout: 258 seconds)
13:03:39	*	juanfra__ quit (Ping timeout: 258 seconds)
13:03:43	*	jaens[m] quit (Ping timeout: 244 seconds)
13:03:45	*	himup[m] quit (Ping timeout: 268 seconds)
13:03:45	*	MTRNord quit (Ping timeout: 268 seconds)
13:05:29	*	hmmm quit (Quit: WeeChat 2.8)
13:07:06	*	reversem3 quit (Ping timeout: 258 seconds)
13:17:04	*	cvoxel quit (Quit: back to work)
13:19:18	*	BitPuffin joined #nim
13:20:09	*	lnxw37d4 joined #nim
13:20:10	*	cvoxel joined #nim
13:21:36	*	xigoi[m] joined #nim
13:22:01	*	juanfra__ joined #nim
13:23:29	*	himup[m] joined #nim
13:23:40	*	cvoxel quit (Client Quit)
13:24:11	*	k0mpjut0r joined #nim
13:24:30	*	VijayMarupudi[m] joined #nim
13:24:42	*	jaens[m] joined #nim
13:24:42	*	hmmm joined #nim
13:25:39	*	Clonkk[m] joined #nim
13:25:42	Tanger	I've had great success with the npeg library so far Zevv, outstanding stuff!
13:25:50	*	Juno[m] joined #nim
13:28:16	Zevv	nice, I'll send out the bill today
13:34:25	*	k0mpjut0r quit (Ping timeout: 240 seconds)
13:34:26	*	BitPuffin quit (Ping timeout: 240 seconds)
13:34:39	*	himup[m] quit (Ping timeout: 246 seconds)
13:34:42	*	lnxw37d4 quit (Ping timeout: 258 seconds)
13:34:43	*	xigoi[m] quit (Ping timeout: 244 seconds)
13:34:55	*	Juno[m] quit (Ping timeout: 240 seconds)
13:34:59	*	Clonkk[m] quit (Ping timeout: 246 seconds)
13:35:00	*	jaens[m] quit (Ping timeout: 246 seconds)
13:35:04	*	VijayMarupudi[m] quit (Ping timeout: 260 seconds)
13:36:40	Tanger	XD
13:39:31	Tanger	While I remember, do you have an idea of what npeg's performance is like on the JS backend Zevv?
13:40:11	*	juanfra__ quit (Ping timeout: 260 seconds)
13:40:12	Tanger	I can't imagine the js version is anywhere near as optimized as C backend
13:50:03	*	Avahe[m] joined #nim
13:50:11	*	Kurre[m] joined #nim
13:50:12	*	cmc[m] joined #nim
13:51:39	*	cadmium[m] joined #nim
13:51:39	*	vindaar[m] joined #nim
13:51:39	*	j-james[m] joined #nim
13:51:46	*	stisa[m] joined #nim
13:51:48	*	planetis[m] joined #nim
13:51:49	*	Avatarfighter[m] joined #nim
13:51:51	*	leorize[m] joined #nim
13:52:34	*	BauxnaMatrix[m] joined #nim
13:52:35	*	dilawar_uchiha[m joined #nim
13:52:35	*	codic joined #nim
13:53:19	*	silvernode[m] joined #nim
13:53:20	*	Zoom[m] joined #nim
13:54:39	*	unclechu joined #nim
13:55:48	*	nyaa8 quit (Quit: byeee~)
13:56:16	*	matthias[m] joined #nim
13:56:28	*	MTRNord joined #nim
13:59:34	*	reversem3 joined #nim
14:03:35	FromDiscord	<hassel> sent a code paste, see https://play.nim-lang.org/#ix=2Ms0
14:04:01	*	lnxw37d4 joined #nim
14:05:15	*	lytedev[m] joined #nim
14:05:39	*	jaens[m] joined #nim
14:05:46	*	BitPuffin joined #nim
14:06:50	*	juanfra__ joined #nim
14:08:09	*	k0mpjut0r joined #nim
14:10:19	*	himup[m] joined #nim
14:10:23	*	oculux quit (Ping timeout: 260 seconds)
14:10:53	*	xigoi[m] joined #nim
14:13:03	*	Clonkk[m] joined #nim
14:13:42	*	Juno[m] joined #nim
14:16:10	*	VijayMarupudi[m] joined #nim
14:16:12	*	Tanger quit (Remote host closed the connection)
14:57:19	*	habamax quit (Ping timeout: 246 seconds)
14:58:54	giaco	sqlite3 docs says that threading mode can be selected at start-time or at run-time, do you know how to do that in nim? https://sqlite.org/threadsafe.html
14:59:05	giaco	or how to know current mode, too
14:59:26	giaco	it says that "The default mode is serialized." but I'd like to check
14:59:55	FromDiscord	<mratsim> @giaco, you need to check the signature of https://sqlite.org/c3ref/config.html
15:05:33	*	jkiesian quit (Ping timeout: 256 seconds)
15:06:46	FromDiscord	<whisperdev> Can someone help me figure out how to compile fswatch?
15:08:56	Zevv	Tanger: I don't know what to compare it to
15:08:56	leorize[m]	mratsim: I don't understand delimited vs undelimited continuations
15:11:07	leorize[m]	mratsim: as far as your description of "resumable" go, it seems comparable to how I'm using cps in nim-sys
15:19:19	*	hmmm quit (Quit: WeeChat 3.0)
15:26:09	*	PMunch quit (Quit: leaving)
15:26:49	leorize	mratsim: so here is something that I feel like I should remind you about: readiness is only one form of "non-blocking" I/O
15:27:11	qwr	leorize[m]: i think that in delimited continuation a limited scope is captured, which can be invoked again
15:28:13	qwr	leorize[m]: or maybe more accuratly, its execution can be continued...
15:30:26	leorize	mratsim: on Windows, and probably Linux's io_uring (haven't read too closely on that one), completion notification based async I/O requires that the buffer you hold be "borrowed" by the operating system, but you will still keep ownership of it and such it can not be freed until the I/O operation is completed
15:31:17	leorize	from a memory safety standpoint, a stack-based buffer doesn't work here and it has implications for cross-platform apis
15:31:19	FromDiscord	<Clyybber> @hassel afaik its not implemented
15:31:23	FromDiscord	<mratsim> undelimited continuation capture the whole program while delimited only some scope is resumable
15:31:56	FromDiscord	<mratsim> undelimited means making any Nim program suspendable which seems a bit overkill
15:32:26	leorize	qwr: thanks
15:32:42	leorize	so I believe the current CPS system is delimited?
15:32:51	FromDiscord	<mratsim> The buffer for completion Io is separate from continuation frame
15:32:54	FromDiscord	<mratsim> yes
15:33:22	FromDiscord	<mratsim> I don't think there is any reason to look into undelimited continuation.
15:33:49	leorize	I don't mind, I think the current system can support me well, if it compiles, that is
15:34:40	leorize	wdym separated from the continuation frame?
15:34:45	FromDiscord	<mratsim> i.e. libraries are free to put their buffers on the heap or stack
15:34:59	qwr	afaik delimited ones are more powerful due composibility (with undelimited continuations any control structures built on them can be broken by raw undelimited continuations)
15:35:28	FromDiscord	<mratsim> and similarly, any scheduler requires the continuation frame on the heap, but iterators that are created and consumed in the same scope can be put on the stack
15:36:08	giaco	mratsim, do you mean manually importing that proc from c?
15:36:13	FromDiscord	<mratsim> I think delimited continuations are also easier to implement and to maintain/verify
15:36:43	FromDiscord	<mratsim> @giaco, the sqlite doc mentions that single-threading or multithreading mode is managed by the sqlite_config proc
15:38:03	giaco	mratsim, yes, and that function is not present in nim sqlite api, that's what I'm asking
15:38:34	leorize	mratsim: there has to be a way to prevent the buffer from being freed during the operation
15:40:03	FromDiscord	<mratsim> @giaco: https://github.com/arnetheduck/nim-sqlite3-abi/blob/master/sqlite3_gen.nim#L4308
15:40:30	FromDiscord	<mratsim> @leorize, if you're continuation hold a ref to the buffer it won't be freed.
15:41:37	leorize	sure, though it does mean that I will have to do away with my dreams of openArray[byte]-only API :P
15:41:43	FromDiscord	<mratsim> We don't have the same problem as Rust with the GC for completionb ased APi (see https://github.com/weavers-guild/weave-io/blob/master/research/async_await_cancellation_rust_matthias247.md#support-for-io-completion-based-operations)
15:42:24	FromDiscord	<mratsim> you might be able to use openarray in struct?
15:42:30	giaco	mratsim, interesting. I wonder why it didn't show up when I searched exactly that signature on github
15:42:36	giaco	thanks
15:42:43	FromDiscord	<mratsim> it's a custom wrapper that we used at Status
15:42:45	*	narimiran quit (Quit: leaving)
15:43:16	FromDiscord	<mratsim> the high level wrapper is here: https://github.com/status-im/nim-eth/blob/master/eth/db/kvstore_sqlite3.nim
15:43:34	*	narimiran joined #nim
15:44:03	FromDiscord	<mratsim> @leorize it's quite important for me as well that IO avoids ref at all cost, or at least push that in the schedulers as they can go nuts with allocation or deal with their own memory pool
15:44:09	*	abm quit (Read error: Connection reset by peer)
15:44:18	*	jkiesian joined #nim
15:44:32	leorize	I want to avoid ref too, openArray is just much more flexible
15:44:43	*	narimiran quit (Client Quit)
15:44:46	FromDiscord	<mratsim> because IO is used also in embedded or kernel drivers that shouldn't use the GC. And also for "Weave-IO" I'dl like tight control over memory
15:45:06	*	narimiran joined #nim
15:45:07	FromDiscord	<mratsim> which is why I didn't use Nim closures
15:45:45	leorize	I wouldn't say that they shouldn't use the GC
15:45:56	FromDiscord	<mratsim> but the forcing the user to hold the buffer might just be requiring a "var openarray".
15:46:23	leorize	most of the time the kernel itself provides a GC
15:46:40	leorize	mratsim: yes, if that works :P
15:47:02	FromDiscord	<mratsim> C++ and Rust will have the same issues with completion-based API as no-GC continuations
15:47:25	leorize	nim-sys cps branch currently break cps because I'm using, uh... everything that is untested by the current implementation
15:50:03	FromDiscord	<mratsim> sent a long message, see http://ix.io/2Mt0
15:50:31	*	abm joined #nim
15:51:03	FromDiscord	<mratsim> This might be an inspiration, didn't look in the internals: https://github.com/spacejam/rio
15:51:20	*	narimiran quit (Quit: leaving)
15:55:32	*	narimiran joined #nim
15:56:15	leorize	I think if we can generate openArray entries for the current Env of CPS it should allow us to implement completion-IO, maybe
16:01:56	FromDiscord	<mratsim> what do you mean?
16:04:00	leorize	keeping the `var openArray` in the continuation environment then rely on the compiler's escape analysis to prove that the buffer outlives the continuation
16:04:17	giaco	would you help understand what is this project doing? https://github.com/arnetheduck/nbindgen why the output is a .h file?
16:04:49	giaco	what are "nim headers" in this context?
16:06:54	FromDiscord	<mratsim> @leorize, ah yes, I would be interested as well for weave
16:07:32	FromDiscord	<mratsim> if you have "var x = @[1, 2, 3]"
16:07:48	FromDiscord	<mratsim> right now you need to cast to ptr UncheckedArray to pass to Weave
16:08:06	FromDiscord	<mratsim> and some way to indicate that we want to capture as var openarray would be nice.
16:09:35	leorize	views are supposed to make this possible
16:10:05	leorize	but since Araq got his hands full with IC, I don't think we are getting fully functional views anytime soon
16:10:50	FromDiscord	<mratsim> we can make kludge that use casting to ptr UncheckedArray and then toopenArray(ptr, len)
16:11:12	FromDiscord	<mratsim> but we need trick to handle mutable seq so that they use var openarray instead
16:12:04	FromDiscord	<mratsim> Maybe we need a "closure RFC"
16:13:06	leorize	we can have tons of RFCs but we only have one Araq
16:13:33	*	Vladar quit (Remote host closed the connection)
16:13:56	leorize	compiler changes have the ultimate bottleneck that is one person
16:14:14	FromDiscord	<XxDiCaprioxX> Hey guys, can any one of you tell me how I can echo a string and int together in one line? somewhat similar to `echo "aaaaa" & someInt`
16:14:26	leorize	s/&/,/
16:14:54	leorize	!eval echo "aaaaa", 42
16:14:59	NimBot	aaaaa42
16:15:08	FromDiscord	<XxDiCaprioxX> okay thanks
16:15:28	FromDiscord	<lqdev> @XxDiCaprioxX alternatively, `echo "aaaa" & $someInt` to turn someInt to a string
16:15:38	FromDiscord	<XxDiCaprioxX> also good idea thanks
16:15:54	giaco	nim project has a bus factor of 1
16:16:00	leorize	and perform two allocations so may not be the best idea
16:16:58	FromDiscord	<mratsim> There are other people who know their way around the compiler, clyybber cooldome zahary krux02
16:17:09	FromDiscord	<mratsim> It's the vision that has a bus factor of 1
16:17:15	leorize	yes but how many of them are full time devs?
16:20:00	FromDiscord	<XxDiCaprioxX> is there a loop that only checks the condition after executing its content?
16:20:04	leorize	RFCs can get the "Accepted RFC" tag but Araq would still be the only person to implement them
16:20:37	leorize	XxDiCaprioxX: we don't have a do-while loop, but it's easy to do a `while true` with a `if cond: break` at the end
16:21:04	FromDiscord	<XxDiCaprioxX> okay and break exits the loop and moves on to the next line, correct?
16:21:13	leorize	yep
16:21:16	*	hmmm joined #nim
16:21:18	FromDiscord	<XxDiCaprioxX> aight thanks
16:23:37	*	jkiesian quit (Ping timeout: 264 seconds)
16:23:51	krux02	leorize, I was full time dev until November 2019
16:25:10	FromDiscord	<SirJosh> sent a code paste, see https://play.nim-lang.org/#ix=2Mta
16:25:30	FromDiscord	<SirJosh> as for "what i'm trying to achieve" i'm just currently prototyping with some random ideas and trying to model them 1:1 in a few target languages as best as possible
16:25:48	FromDiscord	<SirJosh> so no code horror that you need to worry about ending up in production or anything lol
16:25:49	leorize	remove the generic part and you're golden
16:26:21	leorize	the compiler prefers symbols that is more specific
16:26:49	leorize	so `add(seq[byte], byte)` will always be preferred over `add[T](seq[T], T)`
16:27:11	*	tane joined #nim
16:27:38	leorize	if you insist on keeping the `[T]`, then you can write `add[T: byte](seq[T], T)` for example
16:27:46	*	hmmm quit (Quit: WeeChat 2.8)
16:27:57	FromDiscord	<SirJosh> ah snap neat!
16:28:13	krux02	alternatively, you can have a `when` branch in the generic implementation
16:28:23	krux02	when T is int: ...
16:28:48	FromDiscord	<SirJosh> hmm good to know
16:29:32	leorize	krux02: iirc you and Araq disagrees a lot in terms of how compiler development should be done
16:29:43	krux02	yes
16:30:15	FromGitter	<HJarausch_gitlab> How to declare an *external* (Nim) proc, i.e. ⏎ I'd like to declare a proc fct(...) : ... as external to the current module. How do I do that?
16:30:43	krux02	you want to forward declare a funciton from another module?
16:30:57	krux02	you can't do that, you can only forward declar module local
16:31:36	krux02	generally you are supposed to arrange your modules in a way, so that a linear dependency graph can be created
16:32:02	leorize	we also have @zah, but he seems to have stopped working on the compiler, which I suppose is because of the workload he has on Nimbus
16:32:36	krux02	when I started on Nim, zah was already inactive
16:32:44	krux02	don't know if he did something recently.
16:32:54	krux02	It was because of his contract with Status
16:33:12	krux02	it was required that he works full time on status and stop commiting to the compiler.
16:33:13	FromGitter	<HJarausch_gitlab> But, I'd like to create a module which implements some optimization of a function *fct* without passing it to every function in that module as a parameter.
16:33:45	krux02	well, that is a bit tough to do.
16:34:01	krux02	you want to specialize
16:34:04	krux02	?
16:34:04	leorize	sounds like a thing concepts would solve, if it works, that is
16:34:21	krux02	no doesn't sound like cocepts, sounds like template specialization.
16:34:29	krux02	Nim does not have template specialization.
16:34:34	krux02	C++ has.
16:34:55	krux02	that feature was intentionally left out of the language.
16:37:07	FromDiscord	<mratsim> I don't think it's his Status contract, he did work on static because it was a blocker
16:37:14	krux02	@HJarausch is it possible that you were a Professor at RWTH Aachen?
16:37:32	FromDiscord	<mratsim> it's more like, when you do something for work 8-10 hours a day, you might want to do something else on your evnings
16:38:54	leorize	the only other person that is active on complicated features in the compiler is clyybber afaict
16:38:55	disruptek	like what?
16:40:07	FromDiscord	<mratsim> making CPS great again :p
16:40:19	krux02	No I had a chat with him. I think actually wanted to still contribute to the compiler. The people who payed him wanted that his primary focus is somewhere else. And I personally think that is a good.
16:40:20	FromDiscord	<mratsim> cooldome does stuff on destructors as well
16:42:02	*	a_b_m joined #nim
16:42:27	leorize	brand new features are still driven by Araq alone, though
16:42:59	FromDiscord	<mratsim> I do think we need a year of stability
16:43:16	*	blueberrypie6 joined #nim
16:43:33	FromDiscord	<mratsim> we have views, strictFunc, not nil that has been almost working for 18 months, plenty of generics and static crash still out there, the new arc/orc.
16:43:50	leorize	IC is supposed to help with all that given it's frontend simplification
16:44:16	FromDiscord	<mratsim> Fixing those and having more contributors coached into looking into those before introducing newer unstable features would be good for stability
16:44:48	*	Yardanico quit (Remote host closed the connection)
16:44:48	*	letto quit (Remote host closed the connection)
16:45:09	*	letto joined #nim
16:45:18	*	abm quit (Ping timeout: 256 seconds)
16:45:18	*	blueberrypie quit (Ping timeout: 256 seconds)
16:45:18	*	idf quit (Ping timeout: 256 seconds)
16:45:18	*	blueberrypie6 is now known as blueberrypie
16:45:31	FromDiscord	<mratsim> Though I think dynamic dispatch without ref (VTable) is very important to grow/fix the stdlib.
16:45:37	*	Yardanico joined #nim
16:46:07	FromDiscord	<mratsim> but maybe Yuriy interfaced (was it) is enough has a library solutiion
16:46:28	*	idf joined #nim
16:47:37	disruptek	leorize: did you fix the 1.4 archive? it was corrupted last night. all my CIs are red.
16:48:35	narimiran	"I do think we need a year of stability" -- that's roughly the plan for 2021 indeed
16:48:43	leorize	disruptek: try re-running them?
16:49:35	leorize	mratsim: new concepts are supposed to help fix that I believe
16:50:04	FromDiscord	<mratsim> I'm skeptical of new concepts in terms of applicability.
16:50:06	disruptek	leorize: i am doing that now.
16:50:54	leorize	mratsim: they look like interfaces to me, and statically bound at compile time
16:51:09	FromDiscord	<mratsim> So like Rust traits?
16:51:28	leorize	shouldn't be too hard for a macro to make runtime dispatching based on them either
16:51:49	leorize	looks like it, it's simply "checked" generics afaict
16:52:07	leorize	fairly limited compared to what we are having now, but should be expandable as time goes
16:52:21	leorize	the important feature it might bring is that it works
16:52:34	FromGitter	<HJarausch_gitlab> @krux02 Not a professor but a *Privatdozent* - but why does that matter?
16:53:10	krux02	We worked together in teaching C++ to math students
16:53:33	disruptek	leorize: still broken.
16:53:50	leorize	disruptek: link to CI please?
16:54:04	disruptek	https://github.com/disruptek
16:54:19	FromDiscord	<mratsim> The current concepts do work. They are a pain to debug but with a bit of pixie dust they do work.
16:54:23	FromGitter	<HJarausch_gitlab> @krux02 and what are you doing now?
16:54:31	disruptek	no, the current concepts do not work.
16:55:03	FromDiscord	<mratsim> well they work for me then
16:55:31	krux02	After I worked for a year full time for the Nim compiler I now am back at the University.
16:56:06	krux02	Remotely from Home.
16:56:24	leorize	mratsim: if you need pixie dust to get them to work then I would not say that they work
16:56:35	FromGitter	<HJarausch_gitlab> @krux02 in what faculty and in which university?
16:57:22	krux02	still RWTH institut für Textiltechnik Aachen. I am working on a ray tracer to simulate optical fibers.
16:57:36	krux02	but it is temporary. I don't see my future there.
16:58:13	krux02	Friends from university want me to join other companies, something with future.
16:58:24	FromGitter	<HJarausch_gitlab> @krux02 Good luck, then!
16:58:44	krux02	And what are you doing after you are done with C++?
16:58:56	krux02	learning some new programming languages?
16:59:20	*	waleee-cl joined #nim
17:00:24	leorize	disruptek: looks like the linux 64bit archive was broken, I replaced it
17:00:28	leorize	should work now
17:01:55	FromGitter	<HJarausch_gitlab> @krux02 I'm retired since 2015. And yes, I still like to learn new programming languages (having got in touch with more than a dozens of them)
17:01:57	disruptek	aight thanks.
17:02:49	FromDiscord	<dom96> > "I do think we need a year of stability" -- that's roughly the plan for 2021 indeed↵the blog post writes itself 🙂
17:02:49	krux02	I know you are retired. I was tutor in your last semester.
17:03:11	krux02	Unfortunately it was also my last semester, so I can't tell you how it developed after your retirement.
17:04:05	krux02	You certainly remember me. I am the very tall guy with long hair.
17:05:08	leorize	disruptek: have you figured out any viable fix for my `var openArray[T]` usage with cps? or do I have to settle with pointers?
17:05:15	krux02	I certainly did not expect to see you again here.
17:05:21	krux02	Nice to see you still active.
17:08:41	FromDiscord	<mratsim> @leorize the pixie dust is needed to understand why they don't match because the error message is a hard to get/reduce
17:09:05	FromGitter	<HJarausch_gitlab> @krux02 Yes, your image was familiar to me. Now, that you are a professional yourself, what do you think about the style of the C++ course (please an honest comment!) (The style of the C++ course has changed completely, more like most books in programming, while I preferred *learning by doing*.
17:09:33	leorize	mratsim: I get the mysterious matched in one line but not another
17:10:13	FromDiscord	<mratsim> I usually put "static: doAssert Foo is Bar" when i use concept now
17:10:18	leorize	`is` also get magically "not declared" for typedesc so my requirements in concepts no longer work
17:10:34	FromDiscord	<mratsim> just in case someone introduces regression that would affect me
17:10:55	disruptek	leorize: i need to merge a huge branch before i tackle generics, unfortunately, but afaict you don't need openarray in the env.
17:11:11	leorize	I may need it actually
17:11:17	leorize	I need the buffer to stay alive
17:11:26	*	Evolver quit ()
17:11:33	krux02	@HJarausch do you mean the course that we had together?
17:12:02	*	Evolver joined #nim
17:12:09	disruptek	i think the closure solution is almost as nice to use, and should work without any fresh bugs.
17:12:50	FromDiscord	<mratsim> but @leorize isn't it better to pass the buffer to 2 different proc? One that initiate the "transaction" and one that reads and drop the buffer?
17:13:15	disruptek	why does the first proc need the buffer?
17:13:22	FromDiscord	<mratsim> for the low-level at least
17:13:28	FromGitter	<HJarausch_gitlab> @krux02 Yes, but more the general concept
17:13:28	*	natrys joined #nim
17:13:30	FromDiscord	<mratsim> Completion based API requirement
17:13:45	krux02	My honest opinion is, the course was very challenging to overwhelming for people who just started with programming. I completely agree that learning by doing is the right way to learn any programming langugae. But it should be something that is at least on the theoretical level already familiar or easy, so that the students can focus on getting along with the language instead of struggling with understanding the problem.
17:13:48	FromDiscord	<mratsim> for zero-copy, windows and Io uring need a buffer so they copy the input directly inside
17:13:51	*	Vladar joined #nim
17:13:58	disruptek	leorize: it seems like the 1.4 tarball is not working yet. i'm trying again.
17:14:14	leorize	mratsim: isn't in your design we end up copying anyway?
17:14:17	FromDiscord	<mratsim> instead of having a kernel buffer and then a userland buffer in epoll/readiness based API
17:14:23	disruptek	that doesn't answer my question.
17:14:24	FromDiscord	<mratsim> no
17:14:31	disruptek	you don't need the buffer until you need to write into it.
17:14:42	disruptek	ergo, only one proc needs it as an argument.
17:14:44	FromDiscord	<mratsim> you don't copy, you have the nim-sys end user providing the buffer
17:14:59	FromDiscord	<dom96> the kernel copies though, no? So how can you call it "zero-copy"?
17:15:07	FromDiscord	<mratsim> you need it when you initialize the transaction disruptek
17:15:12	disruptek	https://github.com/disruptek/balls/runs/1723038526?check_suite_focus=true
17:15:14	FromDiscord	<mratsim> i don't think you need it after though
17:15:16	leorize	I need the buffer even when I'm not writing into it, because the OS will be doing that until it say that it's done
17:15:29	disruptek	mratsim: one proc.
17:15:48	FromDiscord	<mratsim> @dom96 the kernel copies but it's zero copy overhead over the necessary "recv"
17:16:30	FromDiscord	<mratsim> so @leorize i think the issue of keeping the buffer alive is in the higher-level layer than nim-sys
17:16:37	FromDiscord	<mratsim> likely the IO scheduler.
17:16:54	FromDiscord	<mratsim> but maybe you can write pseudo code
17:16:57	krux02	I think it was very good to give them real challenges, and it is impossible to provide something that evokes the interest of every student the same way, so I don't have a realy answer as well.
17:17:00	FromDiscord	<mratsim> so that we can have a look
17:17:14	disruptek	that's why i think the solution is a magic.
17:17:29	leorize	but nim-sys has to provide the IO scheduler
17:17:58	FromDiscord	<mratsim> why?
17:18:10	disruptek	mratsim: he already has a real impl; no pseudo-code.
17:18:10	leorize	I need to dispatch the completion call
17:18:12	FromDiscord	<mratsim> it's better to decouple low-level APi from the Io scheduler
17:19:15	FromDiscord	<mratsim> how would I use nim-sys from Weave-IO for example?
17:19:28	krux02	@HJarausch: I think the best way to motivate students to do programming is, if they write something that is either fun to use or actually useful.
17:19:46	krux02	Example here is the 4 in a row game that they implemented. I think that was a very good exercise for the students.
17:19:50	leorize	mratsim: you use the blocking APIs, that's the beauty of threads
17:20:03	*	Torro quit (Quit: bye)
17:20:18	krux02	The exercise with encryption was very dry.
17:20:27	FromDiscord	<mratsim> If I want to optimize for latency I likely should use the non-blocking API
17:20:44	leorize	why?
17:20:59	krux02	Also I would leave the usage of lib gmp for the advanced lecture, C++ is already surprising and challenging enough without it.
17:21:13	krux02	keep it simple in the beginning.
17:21:29	leorize	you already got a thread, use the blocking API and the OS will keep it asleep until the operation is done
17:21:49	FromDiscord	<mratsim> Because an optimal scheduler is "as long as there is work to do, all threads are busy".↵The Cilk scheduler has proven that this "greedy scheduler" approach is at most 2x slower than the fastest scheduling possible.
17:22:21	leorize	you can also spin via repeated `poll()` if you must
17:22:32	FromDiscord	<mratsim> But I can't manage millions of sockets, files, pipes and what not at the same time this way
17:22:44	leorize	why are you using threads then?
17:23:04	leorize	at least that's my assumption when I heard about weave-io
17:23:04	FromDiscord	<mratsim> This is not either/or.
17:23:22	krux02	Image processing was also nice. But C++ is a high performance language. Those written programs should have been interactive to see the performance of c++ better.
17:23:27	FromGitter	<HJarausch_gitlab> @krux02 It was always my intention to show some concepts of computer science, as well - since the students wouldn't hear about that elsewhere otherwise
17:23:56	FromDiscord	<mratsim> It's about having a multithreaded IO scheduler but it would be non-blocking as well.
17:24:21	krux02	yea I get your intentions, and they are good intentions. But I think it was a bit too much at the same time for the students.
17:24:36	leorize	mratsim: the lowest level you can go is the OS interface itself
17:24:40	FromDiscord	<mratsim> so a threadpool, like Weave and a scheduler that optimize for latency on top. But I don't want one thread per socket.
17:24:52	krux02	and yes it is sad that they wouldn't soo those CS concepts elsewhere.
17:24:57	leorize	I don't want to give out a selectors-like system, because IMO it's dumb
17:25:30	FromDiscord	<mratsim> sure. What I'm saying is that, ideally nim-sys or flywind ioselector are just thin wrapper over the OS interface, that just give me the OS interface in a consistent API.
17:26:05	FromDiscord	<mratsim> thin as is, openarray[byte] but don't do allocation or scheduling decision, leave them to higher-level schedulers so that they can tune them for their use cases.
17:26:21	leorize	that's a bit too thin
17:26:29	FromDiscord	<mratsim> obviously using continuation is fine.
17:26:32	leorize	my async ops needs to know how to schedule themselves
17:26:49	leorize	ie. the fd is not ready, I want to signal a wait
17:27:07	FromDiscord	<dom96> why do you think selectors-like system is dumb?
17:27:22	FromDiscord	<dom96> It mirrors the selectors APIs of POSIX-compatible systems
17:27:41	leorize	that is exactly why
17:27:50	leorize	selectors-like system assume a readiness-based scheme
17:28:06	FromDiscord	<dom96> yeah, because that's what most OS' implement
17:28:11	leorize	you try to fit iocp in and you become the buffer
17:28:15	leorize	they are moving out of it
17:28:22	FromDiscord	<mratsim> so that's one thing. Signal/condition-variable and locks are meeh. After fighting with Glibc condition variable bugs, i just want to use Futex.
17:28:26	FromDiscord	<dom96> iocp is out of the scope of selectors
17:28:37	leorize	io_uring is a completion-based system too
17:28:47	FromDiscord	<dom96> you build a completion base system on top and fit iocp into that
17:28:51	FromDiscord	<dom96> which is what asyncdispatch does
17:28:59	FromDiscord	<dom96> (edit) "base" => "based"
17:29:04	FromDiscord	<mratsim> but that brings a lot of extra allocations.
17:29:10	leorize	I want my async ops to be efficent, because that is what users use
17:29:41	disruptek	weird.
17:31:32	*	junland quit (Quit: %ZNC Disconnected%)
17:32:44	leorize	disruptek: should work now, I just tested it
17:33:43	FromDiscord	<dom96> My point is that selectors is a fundamental API of the operating system, you have to have this thin wrapper to support Linux/BSD
17:33:55	FromDiscord	<dom96> with io_uring it might be different now, but BSD is still stuck with this AFAIK
17:34:18	leorize	mratsim: my eventqueue implements the APIs that nim-sys will use to schedule itself
17:34:36	FromDiscord	<dom96> but yeah, it is funny that Windows got it right for such a long time
17:34:56	leorize	mrastim: if you want it can be an interface and you may plug your implementation in
17:35:38	FromDiscord	<mratsim> Yeah, it would be nice if the low-level proc can be used to implement alternate schedulers.
17:36:02	leorize	dom96: my point is that all this should be a part of a dispatcher for efficiency reasons
17:36:12	leorize	I'm just skipping the middleware that is selectors
17:36:37	FromDiscord	<mratsim> What difference do you make between a scheduler and dispatcher?
17:37:29	*	vicfred joined #nim
17:37:30	FromDiscord	<dom96> leorize: I don't really see your point. You still need to implement this API to support BSD at least, unless you are planning to drop support for that platform.
17:37:41	*	vicfred quit (Remote host closed the connection)
17:38:32	*	junland joined #nim
17:38:33	FromDiscord	<mratsim> Windows is 90% of home users, Linux is 99% of servers. Doesn't make sense to hold back their performance for 1% of the marketshare left.
17:39:18	FromDiscord	<dom96> Sure, if I was writing an async package today I would drop it too
17:39:56	FromDiscord	<mratsim> Leorize doesn't drop epoll/readiness-based API. But since 90% won't use them, the selector API is skipped.
17:40:43	*	vicfred joined #nim
17:41:28	FromDiscord	<mratsim> so low-level OS APi are stille wrapped. You still have a high-level abstraction. But there is nothing in-between, the OS/API differences are handled in the scheduler directly instead of in the selectors intermediate API
17:42:03	leorize	yep, that is how it's done right now in nim-sys
17:42:31	leorize	CPS enables all this to work really nice
17:42:38	leorize	once it works, that is
17:42:44	FromDiscord	<dom96> Maybe something changed in the selectors implementation since I last used it, but I don't see how that's different. You'll end up with the same implementation.
17:43:21	FromDiscord	<mratsim> The first thing i see when i look into selectors is that it's a ref object
17:43:38	FromDiscord	<mratsim> there is need for ref in a reainess API
17:44:16	FromDiscord	<mratsim> That's the whole zero-cost futures part of Rust
17:44:41	FromDiscord	<mratsim> using the fact that POSIX so far was readiness-based which allowed allocating the buffer only at the final use, and so no alloc.
17:44:45	leorize	dom96: I have more free reign in implementation and can apply the most suitable restrictions
17:45:57	FromDiscord	<mratsim> Because once you don't need heap alloc for futures, you can use Rust iterators/state-machine scheme to have very efficient future chaining.
17:46:00	FromDiscord	<dom96> Sure, that's fine. But calling `selectors` dumb is an over-simplification
17:46:24	FromDiscord	<dom96> In a stdlib it's a great abstractions which allows anyone to write a fast selectors-based server
17:46:34	FromDiscord	<dom96> (edit) "abstractions" => "abstraction"
17:47:10	FromDiscord	<mratsim> I don't call solutions dumb without having a 5k words README with lengthy analysis.
17:47:30	FromDiscord	<mratsim> so leorize you're on your own 😉
17:47:55	leorize	it was more about how I won't implement a selectors in nim-sys :P
17:48:29	disruptek	leorize: it works, thanks.
17:49:13	FromDiscord	<mratsim> I planned to explore raw OS IO API as well for Weave IO but if you expose those in nim-sys that would be helpful
17:49:28	leorize	mainly because I found the abstractions provided by the dispatcher + cps to be more than enough to cover the grounds of nim-sys
17:49:42	FromDiscord	<dom96> @mratsim it's already exposed in the stdlib
17:50:56	FromDiscord	<dom96> also I'm curious why you think there is a need for ref in a readiness API
17:50:58	leorize	you can read this to see how a posix non-blocking read is done with cps: https://github.com/alaviss/nim-sys/blob/cps/src/sys/private/files_posix.nim#L46-L70
17:51:15	FromDiscord	<mratsim> there is no need for ref in a readiness API
17:51:18	leorize	no windows-based version yet since I'm still experimenting
17:51:20	FromDiscord	<mratsim> selectors use ref
17:51:36	FromDiscord	<dom96> was that a typo above?
17:51:47	disruptek	yes.
17:51:53	FromDiscord	<mratsim> yes
17:52:29	FromDiscord	<mratsim> @leorize, can you have an error model/Result instead of exceptions?
17:52:37	leorize	no
17:53:04	disruptek	otherwise there'd be no motivation to upgrade your nim installation.
17:54:42	FromDiscord	<mratsim> Well, I'm also worried of try:except wrappng a suspension point.
17:54:48	leorize	dom96: with cps the dispatcher and the selector just blend together naturally, and remove the need for a separated selectors middleware imo
17:54:58	FromDiscord	<konsumlamm> does anyone happen to know what heapdumprepl is doing?
17:55:03	FromDiscord	<mratsim> those are broken in current closure iterators, Swift async and C++ coroutines also disallow them iirc.
17:55:59	leorize	IMO it's not that selectors is bad, but having it separated from the dispatcher is
17:57:47	FromDiscord	<dom96> Perhaps. Like I said, it was the right call for the stdlib as it allows anyone to skip asyncdispatch altogether and just use selectors directly.
18:03:42	leorize	mratsim: maybe you can copy what --exceptions:goto does? basically automated Result[T] for the most part
18:04:25	FromDiscord	<mratsim> I'll see, i'm too busy to look in-depth at IO low-level stuff right now
18:04:58	FromDiscord	<mratsim> I'm mostly interested to understand the ergonomics or technical blocker in CPS to enable smooth IO atm
18:05:53	*	hoijui quit (Ping timeout: 246 seconds)
18:06:08	FromDiscord	<mratsim> Also to understand if we really need to handle the raw low-level continuations beyond the public API I proposed
18:06:27	disruptek	well, the blocker is you, buddy.
18:07:30	leorize	mratsim: so you would like nim-sys' eventqueue to be replaceable by the end user to hook into their own scheduler? doesn't sound bad to me
18:07:45	disruptek	leorize: that's the beauty of cps.
18:08:00	FromDiscord	<mratsim> What blocker is me?
18:08:31	disruptek	waiting for you to PR type eraser into 0.0.16.
18:08:38	disruptek	you said you would not do so. 🤷
18:08:51	leorize	disruptek: I know, that is why it is designed as it is right now :)
18:08:53	FromDiscord	<mratsim> Yes, and I would like the low-level API to not impose the allocation/exceptions things
18:08:58	leorize	I have to finalize the API first though
18:09:14	leorize	mratsim: get openArray to work first :)
18:09:19	FromDiscord	<mratsim> @disruptek, I said that it was more important to figure out the high level ergonomics.
18:09:33	FromDiscord	<mratsim> because we agreed on what to do for type erasing.
18:09:55	leorize	as for exceptions, I will keep using it just so you have a motivation to have Araq get rid of ref for exceptions :)
18:10:01	disruptek	well, i don't care what your excuse is. the fact is, we don't have the eraser and we agreed to use it.
18:11:08	leorize	exceptions goto already make exceptions perform similar to result codes (because it kinda is)
18:11:25	*	jkiesian joined #nim
18:11:29	FromDiscord	<mratsim> Well, I choose how to use my free time. I don't care what your demands are.
18:11:40	leorize	complain away if you don't want exceptions to be on the heap, it will help the ecosystem at large :)
18:11:52	disruptek	i demand nothing.
18:12:22	disruptek	also, i continue to diverge further from your work without regard to what pain it may cause you.
18:12:40	FromDiscord	<mratsim> You are always asking me to prove things to you.
18:12:57	disruptek	i agreed to your type eraser and even this is not enough for you.
18:13:01	disruptek	i dunno what to tell you, man.
18:13:23	disruptek	i agreed to splitting the code-base to stack-based and heap-based and this is not enough for you.
18:13:50	disruptek	just let me know what would motivate you to spend free time on cps and maybe if that scenario arises, i'll let you know.
18:14:42	FromDiscord	<mratsim> I don't spend all my time on CPS because I'm doing other stuff for Nim ecosystem, in particular related to science.
18:14:51	disruptek	noted.
18:15:12	FromDiscord	<mratsim> Unfortunately, I'm also a blocker here, so I want to remove myself from being a blocker here.
18:15:45	FromDiscord	<mratsim> because I can't keep up as a sole developer with all the things that are output by the various scientific communities.
18:16:20	FromDiscord	<mratsim> Also because automatted wrapper are still a pain, I took a week to wrap stuff manually and show how people can wrap what is meaningful to them.
18:16:51	FromDiscord	<mratsim> and before as mentioned I had internet issue.
18:17:21	FromDiscord	<mratsim> If people can triplicate me so I can do cryptography, science and CPS at the same time, sure.
18:17:55	disruptek	i already said i demand nothing and i don't care what your excuse is. relax.
18:18:12	disruptek	also, don't blame us for continuing to develop without you.
18:18:14	disruptek	that is all.
18:18:17	FromDiscord	<Clyybber> then stop calling it excuse maybe?
18:18:24	FromDiscord	<mratsim> Did i blame you?
18:18:35	disruptek	s/excuse/rationale/
18:18:46	FromDiscord	<mratsim> Did I raise an issue about having to adapt to your changes?
18:19:05	disruptek	look, we don't even have to talk about this.
18:19:19	disruptek	i'm supposed to be renaming my balls again.
18:19:20	FromDiscord	<Daniel> 🍿
18:19:24	FromDiscord	<mratsim> well it seems like we do since you mentionned that I am the blocker
18:19:56	disruptek	no; it's clear that you don't wish that role.
18:21:54	FromDiscord	<mratsim> Type-erasure isn't blocking CPS. That's purely internal and something that we can handle at our leisure.
18:22:22	FromDiscord	<mratsim> High-level API is something we want figured out so that leorize can plan around it
18:22:36	FromDiscord	<mratsim> or anyone who wants to rethink asyncstreams for example
18:23:18	FromDiscord	<mratsim> unless there are architectural change sthat you want to land?
18:24:26	disruptek	i need type erasure because i need to figure out how we do results.
18:39:44	*	jkiesian quit (Ping timeout: 272 seconds)
18:43:43	*	muffindrake quit (Quit: muffindrake)
18:45:43	FromDiscord	<mratsim> understood
18:46:09	ehmry	is there a way to get hide effects of a proc?
18:46:18	ehmry	*a way to hide effects
18:47:57	ehmry	I'm trying to implement a Stream object but the backend is async
18:48:47	giaco	I have a sqlite db with TEXT column that I use to store binary data from nim seq[byte] converted to string. It is correctly inserted as I can read correct bytes using external db manager, but selecting from nim "db.getValue(sql"SELECT data FROM mytable WHERE id=?", 50)" returns a string len 12 for any row no matter the original buffer size (that ranges from 200 to 1500 bytes)
18:52:33	FromDiscord	<Clyybber> ehmry: See https://github.com/nim-lang/RFCs/issues/258
18:52:34	disbot	➥ 'cast' as pragma
18:52:52	FromDiscord	<Clyybber> I think you want {.cast(tags: [Tag1, Tag2, ...]).}: ...
18:54:46	*	narimiran quit (Read error: Connection reset by peer)
18:55:10	*	narimiran joined #nim
18:59:25	ForumUpdaterBot	New thread by Jiyinyiyong: How to detect Chinese character with regex?, see https://forum.nim-lang.org/t/7399
19:01:11	ehmry	on second thought I think I don't want Stream at all if there are futures going around
19:05:51	*	j-james joined #nim
19:06:35	*	j-james quit (Client Quit)
19:07:56	*	j-james joined #nim
19:12:13	*	Vladar quit (Quit: Leaving)
19:15:25	Araq	ehmry, you can also ask for the inferred infects via std / effecttraits
19:17:03	*	j-james quit (Quit: WeeChat 3.0)
19:18:05	*	Vladar joined #nim
19:19:12	ehmry	ok, that was what I was looking for a bit ago
19:22:54	ehmry	the only thing I remember liking about Go was the read/write interface thing, which I'm not if I miss it with Nim
19:23:26	ehmry	Stream seems like a nice idea but I keep trying to implement it with `await` and things get awkward
19:23:38	ehmry	er, `waitFor`
19:24:05	Araq	async streams are in devel now? not sure, PMunch worked on it and I told Miran to merge it
19:24:23	FromDiscord	<Clyybber> Araq: I think that were socket streams no?
19:24:55	Araq	hmm
19:25:03	ehmry	a FutureStream replacement?
19:26:45	FromDiscord	<dom96> Would be nice
19:26:55	*	jkiesian joined #nim
19:28:05	FromDiscord	<konsumlamm> hey Araq, do you remember what heapdumprepl is doing?
19:28:29	Araq	yeah ... no .. maybe
19:28:34	Araq	does it use Sqlite?
19:29:05	leorize	disruptek: please let me know once you got result working on cps
19:29:17	leorize	for now I'll continue using asyncdispatch
19:29:24	FromDiscord	<Clyybber> leorize: Is it blocking you?
19:29:43	leorize	yea, my read need to return how much was read
19:30:12	FromDiscord	<konsumlamm> heapdumprepl? no, it seems to render a graph of the heap or something
19:30:18	leorize	@Clyybber I can use ptr to walkaround `var openArray`, but I don't want ptr int for result
19:30:32	FromDiscord	<Clyybber> Heh, I was about to suggest ptr int
19:30:51	Araq	ah well
19:31:01	FromDiscord	<konsumlamm> looks pretty abandoned though and i couldn't find any references to it
19:31:04	Araq	just forget it, the graphs are enormous
19:31:13	Araq	I never saw anything in them
19:32:33	Araq	the code was written to debug the GCs
19:34:02	disruptek	leorize: i will; it may be a few days.
19:34:43	leorize	np, I need to write tests too
19:34:47	FromDiscord	<konsumlamm> (i'm currently trying to replace all occurrences of intsets with packedsets, that's how i stumpled upon heapdumprepl)
19:35:10	Araq	ugh
19:35:54	Araq	poor human beings who cannot keep up with these renamefests
19:39:42	leorize	Araq: how is not nil nowadays?
19:40:16	leorize	I noticed that you merged alehander `nil check` pr, but I don't know what exactly does it do
19:40:32	Araq	it's opt-in via .experimental: "strictNotNil"
19:45:18	leorize	doesn't seem to be documented in the change log though
19:45:24	*	ee7[m] joined #nim
19:45:38	leorize	~notnil
19:45:39	disbot	no footnotes for `notnil`. 🙁
19:45:47	leorize	~notnil is https://nim-lang.github.io/Nim/manual_experimental.html#strict-not-nil-checking
19:45:48	disbot	notnil: 11https://nim-lang.github.io/Nim/manual_experimental.html#strict-not-nil-checking
19:45:56	Araq	haven't got into it, I merged it to avoid it from bitrotting, will clean it up before the next release
19:50:21	FromDiscord	<Recruit_main707> Araq, today @Clyybber talked about making --warning[GcMem] work with arc.↵(although my interest is to add them to regions), how feasible it is to do this, what kind of stuff needs to be done?
19:51:09	FromDiscord	<Clyybber> @Recruit_main707 It's really just checking for ref/string/seq usage.
19:51:42	FromDiscord	<Clyybber> I'm not sure it actually makes sense for strings because they aren't using a "gc" really, even with orc
19:51:51	FromDiscord	<Clyybber> Because strings cannot create cycles
19:52:18	FromDiscord	<Clyybber> And similar for other types consisting of seq/ref that cannot be cyclic
19:52:53	Araq	https://github.com/nim-lang/RFCs/issues/177#issuecomment-762230174 is what we should do
19:52:55	disbot	➥ Unify Nim's GC/memory management options ; snippet at 12https://play.nim-lang.org/#ix=24Ua
19:53:00	FromDiscord	<Clyybber> Maybe a warning[HeapAlloc] makes more sense
19:54:08	*	Jesin quit (Quit: Leaving)
19:54:28	FromDiscord	<Clyybber> Since it could be for example that seqs will not be a compiler builtin anymore, but instead defined as a library
19:54:38	FromDiscord	<Clyybber> Araq: Yeah, that would be neat too
19:55:10	Araq	I don't like --gc:none nor its warnings, not even C compiler offer a --noMalloc switch
19:56:04	leorize	@Clyybber are you working on any features atm?
19:58:43	FromDiscord	<Recruit_main707> yeah, i also mentioned is kinda dumb you cant use refs of any kind, but having a warning for when you are not deallocating (or at least not allocating it in a MemRegion) something using gc:regions would make stuff easier
20:01:33	FromDiscord	<Clyybber> leorize: I'm working on #16195 atm
20:01:39	disbot	https://github.com/nim-lang/Nim/pull/16195 -- 6fix #16185 ; snippet at 12https://play.nim-lang.org/#ix=2Mu9
20:02:47	leorize	are new concepts and fully functional views on the roadmap for the next release?
20:03:24	Araq	fully functional views need a new RFC, the old design had unfixable problems
20:04:58	Araq	concepts ... I don't know.
20:05:15	FromDiscord	<Clyybber> leorize: See the discussion Araq cooldome and I had on telegram
20:05:26	Araq	the current plan is that 1.6 is about IC
20:07:15	Prestige	Is there a particular plan for 2.0? Reserved for when we want potentially breaking changes?
20:07:31	leorize	looks like I can't wait for cps + openArray to be a thing then
20:08:13	FromDiscord	<Recruit_main707> Prestige: IC /afaik/
20:08:24	Prestige	I thought that was 1.6
20:08:45	Araq	plan for 2.0 is to remove features
20:08:49	leorize	2.0 might do away with macros as they are right now
20:08:50	FromDiscord	<mratsim> we "only" need closure that capture seq/array as openarray
20:09:37	Prestige	Better macro system incoming?
20:09:51	*	Jesin joined #nim
20:09:57	Araq	in my ideal world 2.0 would *only* remove features and we added all the things we need to 1.x.0
20:10:23	leorize	versioned nim modules when :P
20:10:42	Prestige	leorize: before 2.0 :P lol
20:11:01	*	jkiesian quit (Ping timeout: 264 seconds)
20:14:01	FromDiscord	<mratsim> 2.0 removes features and 2.2 adds more
20:15:39	FromDiscord	<dom96> Nim 2.0 is just macros
20:16:09	disruptek	leorize: i'm not convinced that cps + openArray makes sense for anyone.
20:16:20	leorize	there are no macros in nim 2.0, only compiler plugins :P
20:16:33	Araq	Prestige, better macro system ... yeah what leorize said
20:16:48	Prestige	Interesting
20:16:52	FromDiscord	<Clyybber> eh, macros are already compiler plugins. They just run in the VM
20:17:00	FromDiscord	<dom96> Yeah
20:17:21	leorize	packedast will allow macros to be run as actual compiled programs afaict
20:17:22	FromDiscord	<dom96> Just expose more through macros. Job done. 😄
20:17:24	FromDiscord	<Clyybber> But of course compiler plugins could be extended to hook between certain passes
20:17:41	Araq	dom96: that's one way of phrasing it, yes
20:17:47	Prestige	I assume you can achieve the same work with it - I haven't seen a "compiler plugin" (aside from macros)
20:17:51	FromDiscord	<mratsim> @disruptek, closure being able to capture openarray would be helpful to update a seq in general.
20:17:56	FromDiscord	<mratsim> or a buffer
20:18:16	FromDiscord	<mratsim> this is the reason why I need to cast seqt to ptr UncheckedArray in Weave examples.
20:18:52	leorize	disruptek: it enables a safe interface to access a buffer
20:19:13	leorize	it can be a pointer, a string, a seq, or whatever, we have conversion utilities to turn them into openArray
20:19:17	FromDiscord	<mratsim> Nim 2.0: interfaces 😉
20:19:57	FromDiscord	<Clyybber> There is no need for compiler magic for interfaces IMO
20:20:33	leorize	the responsibility of what type to choose and what kind of buffer to use will then be pushed to the caller as only they know what they want
20:21:08	FromDiscord	<Clyybber> Araq: Do we include seq/string indexing under the term array indexing ?
20:21:28	FromDiscord	<dom96> Nim 2.0 will just have interfaces in the stdlib. Implemented using macros 😄
20:21:35	disruptek	leorize: all i'm saying is that continuations holding openArray fields is probably not a thing.
20:22:30	*	natrys quit (Quit: natrys)
20:22:33	leorize	disruptek: I don't care too much about the specifics, as long as you can keep the safety analysis and guarantees of openArray I will take it
20:23:04	disruptek	yes, i want this too.
20:23:12	*	hoijui joined #nim
20:23:29	FromDiscord	<Clyybber> then closures need to inherit the lifetimes of their captured variables first, thats a thing on my TODO :p
20:24:56	disruptek	right now i'm trying to figure out how jason gets "correctly" excluded from dist/1.[02].
20:25:38	FromDiscord	<iWonderAboutTuatara> Would it be possible for contributions to Nim or opensource Nim projects/ecosystem to be marked as community service?
20:26:01	disruptek	dui, huh
20:26:59	leorize	openArray and var params working with cps would be the best thing
20:27:30	FromDiscord	<Clyybber> @iWonderAboutTuatara Is that a github feature?
20:27:57	narimiran	so you make a 5 PRs instead of going to prison?
20:28:28	disruptek	minor possession, huh
20:30:17	FromDiscord	<iWonderAboutTuatara> lmao
20:30:19	FromDiscord	<iWonderAboutTuatara> no for colleges
20:30:26	FromDiscord	<iWonderAboutTuatara> @Clyybber don't think so but would be nice
20:30:50	FromDiscord	<Clyybber> oh, I wasn't familiar with the term :D
20:31:02	FromDiscord	<iWonderAboutTuatara> wait which term?
20:32:53	FromDiscord	<Clyybber> community service
20:33:25	FromDiscord	<iWonderAboutTuatara> oh I see
20:33:35	FromDiscord	<iWonderAboutTuatara> it's a very vague term, doesn't mean too much
20:35:38	Zevv	https://github.com/edubart/minicoro
20:37:12	FromDiscord	<lqdev> why does getOccupiedMem() return 0 under ARC?
20:38:33	Zevv	lqdev: how's the parsing adventure going then
20:38:58	FromDiscord	<lqdev> Zevv: ah pretty good so far, just fixed a stupid memory corruption bug. lesson learned: never use copyMem with refs
20:39:12	Zevv	didn't yo mama tell ya
20:39:18	FromDiscord	<lqdev> just copy your values around in for loops like a good boy
20:39:31	Zevv	right
20:40:08	Zevv	is your code public?
20:40:22	FromDiscord	<lqdev> not yet, will need to do some testing before i can call it ready for publishing
20:40:37	Zevv	cool
20:40:42	FromDiscord	<lqdev> but it's almost complete
20:41:02	FromDiscord	<lqdev> my VM can compute factorials like a champ
20:41:16	Zevv	hooray \o/
20:47:10	*	wasted_youth quit (Read error: Connection reset by peer)
20:48:26	*	tane quit (Quit: Leaving)
20:48:55	*	jkiesian joined #nim
20:50:39	giaco	nim on quantum computer, anyone?
20:50:53	disruptek	btdt
20:50:59	FromDiscord	<Clyybber> maybe
20:51:05	disruptek	uncertain.
20:51:08	FromDiscord	<Daniel> quantum computer only works with cats
20:51:23	disruptek	disbot: are you awake?
20:51:24	disbot	very doubtful.
20:51:32	Zevv	will never work, will trigger all possible compiler bugs at the same time
20:52:07	FromDiscord	<Clyybber> but will also find all compiling snippets at the same time
20:52:18	FromDiscord	<Clyybber> disbot: are you asleep?
20:52:19	disbot	without a doubt.
20:52:29	FromDiscord	<Clyybber> disbot: are you awake?
20:52:30	disbot	my sources say no.
20:52:36	FromDiscord	<Daniel> disbot: are you sentient?
20:52:37	disbot	better not tell you now.
20:52:37	FromDiscord	<Clyybber> incredible
20:53:05	FromDiscord	<Clyybber> disruptek: How is life as a mechanical turk?
20:53:09	FromDiscord	<Daniel> distbot: are you drunk?
20:53:17	FromDiscord	<Daniel> (edit) "distbot:" => "disbot:"
20:53:18	disruptek	answer cloudy; ask later.
20:53:33	*	awici[m] joined #nim
20:53:34	giaco	disruptek: what's the answer to the Ultimate Question of Life, the Universe, and Everything
20:53:40	giaco	I mean disbot
20:53:41	disruptek	42 ofc
20:53:42	giaco	damn!
20:54:20	FromDiscord	<Daniel> regarding quantum computing, ....doesn't seem viable in near future, or maybe it will have to be reworked.
20:55:33	FromDiscord	<mratsim> https://en.wikipedia.org/wiki/Q_Sharp
20:55:34	FromDiscord	<Daniel> (edit) "reworked." => "reworked....probably wont have "quantum" in its name"
21:00:23	giaco	is it possible with dq_sqlite to extract a cell that contains a blob with null bytes? All I get is a truncated string. Tried with getRow and getValue
21:03:28	leorize	you don't use db_sqlite and instead use ndb
21:04:36	giaco	leorize: I'm actually using norm, that uses ndb, but there's a push request of 1 character waiting to be accepted for that https://github.com/xzfc/ndb.nim/pull/17
21:04:37	disbot	➥ Export "o" field of DbValue type.
21:08:01	*	narimiran quit (Ping timeout: 256 seconds)
21:08:42	*	hnOsmium0001 joined #nim
21:18:02	*	oculux joined #nim
21:19:57	*	jkiesian quit (Ping timeout: 272 seconds)
21:24:47	giaco	leorize: here's a small example that shows that ndb fails too https://play.nim-lang.org/#ix=2Muw
21:27:13	leorize	@mratsim: since you manage to figure out concepts, can you tell me how to fix this? https://github.com/alaviss/nim-sys/commit/bfd0e0bdbc63ac3e465ec4a3d75e2f074e49992d
21:27:48	leorize	you can see CI for the error messages
21:28:00	*	hmmm joined #nim
21:29:06	leorize	tconcept pass, even if it wasn't run on CI
21:29:30	FromDiscord	<mratsim> so the FD(fd) and T(fd) are conversion constraint?
21:30:37	*	hoijui quit (Ping timeout: 272 seconds)
21:30:41	leorize	yea, I just need them to be valid
21:34:11	*	jkiesian joined #nim
21:37:06	FromDiscord	<mratsim> I don't have time today, ping me tomorrow
21:37:13	leorize	sure, thanks
21:39:17	FromDiscord	<mratsim> it's very possible that concept only takes normal proc into account and type converter were missed though
21:40:12	disruptek	i doubt it; i believe concepts uses `compiles`, and this is part of the reason replacement is desired.
21:40:26	leorize	yea but tconcept, which contains `FD is AnyFD` and `SocketFD is AnyFD` tests, passes
21:42:43	disruptek	haxscramper: is your path code compatible with 1.0 and where do i find it?
21:48:30	*	NimBot joined #nim
21:53:07	FromDiscord	<haxscramper> https://haxscramper.github.io/hmisc/src/hmisc/other/oswrap.html
21:53:20	disruptek	rip
21:53:22	FromDiscord	<haxscramper> And I have no idea about 1.0 compatibility
21:53:39	disruptek	we bumped gram because hmisc doesn't support 1.2, so...
21:53:47	disruptek	seems unlikely that it will work for me.
21:54:12	FromDiscord	<lqdev> Zevv: done! https://github.com/liquidev/cflang
21:55:01	*	xet7 quit (Remote host closed the connection)
21:56:08	*	xet7 joined #nim
21:58:47	FromDiscord	<cji> sent a long message, see http://ix.io/2MuF
22:00:55	*	tane joined #nim
22:02:17	FromDiscord	<lqdev> that's the preferred way, yes
22:02:34	FromDiscord	<Clyybber> leorize: Can you check https://github.com/nim-lang/Nim/pull/16748/files since you use haiku yourself ?
22:02:35	disbot	➥ [backport 1.0.x] fix config/nim.cfg: `@if not bsd or haiku:` was buggy
22:02:56	FromDiscord	<lqdev> from what i heard i think {.inject.} may also work but don't quote me on that
22:03:39	FromDiscord	<Clyybber> Yep, inject will work too
22:04:02	leorize[m]	@Clyybber it's fine, he just turn multiple ifs into if-elseif chains
22:04:06	*	leorize quit (Ping timeout: 268 seconds)
22:05:18	FromDiscord	<Clyybber> ah I see. So the bug was the not bsd or haiku not being not(bsd or haiku)
22:08:15	FromDiscord	<cji> @lqdev thanks!
22:11:45	*	lritter joined #nim
22:19:03	FromDiscord	<Clyybber> disruptek: damnit, you broke my build just when I fixed it
22:19:11	disruptek	what?
22:21:45	FromDiscord	<Clyybber> bump
22:22:00	disruptek	broke it how?
22:22:25	disruptek	!pulls author:disruptek
22:22:28	disbot	https://github.com/nim-lang/Nim/pull/16517 -- 3disable disruptek, add frosty, testes (disabled)
22:22:28	disbot	https://github.com/nim-lang/Nim/pull/16390 -- 5add jason to important packages
22:22:28	disbot	https://github.com/nim-lang/Nim/pull/16323 -- 5incremental compilation for the frontend 7& 27 more...
22:22:49	disruptek	why don't you merge 16517 so i don't have to worry about you, chucklehead?
22:23:20	FromDiscord	<Clyybber> bump/tests/tbump.nim(5, 8) Error: cannot open file: bump
22:23:31	FromDiscord	<Clyybber> its the switch to balls I presume
22:23:35	disruptek	of course.
22:23:51	FromDiscord	<Clyybber> ah issue on our end I guess
22:23:54	disruptek	of course.
22:24:02	disruptek	and one i've already patched and PR'd.
22:24:40	FromDiscord	<Clyybber> Does it need to be installed for the test to work?
22:24:50	disruptek	what kind of a question is that?
22:25:11	FromDiscord	<Clyybber> Sry, I should look myself
22:25:17	FromDiscord	<Clyybber> mobile laziness
22:25:34	disruptek	balls has to be compiled before it can be run, yes.
22:25:59	FromDiscord	<Clyybber> ah its somehow a path issue
22:26:04	disruptek	indeed.
22:26:07	disruptek	#16447
22:26:08	disbot	https://github.com/nim-lang/Nim/issues/16447 -- 3testament fails for packages that require $NIMBLE_DIR/bin in PATH ; snippet at 12https://play.nim-lang.org/#ix=2LQ3
22:26:19	FromDiscord	<Clyybber> your tbump expects bump to be installed as a package
22:26:30	FromDiscord	<Clyybber> no its not the binary issue here
22:26:36	disruptek	yes, it is.
22:26:53	FromDiscord	<Clyybber> but the line where it fails is "import bump"
22:26:59	disruptek	so what?
22:27:02	FromDiscord	<Clyybber> not "execute bump"
22:27:05	disruptek	you're not running my test runner.
22:27:30	FromDiscord	<Clyybber> I'm just saying that its a different issue I think
22:27:45	disruptek	well, you can think and say what you want, but i've explained the issue.
22:28:01	disruptek	tbump does not run bump.
22:28:01	FromDiscord	<Clyybber> please explain why the import fails
22:28:09	disruptek	because there's no path to bump.nim.
22:28:14	disruptek	balls adds it.
22:28:17	Araq	the issue is that my IC refactoring break your code
22:28:20	disruptek	tbump.nim.cfg was removed recently.
22:28:37	FromDiscord	<Clyybber> Araq: No its in my PR
22:28:45	FromDiscord	<Clyybber> the issue can be fixed
22:28:47	Araq	and I haven't yet looked why, but I did notice that the code never worked correctly anyway
22:28:51	FromDiscord	<Clyybber> I did it for cligen once
22:29:21	FromDiscord	<Clyybber> Its just a --path:. in importantpackages
22:29:23	Araq	btw tests\exception\t13115.nim fails on my machine
22:29:56	*	leorize joined #nim
22:30:08	Araq	maybe we need an AraqsMachine special case for testament
22:30:27	Araq	as I couldn't care less if my exception messages handle binary zero properly
22:30:52	Araq	sure, let's have binary data in human readable error messages, what could go wrong
22:31:20	FromDiscord	<Clyybber> lol
22:31:22	Araq	BUG! My teeth hurt when I eat stone.
22:32:11	FromDiscord	<Clyybber> disruptek: aah thats the issue
22:32:17	disruptek	yes.
22:32:26	FromDiscord	<Clyybber> Araq: Are you testing on windows?
22:33:09	disruptek	this glass tastes like blood. fix it!
22:33:19	Araq	I'm always on Windows, don't want to debug my OS when I debug my compiler.
22:33:31	*	disruptek sighs.
22:33:40	disruptek	malloc(): unsorted double linked list corrupted
22:34:31	Araq	tomorrow I'll be on OSX though, Apple said I'll get an M1 tomorrow
22:34:48	FromDiscord	<Clyybber> yeah I heard you got one thats why I wondered
22:35:30	FromDiscord	<Clyybber> disruptek: Inb4 are you going to do the same for sigv4?
22:35:43	FromDiscord	<Clyybber> Since thats another pkg we test from you
22:35:48	disruptek	of course. it's important to me that my software works.
22:35:55	FromDiscord	<Clyybber> And I can update it in advance then
22:35:56	FromDiscord	<Clyybber> Ok
22:35:57	disruptek	it's not so important to me that your compiler passes tests.
22:36:03	FromDiscord	<Clyybber> sure
22:36:23	disruptek	especially when, as i said, i already submitted a PR to disable all my shit in important_packages.
22:36:29	disruptek	nothing more i can do.
22:36:52	FromDiscord	<Clyybber> yeah, I'm not asking you to do anything
22:37:16	Araq	so ... it occured to me that my IC related backend changes could simply always be done
22:37:31	FromDiscord	<Clyybber> The DCE?
22:37:33	Araq	yeah
22:37:37	FromDiscord	<Clyybber> cool
22:37:40	Araq	it's much more elegant this way...
22:38:25	Araq	btw apparently we generate 5000 destructors for system.nim
22:38:43	Araq	that doesn't feel right...
22:38:59	disruptek	i like it; i feel lonely with less than 3500.
22:39:38	FromDiscord	<Clyybber> Araq: I mean everything uses something from system.nim no?
22:39:49	FromDiscord	<Clyybber> Or is that really system.nim alone?
22:39:58	Araq	system.nim alone
22:40:03	FromDiscord	<Clyybber> huh
22:40:16	*	grobe0ba quit (Ping timeout: 240 seconds)
22:40:22	Araq	well we have lots of trivial destructors I guess
22:40:46	Araq	but we always do it, even with --gc:refc etc
22:41:42	FromDiscord	<Clyybber> we dont need to generate the trivial destructors even with gc arc no?
22:42:10	FromDiscord	<Clyybber> eh nevermind
22:42:39	Araq	we don't need to, but we do
22:42:42	Araq	I think.
22:43:25	Araq	but there is a more interesting problem: *when* do we attach a non-trivial/trivial destructor to a type?
22:43:57	Araq	the spec says "when the type T is used in context X"
22:45:07	Araq	but this violates modularity, doesn't it? what if type T is not used in the module at all but exported and then some later module uses T?
22:45:45	FromDiscord	<Clyybber> Then we do it at definition instead?
22:46:19	Araq	can't work, you cannot even put the =destroy inside an object declaration
22:47:00	FromDiscord	<Clyybber> I dont understand
22:47:14	Araq	we need a rule like: at the end of a module =destroy for T is generated
22:47:17	FromDiscord	<Clyybber> Do you mean attach in the context of the compiler
22:47:20	FromDiscord	<Clyybber> ah
22:48:10	FromDiscord	<Clyybber> Because the destroy gets put into the rodfile?
22:48:24	Araq	of course
22:50:51	FromDiscord	<Clyybber> Must be careful with recursive imports I think
22:52:12	*	grobe0ba joined #nim
22:52:19	FromDiscord	<Clyybber> Araq: Does DCE mean that system.nim.c will need to be recompiled if a sym of it becomes unused?
22:53:02	FromDiscord	<Clyybber> If so, it might make sense to have a switch to disable it, at the cost of binary bloat
22:53:09	Araq	yes and there is no alternative.
22:53:46	Araq	you can argue that when you use symbol X from M that was previously unused we can generate M2.nim.c that only contains X.
22:54:14	Araq	But since no such solution exists for removing a symbol, there is no need to implement that solution
22:54:53	Araq	and I suspect it's rare that you change what you use from system.nim
22:55:18	FromDiscord	<Clyybber> What about just not removing it. People dont care if their binary size grows if they just want to build and debug fast
22:55:54	FromDiscord	<Clyybber> We could warn then once the bloat reaches a critical point
22:57:25	Araq	because then you need to recompile "just in case" when you care for size
22:58:35	Araq	I usually design for "no switch required, not yet another thing you need to keep in mind"
22:58:56	FromDiscord	<Clyybber> Yeah fair. We could do it behind the scenes
22:59:14	FromDiscord	<Clyybber> and only for --debug not release or danger
22:59:39	Araq	also, the better solution is to simply slim down system.nim
23:00:10	Araq	assert and io at least should have been their own imports
23:01:34	*	tane quit (Quit: Leaving)
23:02:02	*	a_b_m quit (Quit: Leaving)
23:43:28	FromDiscord	<juliuskiesian> i hope i'm not the only one who's running into this, but the vscode extension actually renames stuff in the std lib when i try to rename a var in my project.
23:47:11	*	Vladar quit (Quit: Leaving)
23:47:41	*	idf quit (Ping timeout: 256 seconds)
23:47:42	*	idf_ joined #nim
23:48:36	*	junland quit (Ping timeout: 256 seconds)
23:48:51	*	junland joined #nim
23:50:03	FromDiscord	<juliuskiesian> my `lib/core/macros.nim` got messed up twice in the past two weeks.
23:50:37	leorize	which extension are you using?
23:50:40	disruptek	which vscode extension are you using?
23:50:45	disruptek	heh
23:50:49	disruptek	^ saem
23:51:16	FromDiscord	<juliuskiesian> kosz78.nim
23:51:34	saem	notit!
23:51:45	FromDiscord	<juliuskiesian> v0.6.6
23:51:46	saem	Guaranteed mine has the same bug, though
23:52:28	disruptek	my entire day was spent on bullshit.
23:52:46	saem	I mean you can give my extension a shot and see if it poops the bed or not.
23:53:09	FromDiscord	<juliuskiesian> sure, which one is it?
23:53:16	saem	disruptek: we're having the same day
23:53:39	FromDiscord	<juliuskiesian> nimsaem, i assume?
23:53:43	saem	julius: the good one, obviously
23:53:45	saem	LoL
23:53:47	saem	Yeah
23:53:57	FromDiscord	<juliuskiesian> do i need to remove the old one, then?
23:54:12	saem	Yeah, I think you can just disable it
23:54:31	saem	Might have to remove, I don't use it for obvious reasons
23:55:18	saem	Can't say for sure.
23:55:38	disruptek	someone told me that a fork is the opposite of collaboration.
23:55:53	disruptek	if that's the case, perhaps it explains all the cutlery here:
23:55:56	disruptek	!repo nimble
23:55:58	disbot	https://github.com/nim-lang/nimble -- 9nimble: 11Package manager for the Nim programming language. 15 789⭐ 133🍴 7& 29 more...
23:56:23	Araq	cutlery?
23:56:32	disruptek	the knife cuts both ways.
23:56:36	saem	Anyhow, if the bug happens there please raise an issue or better yet give a fix a whirl.
23:56:59	Araq	never heard of such a bug before
23:57:12	FromDiscord	<juliuskiesian> saem, it happens there, too.
23:57:16	saem	Sweet
23:57:19	saem	I guess
23:57:23	Araq	what I noticed is that VScode can trunc files when your HD is full
23:57:31	Araq	which kinda makes sense I guess
23:57:55	disruptek	yeah, i never would have guessed that this isn't what the user wants.
23:58:15	saem	Imma say that FS full issues aren't something to prioritize rn. 😄
23:58:21	disruptek	afaic, not using every bit of my hard-drive is a bug.
23:58:27	FromDiscord	<juliuskiesian> saem, both extensions seem to have something agains L550 in macros.nim.
23:58:48	FromDiscord	<juliuskiesian> they always find a way to mess up that line.
23:59:12	FromDiscord	<juliuskiesian> (edit) "agains" => "against"