alichay has quit [Remote host closed the connection]
protty has quit [Quit: Ping timeout (120 seconds)]
blackbeard420 has quit [Quit: ZNC 1.7.5 - https://znc.in]
blackbeard420 has joined #zig
nore has quit [Ping timeout: 240 seconds]
ave_ has quit [Quit: Connection closed for inactivity]
scientes has quit [Ping timeout: 240 seconds]
nore has joined #zig
redj has joined #zig
<daurnimator>
mq32: add completion support to your arg parser :)
<daurnimator>
mq32: i.e. have a method "pub fn writeCompletions(enum {bash, fish, zsh}, myoutstream) !void`
a_chou has joined #zig
a_chou has quit [Client Quit]
protty has joined #zig
<protty>
andrewrk: Realized the issue: the node's `prev` wasn't being reset after being notified so the queue ended up in an error state.
lunamn_ has quit [Quit: leaving]
lunamn has quit [Quit: leaving]
mahmudov has quit [Remote host closed the connection]
protty has quit [Remote host closed the connection]
alexnask has quit [Ping timeout: 268 seconds]
jjido has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
alexnask has joined #zig
alexnask has quit [Read error: Connection reset by peer]
alichay has joined #zig
nikoala has quit [Ping timeout: 265 seconds]
vegecode has joined #zig
<vegecode>
Can anyone tell me what the recommended way to do type punning is? I'm looking for something equivalent to teh following:
<vegecode>
typedef union
<vegecode>
crap how do i do small code snippets
<vegecode>
I'm trying to do the equivalent of a union in c where one union member is a struct of bitfields like for flags and then the other member is an integer that is the width of all teh flags combined
<vegecode>
Common C idiom for representing lfags
<vegecode>
I think I can make it be a union and then just change the active member by doing a write to the other member with undefined. Is there a better way? Do I just have to shift and mask everything?
<vegecode>
maybe bit casting between a packed struct and the full width type?
<fengb>
packed struct for the bits. Depending on how often you use the full type, a packed union or bitcasting
<mikdusan>
but zig union is one-field active at a time rather than a "bitcast" exchange hub :)
<fengb>
Not packed union
<vegecode>
does a packed union skip the safety check of which field is active then?
<mikdusan>
fengb: ah nice
<fengb>
Yes
<vegecode>
skip it intentionally I mean
<vegecode>
thank you sir!
<vegecode>
One more thing, before I pollute the issue tracker, is it reasonable to request some movement on a proposal? Either accpet or reject. Specifically #4284
<mikdusan>
lookin at #4284 I think you already have an answer: on Jan 25 it was added to 0.7.0 milestone
<vegecode>
That was the day it was opened and the "accepted" label was never added
<vegecode>
I sure hope it's not accepted. I'm quite opposed.
<andrewrk>
Milestones on non accepted proposals mean it's not planned to make a decision until that release cycle
<vegecode>
perfect, thanks
ur5us has quit [Ping timeout: 240 seconds]
ur5us has joined #zig
<daurnimator>
andrewrk: can has merges? :)
<fengb>
`self.code % 16 + '0'` lol whoops I fail at hexamath
<mikdusan>
so 0xffff = 0x???? and 0xaaaa = 0x:::: ... I'm ok with that
waleee-cl has quit [Quit: Connection closed for inactivity]
<fengb>
`i32.const 420` => `.I32 = -15964`
<fengb>
I don't think this is working quite right
ur5us has quit [Ping timeout: 240 seconds]
<fengb>
`else if (0x40 != 0` I think I forgot a variable
<fengb>
I am so smrt
<BaroqueLarouche>
now I imagine fengb's house in fire
<daurnimator>
forgetting a variable: not even once.
<fengb>
Well, I'm glad I didn't build my own house
dddddd has quit [Remote host closed the connection]
vegecode has quit [Remote host closed the connection]
slowtyper has joined #zig
slowtyper has quit [Quit: WeeChat 2.7.1]
rom1504 has quit [*.net *.split]
gchristensen has quit [*.net *.split]
M-ou-se has quit [*.net *.split]
mht has quit [*.net *.split]
shachaf has quit [*.net *.split]
xvilka has quit [*.net *.split]
pmwhite has quit [*.net *.split]
AlexMax has quit [*.net *.split]
dtz has quit [*.net *.split]
jicksaw has quit [*.net *.split]
halbeno has quit [*.net *.split]
s-ol has quit [*.net *.split]
Cloudef has quit [*.net *.split]
euantor has quit [*.net *.split]
wjlroe has quit [*.net *.split]
shodan45 has quit [*.net *.split]
Prf_Jakob has quit [*.net *.split]
jrl has quit [*.net *.split]
diginet has quit [*.net *.split]
cow-orker has quit [*.net *.split]
commander has quit [*.net *.split]
nickster has quit [*.net *.split]
andrewrk has quit [*.net *.split]
wtw has quit [*.net *.split]
notjones has quit [*.net *.split]
Sahnvour has quit [*.net *.split]
benaiah has quit [*.net *.split]
johnLate has quit [*.net *.split]
Ekho has quit [*.net *.split]
idxu has quit [*.net *.split]
hspak has quit [*.net *.split]
niftynei has quit [*.net *.split]
lukeholder has quit [*.net *.split]
edr has quit [*.net *.split]
dom96 has quit [*.net *.split]
cota has quit [*.net *.split]
fengb has quit [*.net *.split]
ianc6209 has quit [*.net *.split]
companion_cube has quit [*.net *.split]
guan has quit [*.net *.split]
grayhatter has quit [*.net *.split]
_whitelogger has joined #zig
return0e has joined #zig
CommunistWolf has joined #zig
blackbeard420 has joined #zig
doublex_ has joined #zig
BaroqueLarouche has joined #zig
lqd has joined #zig
Snektron has joined #zig
tralamazza has joined #zig
APic has joined #zig
jmiven has joined #zig
Cadey has joined #zig
mixi has joined #zig
torque has joined #zig
mforney has joined #zig
APic has quit [Max SendQ exceeded]
APic has joined #zig
alexnask has joined #zig
alichay has quit [Quit: Konversation terminated!]
marmotini_ has joined #zig
marmotini_ has quit [Ping timeout: 255 seconds]
<alexnask>
andrewrk, Is it expected behavior that the calling convention and type of inferred async functions is .Unspecified and non-async fn ptr?
<mq32>
<daurnimator> mq32: add completion support to your arg parser :)
<mq32>
never used something like that, can you elaborate?
<daurnimator>
mq32: do you have luarocks installed?
<mq32>
maybe? D
<mq32>
nope
<daurnimator>
mq32: kubectl?
<mq32>
negative
<daurnimator>
uh... okay. so many commands have the ability to output bash/fish/ksh completions for themselves
<daurnimator>
which you tell your shell about, so you can type (e.g.): luarocks in<TAB> and it completes to `install`
<mq32>
hm, sounds useful
<daurnimator>
==> same for all arguments of the program `luarocks --c<TAB>` => `luarocks --config`
<daurnimator>
depending on your shell configuration, people even set it up to have pretty menus and things
<mq32>
can you point me to some docs for this?
* mq32
uses only bare-bones completion of bash
<daurnimator>
mq32: I can help most with luarocks's version of it.... cause adding that was the GSoC project I mentored last year :)
<dimenus>
i fully deleted my build folder (i'm not doing a system wide install)
<dimenus>
it's quite possible, in my wayland example i'm only linking libc and i got that error
<andrewrk>
I bet it is 2044
<dimenus>
you're so quick with the issues btw :)
<andrewrk>
it helps when you have them all memorized
<andrewrk>
zig is small enough that it fits in my head
<dimenus>
but yeah, clean build and it works as expected
<andrewrk>
glibc goes nuts with same-named .h files and relying on include order
<andrewrk>
I was looking at porting cache_hash.cpp into zig last night (coincidentally with LeRoyce's PR) and I believe it can be done in O(1) memory, allocator needed
<dimenus>
nice
<fengb>
Debugging raw bytecode is so miserable
<andrewrk>
work on features to make debugging raw bytecode not miserable
<fengb>
Yeah soon™
<andrewrk>
it will pay off big time
<fengb>
Gonna need to revive my sexpr parser so I don't have to write bytecode all the time
<alexnask>
@andrewrk, Would you be ok with a SCOPE_DEFER {...} macro (that creates an object that runs a lambda on destruction) in utils.hpp?
<alexnask>
I want to start freeing a bunch of temporaries
<andrewrk>
alexnask, nah no macros
<alexnask>
Ok
<andrewrk>
sorry
<andrewrk>
I've been thinking, a long term stretch goal of stage1, once we are shipping self-hosted, would be to reduce it to C (not C++) and have it generate C code rather than using llvm
<andrewrk>
then we would be able to bootstrap the (vaporware) self-hosted non-llvm-backend compiler from only a C compiler
dimenus has quit [Remote host closed the connection]
<andrewrk>
which is not a pointless endeavor - it would make zig viable to use as a compiler in a fledgling OS or new CPU architecture
<alexnask>
Right (I suppose it would generate C89 too then :P)
<alexnask>
For maximum portability
<andrewrk>
yeah, you get the idea
<metaleap>
from the outside, i'm curious about your eagerness to be potentially/partially/incrementally llvm-independent. does ll not live up to the hype, does it incur friction? too buggy? or is it just general desire for fewer dependencies in-the-very-long-run-future?
<mikdusan>
instead of [stage1 -> C -> stage2] why not [stage2 -> C -> stage2]
<companion_cube>
a language striving to be simple and standard shouldn't depend on a beast like llvm (or rather it shouldn't be a single point of failure?)
<mikdusan>
as in stage2 generates C code
<andrewrk>
mikdusan, because the generated C isn't "source" code and cannot be audited as such
<andrewrk>
it's tempting though. idea would be to do it once, and then sort of "de-obfuscate" it and "sourceify" it
<andrewrk>
sort of what go did when they translated from C, except in reverse
<fengb>
LLVM is an implementation detail, not a spec. We use it for a bunch of really nice features, but it's also has its own quirks and challenges
<andrewrk>
mikdusan, also we'd need to build N versions (one for each target) and then merge them with #ifdefs
<companion_cube>
fengb: well it's kind of influencing the specs of what's an intrinsic, isn't it?
<fengb>
In the near-term, we'd like faster debug builds, which LLVM is somewhat of a bottleneck. Odin and Jai have the same issues
<companion_cube>
with a C backend, how do you provide portable simd? :/
<andrewrk>
companion_cube, same way you emit portable simd when the target cpu does not have simd features
<companion_cube>
by emulating them? fair enough :)
<fengb>
companion_cube: sort of. Intrinsics are what I'd consider "some CPUs have instructions, but we can provide a userland implementation if necessary"
<companion_cube>
ok, I guess that's fair
<companion_cube>
as long as the semantics is the same
<companion_cube>
(same as `popcount` and the likes, actually)
<mikdusan>
if we [stage2->C] to bootstrap onto a platform, that is done just once, and the codegen benefits from all zig userland smarts to see what the target is, etc. and one'offs some C. then we build a (stage2)0 with target host buildtools. and bob's your uncle.
<fengb>
mikdusan: are you british?
<mikdusan>
fengb: nah I just like that phrase
<andrewrk>
metaleap, a lot of the communities that zig wants to reach (people who are currently using C) feel that LLVM is too much of a monoculture
<andrewrk>
...which is true. we need a language spec that is independent of LLVM. and a non-llvm-backend is proof that it is viable
<metaleap>
ah well what i heard in terms of criticism so far is that its a c-focused monoculture :D
<companion_cube>
also, yeah, but that matters more if you either want to use optimizations that are infeasible in C, or if you want GC support :)
<metaleap>
not a bad aspiration for sure, was just curious if there were "burning" reasons or merely "lets become proper-er over time" reasoning
<fengb>
Does Rust have a non-LLVM compiler?
<companion_cube>
nope
* BaroqueLarouche
hopes to find/have time to write non-llvm backend for retro consoles when self-hosted compiler is shippable
<companion_cube>
and it's imho a shortcoming of rust
<TheLemonMan>
they're working on a cranelift-based backend
<fengb>
metaleap: the goal was always there. LLVM merely enables us to ship "production-grade" code earlier
<companion_cube>
what I'm wondering is, wouldn't a wasm (direct) backend be a viable first-class backend?
<andrewrk>
mikdusan, I think that's something worth considering once we are shipping self-hosted and have a C backend
<companion_cube>
no one said languages should only target native architectures
<fengb>
We've also inherited some LLVM baggage... like old coroutines...
<andrewrk>
the goal of bootstrapping is to get from hand-typing hex code directly into an executable file, to a fully featured self-hosted compiler
<andrewrk>
wasm introduces unnecessary steps in that sequence
<fengb>
what about CPUs that natively execute wasm? 🙃
<fengb>
Oh speaking of which, I managed to get my wasm vm execute 1 function with 1 instruction! No longer complete vaporware
<andrewrk>
it took 9.5 hours to run the full zig test suite with debug build of llvm
<mikdusan>
passed?
<andrewrk>
yes
<mikdusan>
you need more ghz
<andrewrk>
need more utilizing cpu cores :)
<fengb>
How important is Zig to LLVM? :P
<mikdusan>
andrewrk: what does your gut say the main focus for 0.6.0 -> 0.7.0 will be?
<andrewrk>
stabilizing the language, specification, and self-hosted compiler
<TheLemonMan>
andrewrk, the source instruction looks fine https://clbin.com/AfBBi , I don't remember if the GenConst with (runtime value) is ok at IR level
<andrewrk>
I think a GenConst with runtime value is suspect
o_O has joined #zig
<andrewrk>
TheLemonMan, btw did I ever show you how to use the IR debugging features? size_t break_debug_id; // for debugging purposes
<andrewrk>
damn, 0.5.0 release notes says "it's time to begin on networking in the Standard Library. I expect to complete this along with at least an early prototype of the package manager during the next release cycle."
<fengb>
You have a month to catch up :P
<andrewrk>
we've completed so many things this cycle, and yet reading this old roadmap makes me feel like no progress has been made
<TheLemonMan>
alexnask, awesome work on #4643!
<alexnask>
Thanks
<alexnask>
Tackling #3237 now
<TheLemonMan>
seeing all those red crosses on the PR page makes me really sad
<andrewrk>
TheLemonMan, I'm slightly annoyed at source hut for not using a more stable freebsd version
<andrewrk>
the download page is still getting updated when everything else passes
<mikdusan>
would be nice if they added a 12.1 image
Akuli has joined #zig
<andrewrk>
if this continues for another day I'll disable freebsd builds entirely
<fengb>
Just highlight all the breaking changes
<TheLemonMan>
isn't 12 the latest stable release?
marmotini_ has joined #zig
<mikdusan>
i'm not sure exactly but their 12.1 image somehow fails with pkg commands, and yet "uname -r" matches my vm, and my pkg commands seem to work
<TheLemonMan>
maybe your pkg is new enough to include the patch that fixed the problem
<andrewrk>
IMO it's source hut's responsibility to choose a working version of freebsd
<andrewrk>
and it's zig's responsibility to choose a working CI
<TheLemonMan>
a lot of CI services are affected by that problem
<TheLemonMan>
shit happens™
<andrewrk>
yeah I'll give it a couple more days
marmotini_ has quit [Read error: Connection reset by peer]
<Guest20151>
Assertion failed at /deps/zig/src/list.hpp:31 in at. This is a bug in the Zig compiler.
<Guest20151>
can we make it that this does not happen when simply moving a number of defs to new files and renaming them and all references.... too often when i want to refactor things a bit and clean up, becomes a battlefield
<TheLemonMan>
good job, you broke it
<andrewrk>
Guest20151, yes, we can fix all the bugs
<andrewrk>
it will just take time
dimenus has joined #zig
<Guest20151>
here's a hint if someone wants to glimpse at your list.hpp:31. i had a couple .zig files all `pub usingnamespace @import`ing a single .zig file (which exists) that itself `pub usingnamespace @import`ed one which.. didnt exist (renamed) :D
<Guest20151>
maybe can do a repro later
<TheLemonMan>
without a repro there's little or no chances for this bug to be fixed
<Guest20151>
somehow hexchat irc lost my nick, brb. yeah i put a repro in my todo.txt, cant lose flow now =)
Guest20151 has quit [Quit: Leaving]
metaleap has joined #zig
<TheLemonMan>
bleh, a lot of std.target.VersionRange stuff ends up in freestanding binaries
jjido has joined #zig
<andrewrk>
in release builds? that should not be the case
<andrewrk>
in debug builds, functions called at compile time get runtime versions generated as well. I believe there is an open issue for this. it's tricky because, maybe you wanted to call that function in a debugger?
<TheLemonMan>
yeah, this is related to the missing way to mark some functions as comptime-only
<TheLemonMan>
maybe with a custom calling convention enum?
<andrewrk>
that's an important issue, not sure how to solve yet
<andrewrk>
that's a nice idea
<andrewrk>
that might even solve #425 as well
<mikdusan>
oooh does that mean if a fn is marked comptime-only, I no longer need to `comptime callme()` ?
<TheLemonMan>
that's the idea, I sprinkle `comptime` everywhere because I'm never sure
<andrewrk>
mikdusan, more like `wants-to-be-comptime`
<andrewrk>
if all the parameters are comptime, then it comptime-calls the function
<andrewrk>
you would almost certainly want this for all pure functions
<mikdusan>
^^ with master or with this new feature in mind?
<andrewrk>
with this new feature in mind
<andrewrk>
I think I'm sidetracking the conversation though, because the thing I'm talking about is not really a calling convention, whereas TheLemonMan's idea is
<andrewrk>
if something *must* be called at comptime, that is a calling convention. it cannot be called at runtime with a different calling convention
<andrewrk>
however all this os version range stuff would not be comptime only
<andrewrk>
and even with "wants-to-be-comptime", it would still leave the issue of generating unneeded runtime functions in debug builds
<andrewrk>
so there are a few different but related things here
<andrewrk>
I haven't thought of an elegant solution yet
_Vi has quit [Ping timeout: 272 seconds]
jjido has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
mixi_ has joined #zig
mahmudov has joined #zig
dimenus has quit [Read error: Connection reset by peer]
hugbubby has joined #zig
<hugbubby>
I'm getting the error "container 'std.heap' has no member called 'page_allocator'
<hugbubby>
Trying to use the code suggested in teh Choosing your Allocator documentation. Does page_allocator not actually exist?
<hugbubby>
the line is: var arena = std.heap.ArenaAllocator.init(std.heap.page_allocator);
<metaleap>
using a real old version per chance hugbubby?
<andrewrk>
they've really done a lot of cleaning and consolidation
<alexnask>
nice
<andrewrk>
that's for arm, aarch64, i386, and x86_64 btw
<hugbubby>
How do I unsafely access freed heap memory w/ zig uwu
<hugbubby>
promise that's what I want to do
<andrewrk>
mingw-w64 is the only libc zig ships where we don't have to do a bunch of chores to make it multi-arch
<TheLemonMan>
does it still compile for i386? I remember having to patch some header to make it compile with clang
<TheLemonMan>
hugbubby, use a pointer?
<hugbubby>
It's segfaulting >.<
<andrewrk>
don't take this the wrong way, I'm just excited to use this idiom...
<andrewrk>
play stupid games - win stupid prizes
<alexnask>
:D
<TheLemonMan>
show some code
<TheLemonMan>
maybe you're getting a SIGBUS after a page is unmapped
<andrewrk>
TheLemonMan, do you remember any words from the commit message?
<andrewrk>
or files that it touched
<andrewrk>
found it: f83bb3dd9e68899ca39cdddb7d84c2ad1c833d55
dimenus has joined #zig
<dimenus>
andrewrk: I am not able to replicate joachimschmidt's issue with tetris
<dimenus>
it compiles quickly and without issue
<andrewrk>
dimenus, I think it depends on system c .h files
<andrewrk>
also zig caches @cImport
marijnfs has joined #zig
<wilsonk>
anybody else getting an error when trying to build the clashproto game on linux? The cross compiled version for windows-gnu builds and runs fine, but I get a 'c import failure' for soundio/soundio.h and an error about 'container "c" has no member SDL_SetHintPriority' also?
ur5us has joined #zig
<andrewrk>
wilsonk, make sure your zig version is updated and you have all the dependencies installed on your system
<andrewrk>
this is exactly the thing a package manager will solve fwiw
<fengb>
Which will have a prototype ready by next release ;)
<andrewrk>
the windows build is working because those dependencies are using packages
<wilsonk>
oh, I thought it was just 'supposed to work' out of the box the way it was written
<andrewrk>
yeah, on windows
<andrewrk>
it doesn't have sdl2 and libsoundio packaged for other OS's yet
<hugbubby>
http://dpaste.com/15GJ1B2 <-- this is the code taht triggers segfault instead of just going for dat sweet unsafe data
marijnfs_ has joined #zig
marijnfs has quit [Ping timeout: 258 seconds]
<wilsonk>
ok, so the second error goes away once I installed the dev version of libsoundio. Works fine now. Thanks
<hugbubby>
Is there a way I can free w/out setting space to 0
<hugbubby>
*undefined
<hugbubby>
Or is this a behavior of all allocators
<fengb>
Um... that'd be a use-after-free bug. What's the usecase?
<andrewrk>
ArenaAllocator doesn't yet have a way to "reset", while keeping all the memory reserved. that would be a nice API addition
<fengb>
Actually in this case I think it's probably page_allocator punting the page back to the OS
<mikdusan>
I think maybe you'll have to make your own allocator that does not markup after free
<andrewrk>
but it would certainly @memset everything to undefined
<hugbubby>
Atm the use case is I'm deliberately attempting to do that bug ;p, but also I would think memsetting upon free is a little costly speed wise
<hugbubby>
Isn't it?
<andrewrk>
yes it is. debug mode pays a lot of cost to gain easier debugging
<fengb>
I'd expect garbage instead of a hard crash
<mikdusan>
fengb: but I should point out that's only for the free/destroy's ... deinit is a different story.
<hugbubby>
So.... Does --release small @memset or not. I'm still in the dark here
<andrewrk>
hugbubby, if you're using std.mem.Allocator interface, free() and destroy() are defined to make the memory contents undefined and whether or not memory access will trap is allocator-implementation-defined
<metaleap>
usingnamespace must always have an operand resolving to a struct right? or any container? (all containers can have decls after all)
<andrewrk>
so your use case is deliberately thwarted by the Allocator interface
<andrewrk>
your use case requires resorting to non standard hacks
<metaleap>
(asking for my ast traversals driving the wip LSP prototype =)
<mikdusan>
thwarted is one of those words if you read more than once, it stops looking like a real word
<andrewrk>
metaleap, any container
<metaleap>
andrewrk thx
<hugbubby>
memsetting freed memory seems like its a little handholdey to me tbh
<TheLemonMan>
handholdey? people want even more safety
<mikdusan>
same can be said for array bounds checking
<fengb>
We might as well use a GC at this point
<hugbubby>
^
<hugbubby>
should be the allocator's choice
<fengb>
Real programmers write to empty bus
<mikdusan>
hugbubby: iirc there was talk of moving the responsibility of @memset(undefined) to allocator implementation
<mikdusan>
see that? i just mention it and boom. autolinked!
<andrewrk>
that was a re-paste too, I'm so good I linked it before you even said it :P
<alexnask>
:D
<mikdusan>
ahh i missed the first one oh my
<andrewrk>
my fault for posting a URL with no summary
<hugbubby>
I just figure ppl who are going to make their own allocator are also the high frequency trading types who dislike array bounds checking and memsetting unused memory to 0
<hugbubby>
so they can get that 10ns performance boost
<TheLemonMan>
I wrote my own allocator and I do like safety features
<mikdusan>
hugbubby: the good news is that an allocator could condition on build-mode for reset policy
<hugbubby>
^ see that'd be nice
<mikdusan>
but we're at version 0.5.0 still
<mikdusan>
patche welcome
<mikdusan>
*s
<hugbubby>
What do you guys plan to use zig for?
<BaroqueLarouche>
Games mostly
<andrewrk>
BaroqueLarouche has a tutorial repository on making game boy advance games in zig
<mq32>
hugbubby: anything lowlevel
<companion_cube>
sat solvers!
marijnfs has joined #zig
marijnfs_ has quit [Ping timeout: 268 seconds]
<andrewrk>
I'm going to rewrite my node.js music player server in zig, make a digital audio workstation, an arcade game, experiment with evolving digital organisms, build my static websites, and all the rest of the open source projects in my entire life
<hugbubby>
Maybe when I get rich I'll spend my free time rewriting bloated slow linux ecosystem components in zig idk
<andrewrk>
what did I do to deserve a tabs v spaces discussion happening today
<shakesoda>
hah
<BaroqueLarouche>
haha
<TheLemonMan>
tabs are 4 spaces, therefore tabs > spaces
<mikdusan>
I have no idea why you guys are talking about an old softdrink
<shakesoda>
my tabs may be 3 spaces, but the logic still checks out
<marijnfs>
shit, thats pretty good
<fengb>
I had a friend argue that tabs are 3 spaces. He is no longer a friend
<shakesoda>
uh oh
<hugbubby>
the garbage collectors want you to use terminal tabs
<companion_cube>
I like my tabs to be $(random 3 5) spaces
<TheLemonMan>
tabs are &ts spaces
<hugbubby>
they also want you to use many instances of the same terminal like a barbarian instead of one tmux term
<fengb>
Has anyone used fibonacci tabstops? :P
<companion_cube>
wait, do you only have one terminal?
<hugbubby>
I have one terminal. Every time I open a terminal, it either creates a tmux session or attaches to the existing one. I create new windows/splits when I need to do new things, like I do with a browser
<hugbubby>
It's the way god intended
<shakesoda>
i just make more terminal windows if i need them
<shakesoda>
they even get allowed to do barbaric things like overlap
<companion_cube>
hugbubby: ah but you still have multiple terminal windows? 😌
<fengb>
I don't actually run executables. They're all applets running in my browser
<hugbubby>
I do not. I have the one, 99% of the time. Its basically useless
<hugbubby>
The only time I use the second behavior is when I go to a different tty with alt-fn-x
<TheLemonMan>
everyone knows productivity scales linearly with the amount of terminals open
<hugbubby>
thx for the lang, I like making excuses to code in low level languages and you have enabled me
<alexnask>
Not sure how we get to that realloc without getting to realloc with old_count=0 and new_count=2 though Im looking into it
<alexnask>
new_count=1*
<mikdusan>
somebody is hammering the merge button
<alexnask>
Also not sure how it passed the tests on my machine and on the CI servers
<alexnask>
:p
<andrewrk>
alexnask, building stage2 isn't covered by tests currently, it's bit rotted a bit
<andrewrk>
I was just about to go in there and freshen things up
<andrewrk>
hugbubby, glad to hear it
<mikdusan>
alexnask: please tell me that's not how your fonts looks and is an artifact of image uploading
<TheLemonMan>
antialiasing is for the weak
<mikdusan>
ah yeah I got the same crash with `zig build --help`
<tdeo>
should the std.os wrapper for sendmsg set the MSG_DONTWAIT flag on the header with an event loop instance? it currently doesn't, so it seems like non-blocking io won't work with it automatically
<tdeo>
oh never mind that :( i think this wrapper was added my me in my local branch a few weeks ago and i forgot about it
<tdeo>
(but should it? i want to pr this when i'm done)
marijnfs_ has joined #zig
marijnfs has quit [Ping timeout: 240 seconds]
<mikdusan>
alexnask: so on your crash: try reversing 371c21aa70fc61ef703e34079ce6de6c52ec91df
hugbubby has quit [Ping timeout: 268 seconds]
<alexnask>
Yeah the thought crossed my mind although I think the issue is that we are only allocating src_fiel_count on the struct_val and not src_field_count + gen_field_count
<alexnask>
So I will try that first
<alexnask>
Or something like that
<TheLemonMan>
hm, gen_field_count should always be <= src_field_count
<mikdusan>
ok I regressed something.
marijnfs has joined #zig
marijnfs_ has quit [Ping timeout: 260 seconds]
<TheLemonMan>
alexnask, I'm still seeing #3878 in #4619 :(
<TheLemonMan>
I've applied your patch to the tuple_mult code path but the deref fetches an undefined value
<alexnask>
Weird, my tests use values from a tuple passed as a comptime var parameter
<TheLemonMan>
the problem is still there if you try to cast the result of an addition between two tuples to an array :\
<TheLemonMan>
it's like the resulting fields are never written back to the result_loc
jjido has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<alexnask>
mikdusan, I think my commit is the problematic one, testing a couple of ideas and will let you know asap I have something
<mikdusan>
alexnask: no probs. I'm putting some hooks on that container usage to see if I can prove/disprove safe-to-put-into-arena
marijnfs_ has joined #zig
marijnfs has quit [Ping timeout: 265 seconds]
marijnfs_ has quit [Read error: Connection reset by peer]
<andrewrk>
alexnask, I think checking the AST node is a smell
Akuli has quit [Quit: Leaving]
<andrewrk>
gonna revert that for now, sorry
<alexnask>
Ofc
<TheLemonMan>
andrewrk, if I have an "empty" result_loc with a struct type that's initialized as undefined, now if I fill it in field-by-field do I have to somehow reset its undefined state first?
<andrewrk>
by empty do you mean ResultLocIdNone?
<TheLemonMan>
yep
<andrewrk>
ResultLocIdNone means there is no parent result loc, and if there needs to be a result location, then it will be created as a temporary (as a stack variable). For examle, the result location of a function call parameter
<andrewrk>
I'm not sure I follow- in zig semantics, an undefined struct means all the fields are undefined. so if you define all the fields then the struct is no longer undefined
<TheLemonMan>
yeah, that's what I expected, but even though I'm filling in this structure all I get back is `undefined`
<andrewrk>
in the compiler guts, you would need to set the ZigValue::special to not be ConstValSpecialUndef anymore on the struct value
<andrewrk>
but I would expect that to happen when the fields are allocated
<andrewrk>
hmm this is making me realize that my plan for how to safety detect undef of structs at runtime is flawed
<TheLemonMan>
there's something that's not setting that flag then :\ no idea where the problem is tho
<andrewrk>
it should be when x_struct.fields is allocated
<andrewrk>
similar to expand_undef_array - the value itself no longer becomes undefined, and each field becomes undefined instead
<mikdusan>
1 liner: nuke their file which overrides "quarterly" to "latest"
doublex has joined #zig
doublex_ has joined #zig
recombinant has joined #zig
doublex_ has quit [Remote host closed the connection]
<andrewrk>
👏 👏 👏
Sphax has left #zig ["WeeChat 2.4"]
<andrewrk>
now to find out if we've broken freebsd in the last 3 days
<mikdusan>
I'll check my local vm
<TheLemonMan>
I'm logging out for today, if you have any news about the result_loc thingie please write it down in the PR
<mikdusan>
confirmed: `install` fails same error on my freebsd vm
<TheLemonMan>
peace
TheLemonMan has quit [Quit: "It's now safe to turn off your computer."]
<andrewrk>
peace
<andrewrk>
mikdusan, when in doubt, don't put stuff in the arena. basically anything that needs the concept of "reallocation" should not go in the arena
marijnfs has joined #zig
<marijnfs>
for (var i = 0; i < 10; i += 1) {} gives me: Expected expression, found var
<marijnfs>
howcome
<alexnask>
This is not zig code :-)
<alexnask>
you want `var i = 0; while (i < 10) : (i += 1) { }`
<marijnfs>
ah k
<marijnfs>
i was grepping zig code, but now see it was javascript:)
<fengb>
Proof that zig is inferior to JS? 🙃
<alexnask>
:D
recombinant has quit [Quit: Leaving]
Sphax has joined #zig
<marijnfs>
If i dont define var i : i64 = 0 i get issues
<marijnfs>
just var i = 0; gets me error: variable of type 'comptime_int' must be const or comptime
<marijnfs>
:O
Sphax has quit [Client Quit]
<alexnask>
a right mb
Sphax has joined #zig
<alexnask>
literals are comptime only types
<alexnask>
integer literals are comptime_int which is a comptime only type **
marijnfs has quit [Ping timeout: 240 seconds]
marijnfs has joined #zig
<marijnfs>
what is the comptime literal, the '0'?
<alexnask>
Yes, 0 is an integer literal :)
marijnfs_ has joined #zig
marijnfs has quit [Ping timeout: 255 seconds]
metaleap has quit [Quit: Leaving]
<marijnfs_>
can't comptime be assigned to a runtime var?
<marijnfs_>
other way around I understand is not possible, but this seems strange
<alexnask>
marijnfs_, comptime_int, type, comptime_float are comptime-only types
<alexnask>
But comptime_int and comptime_float can be casted to regular types (they coerce to integer/fp types)
<alexnask>
marijnfs_, If you explicitly make a comptime u64 (for example) you can implicitly use it in a runtime context
<mikdusan>
andrewrk: weird. I needed to do the same thing 1141bfb21b82f8d3fc353e968a591f2ad9aaa571 to FreeBSD