#jsmess 2021-12-29,Wed

↑back Search

Time Nickname Message
03:37 🔗 DFJustin 0.239 has dropped
06:06 🔗 SketchCo1 Hey there.
06:06 🔗 SketchCo1 So, I started the work on it.
06:06 🔗 SketchCo1 I started the compiles, etc.
06:06 🔗 SketchCo1 And of the first two, both break
06:06 🔗 SketchCo1 DFJustin: If you have any thoughts
06:24 🔗 SketchCo1 Ah I see
06:30 🔗 SketchCo1 I see what you said up there
06:30 🔗 SketchCo1 We are going to have major issues implementing this when we're not on the same time zones at al.
06:41 🔗 SketchCo1 But, anyway
06:41 🔗 SketchCo1 I have a2600 going.
06:41 🔗 SketchCo1 It blows up thusly:
06:42 🔗 SketchCo1 Starting Atari 2600 (NTSC) ':'
06:42 🔗 SketchCo1 loader.js?v=51c75c55:1053 exception thrown: 11235952
06:42 🔗 SketchCo1 atari_2600_pitfall_1983_cce_c-813:1 Uncaught (in promise) 11235952
06:43 🔗 SketchCo1 But otherwise, it's booting 0.239 right
06:45 🔗 SketchCo1 I simply renamed a2600.js to mamea2600.js, renamed a2600.wasm to mamea2600.wasm, gzipped, uploaded, it works.
06:45 🔗 SketchCo1 If that is causing something to explode let me know
07:09 🔗 DFJustin you didn't update emscripten
07:11 🔗 DFJustin the minimum requirement is now 2.0.25 but I would recommend the latest 3.0.1
07:11 🔗 DFJustin sorry I wasn't as explicit about that as I thought
07:16 🔗 DFJustin versions over 2.0.7 will fail to install on fos due to outdated python so you will need to use teamarchive1 or another box
07:19 🔗 SketchCo1 OK.
07:20 🔗 SketchCo1 I can just dump emscripten in, right, it'll go over the thing
07:20 🔗 DFJustin https://emscripten.org/docs/getting_started/downloads.html#updating-the-sdk
07:21 🔗 DFJustin you will need to do a clean rebuild of mame after the update, easiest way is to rm -rf mame/build
07:22 🔗 DFJustin I'm not sure what you mean by that question
07:22 🔗 SketchCo1 Emscripten answered it
07:22 🔗 SketchCo1 In teamarchive1 I'm extremely clean on the setup
07:23 🔗 SketchCo1 So it's all in /1/EMSCRIPTEN
07:23 🔗 SketchCo1 I went in there and followed update instructions, it told me to do a git pull and then so on
07:23 🔗 SketchCo1 rm -rf mame/build done
07:24 🔗 DFJustin after rebuilding now it might just work, there were two other problems I thought might occur but the a2600 is getting past both of them
07:24 🔗 SketchCo1 Emscripten is now at 3.1.0
07:24 🔗 SketchCo1 I'm running the thing now, hopefully it juuuuust works
07:25 🔗 DFJustin actually I take that back the buffer undefined error could still happen
07:25 🔗 SketchCo1 https://twitter.com/textfiles/status/1476078725531897856 this caused the usual joy
07:26 🔗 SketchCo1 Suppose I should double-check the compile settings in my MASTERBUILDER script
07:27 🔗 SketchCo1 emmake make -j4 SUBTARGET=$driver SOURCES=src/mame/drivers/${driver}.cpp NOWERROR=1 REGENIE=1
07:28 🔗 SketchCo1 So, I don't think the rename trick is bad to do this round, do you
07:28 🔗 DFJustin looks good, you can increase -j4 to -j8 or decrease it to -j2 depending on how many cores the box has and how busy you want to make it
07:28 🔗 SketchCo1 I just try to avoid messing with loader.js, it's brittle
07:28 🔗 DFJustin I didn't think the rename trick would work but it seems to have
07:30 🔗 DFJustin I expected it to be possible to rename the js but not the wasm
07:31 🔗 DFJustin there may be a loader change needed but I don't think it would be in loader.js itself
07:39 🔗 SketchCo1 The first build always takes foreverrr
07:43 🔗 SketchCo1 Linking 1942.html...
07:43 🔗 SketchCo1 warning: undefined symbol: glPopAttrib (referenced by top-level compiled C/C++ code)
07:43 🔗 SketchCo1 warning: undefined symbol: glPushAttrib (referenced by top-level compiled C/C++ code)
07:45 🔗 SketchCo1 We started 1942 at Wed Dec 29 07:24:32 UTC 2021 and finished at Wed Dec 29 07:45:29 UTC 2021
07:47 🔗 DFJustin those warnings are OK
07:47 🔗 SketchCo1 1942 is getting uploaded.
07:47 🔗 SketchCo1 As usual, it will take a small bit of time to reflect
07:48 🔗 SketchCo1 And I backed up all the emularity in case of fallout so it's easy to back out
07:50 🔗 SketchCo1 https://archive.org/details/arcade_1942
07:50 🔗 SketchCo1 Works.
07:52 🔗 DFJustin nice
07:53 🔗 SketchCo1 I mean, I don't see a huge speed jump.
07:53 🔗 SketchCo1 But I assume that's some thing for specific stuff
07:54 🔗 SketchCo1 We started 1943 at Wed Dec 29 07:45:29 UTC 2021 and finished at Wed Dec 29 07:48:35 UTC 2021
07:54 🔗 SketchCo1 We started 1945kiii at Wed Dec 29 07:48:35 UTC 2021 and finished at Wed Dec 29 07:51:56 UTC 2021
07:54 🔗 SketchCo1 So, based on that, about 3 minutes a driver.
07:55 🔗 DFJustin apple2gs was specifically sped up this release but other than that probably nothing major on the mame side
07:55 🔗 SketchCo1 Oooo
07:55 🔗 DFJustin may be some benefit from newer emscripen/clang or could be worse
07:55 🔗 SketchCo1 See, I was led to believe from the phrasing it'd be a bunch of things.
07:55 🔗 DFJustin 1942 was already solid 100% I think so not a good testcase
07:55 🔗 SketchCo1 Yeah, the only main good test cases are dc
07:55 🔗 SketchCo1 And that one insane linux-on-a-chip one
07:56 🔗 SketchCo1 Where people scream in the reviews
07:59 🔗 DFJustin testing with the windows version, looks like a decent speedup on apple2e as well
08:04 🔗 SketchCo1 Wrote a little script to upload the items as we go.
08:12 🔗 SketchCo1 One thing this does come to light is that it's time to get things working doc-wise with the arcade machines
08:12 🔗 SketchCo1 I went ahead and really blasted up the set
08:13 🔗 SketchCo1 jackpotlounge is handled.
08:13 🔗 SketchCo1 But internetarcade and internetarcadeturbo
08:13 🔗 SketchCo1 That's 2,800 games
08:22 🔗 SketchCo1 I think we have your bug in the a2600
08:23 🔗 SketchCo1 Cannot read properties of undefined (reading 'ENOTSUP')
08:25 🔗 DFJustin that's something different
08:28 🔗 DFJustin at first glance I don't know what's causing it
08:29 🔗 DFJustin everything I ran into myself would have affected every system equally
08:32 🔗 SketchCo1 a7800 blows up the same way
08:32 🔗 Sgeo has quit IRC (Read error: Connection reset by peer)
08:33 🔗 SketchCo1 If I had to guess, the arcades are working and the "add this medias" are not
08:33 🔗 SketchCo1 Because it's trying to tell the browserfs something bad
08:34 🔗 SketchCo1 I'll see if bai is around
08:36 🔗 SketchCo1 Bai, we're getting explosions when things reach out for other things
08:36 🔗 SketchCo1 i.e. arcade machines seem fine.
08:36 🔗 SketchCo1 Two game consoles did not.
08:36 🔗 SketchCo1 (a2600 and a7800)
08:36 🔗 SketchCo1 In those, we get:
08:37 🔗 SketchCo1 browserfs.rollup.js:2399 Uncaught (in promise) TypeError: Cannot read properties of undefined (reading 'ENOTSUP') at $.readlink (browserfs.rollup.js:2399) at Object.readlink (mamea2600.js.gz:1) at Object.doReadlink (mamea2600.js.gz:1) at ___syscall_readlink (mamea2600.js.gz:1) at 04b85052:0x51b1b2 at 04b85052:0x51b034 at 04b85052:0x26135d at 04b85052:0x21693b at 04b85052:0x18307f at 04b85052:0x66f4fb
08:37 🔗 SketchCo1 $.readlink @ browserfs.rollup.js:2399
08:37 🔗 SketchCo1 readlink @ mamea2600.js.gz:1
08:37 🔗 SketchCo1 doReadlink @ mamea2600.js.gz:1
08:37 🔗 SketchCo1 ___syscall_readlink @ mamea2600.js.gz:1
08:37 🔗 SketchCo1 $func9869 @ 04b85052:0x51b1b2
08:37 🔗 SketchCo1 $func9868 @ 04b85052:0x51b034
08:37 🔗 SketchCo1 $func4232 @ 04b85052:0x26135d
08:37 🔗 SketchCo1 $func3747 @ 04b85052:0x21693b
08:37 🔗 SketchCo1 $func3359 @ 04b85052:0x18307f
08:37 🔗 SketchCo1 $func11409 @ 04b85052:0x66f4fb
08:37 🔗 SketchCo1 $func23507 @ 04b85052:0xbd3c6b
08:37 🔗 SketchCo1 $func11029 @ 04b85052:0x62955c
08:37 🔗 SketchCo1 $func11030 @ 04b85052:0x629aed
08:38 🔗 SketchCo1 $wk @ 04b85052:0xc3292d
08:38 🔗 SketchCo1 Module._main @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 callMain @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 doRun @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 run @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 runCaller @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 removeRunDependency @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 receiveInstance @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 receiveInstantiationResult @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 Promise.then (async)
08:38 🔗 bai just crawled into bed so I can't dig too deep....readlink implies we're using symlinks? and maybe the fs type we're using doesn't support them?
08:38 🔗 SketchCo1 instantiateArrayBuffer @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 instantiateAsync @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 createWasm @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 (anonymous) @ mamea2600.js.gz:1
08:38 🔗 SketchCo1 04b85052:0xe96ed Uncaught RuntimeError: null function or function signature mismatch at 04b85052:0xe96ed at Object.Module.__ZN13sound_manager4muteEbh (mamea2600.js.gz:1) at MAMERunner.value (loader.js?v=51c75c55:1163) at loader.js?v=51c75c55:1685
08:40 🔗 SketchCo1 https://archive.org/details/arcade_actfancr also blows up
08:41 🔗 SketchCo1 So, I'll give us a day to figure it out, and then I need to back these out.
08:41 🔗 SketchCo1 (I can put in a few of the others, of course, but these are definitely indications it's hit or miss.)
08:41 🔗 SketchCo1 I'm sure it's a simple solution.
08:42 🔗 DFJustin actfancr is something different again
08:42 🔗 bai localStorage, IndexedDB, and MountableFilesystem don't support symlinks...in fact it looks like "emscripten" is the only fs backend that does, I'm not really sure ehat that means with our setup
08:42 🔗 SketchCo1 Well, I'm going to do this
08:42 🔗 DFJustin a2600 runs for me with a different loader so I think it's emularity-related
08:42 🔗 SketchCo1 I'm going to backout a2600
08:42 🔗 SketchCo1 Just because that has lots of stuff people use
08:43 🔗 SketchCo1 I'm going to leave a7800 broken
08:43 🔗 SketchCo1 Since nobody cares about a7800
08:43 🔗 DFJustin harsh
08:43 🔗 SketchCo1 (Especially as we have it set up)
08:43 🔗 SketchCo1 Well, our 7800 is super crippled because atari basically told me to take 100% down
08:44 🔗 SketchCo1 But a2600 actually has users on our site
08:44 🔗 SketchCo1 and if stuff breaks they WILL start filling up an actual staff we have now
08:44 🔗 SketchCo1 it used to be it would forward to me
08:45 🔗 SketchCo1 now 12 people will take a shot at it
08:45 🔗 SketchCo1 a2600 at normal.
08:45 🔗 SketchCo1 So, I'm going to bed myself, maybe we can think out a7800 and what's causing that issue
08:45 🔗 SketchCo1 I mean, covid and all
08:50 🔗 SketchCo1 So, right now, I'm going to
08:50 🔗 SketchCo1 - Let the compiler run
08:51 🔗 SketchCo1 - Cherry-pick a few ones (apple2gs, apple) that were cited as faster
08:51 🔗 SketchCo1 - Put them in, back out if they die
08:51 🔗 SketchCo1 My priorities right now need to be recovery from covid and podcast recording before end of month.
08:51 🔗 SketchCo1 But I can try a few.
08:52 🔗 SketchCo1 I see I don't add 39-in-1 which is silly. After it's done compiling, I'll do it, just to see if it's faster.
08:52 🔗 SketchCo1 https://archive.org/details/arcade_39in1
08:52 🔗 SketchCo1 I love this fucking thing, it is chef's-kiss-annoying to people
08:53 🔗 SketchCo1 Currently 17% speed
08:54 🔗 DFJustin ok, best wishes on your recovery
08:54 🔗 SketchCo1 Thanks so far
08:56 🔗 bai a quick search through the mame codebase for calls to readlink doesn't turn uo anything that hasn't been there for 6+ years so I don't think anything changed there
08:59 🔗 bai I'll do some more digging tomorrow
15:12 🔗 Sgeo has joined #jsmess
17:34 🔗 SketchCo1 Sleep was for the weak, and I am, weak
17:34 🔗 SketchCo1 The compiler is down to the d's
17:42 🔗 SketchCo1 Suppose I should bite the bullet and try apple2e
17:44 🔗 SketchCo1 Apple2e clicking in, backout at the ready
17:51 🔗 SketchCo1 browserfs.rollup.js:2399 Uncaught (in promise) TypeError: Cannot read properties of undefined (reading 'ENOTSUP')
17:52 🔗 SketchCo1 I've pulled apple2e back
17:52 🔗 SketchCo1 Also stopping the compiling because clearly this problem is endemic
17:53 🔗 SketchCo1 I am going to make 39in1 just in case, though :)
18:04 🔗 SketchCo1 Ha ha, 39in1 is SLOWER.
21:57 🔗 SketchCo1 Anyway, we can revisit, but the rush is over - the disadvantages outweigh advantages, and it's worth figuring out what's busting.
21:57 🔗 SketchCo1 And also, since we have bounced up to Emscripten 3.1.0, I bet alon might have some generalized hints for us.
22:02 🔗 SketchCo1 I've reached out to Alon to ask.
22:52 🔗 SketchCo1 Just throwing some things at him, to see if anything jumps out
22:59 🔗 SketchCo1 Where's the compilation lines in the code for mamedev again?
23:23 🔗 DFJustin https://docs.mamedev.org/initialsetup/compilingmame.html#emscripten-javascript-and-html
23:23 🔗 DFJustin https://github.com/mamedev/mame/blob/mame0239/scripts/genie.lua#L1150
23:24 🔗 DFJustin not sure what question you were asking but guessing it's one of those two
23:35 🔗 SketchCo1 Second one was
23:36 🔗 SketchCo1 I am always on the hunt for some amazing upgrades emscripten made, which we didn't know to add GIRAFFENECK=1 to the options
23:36 🔗 SketchCo1 Followed by being told adding USE_WEBGL2 in any context is like making an infant smoke
23:37 🔗 SketchCo1 John Vilk says he'll come look
23:38 🔗 DFJustin there's some simd option I was meaning to look at
23:43 🔗 SketchCo1 I do wish the runtime was smaller
irclogger-viewer