log☇︎
1100+ entries in 0.143s
asciilifeform: ( the FG master/slave mechanism is only used in test benching )
asciilifeform: the primary headache of old FG is imho the very modest output rate, which makes the thing take ridiculous length of time to test on my bench
mod6: http://btcbase.org/log/2018-10-22#1865239 << This would be a neat test. ☝︎
asciilifeform: afaik this is not a test that has actually happened in trb history ( longest naturally-occurring reorg since trb first saw use in the battlefield was, iirc, 8 blox or so ? )
asciilifeform: seems that he wanted to test whether trb's miner component actually worked.
asciilifeform: my current best 4096bit modexp is still 22 sec. on the 'standard' test iron. but with barrett ( next ch. ) cuts down to ~5sec iirc; and with asm, hypothetically <1s
asciilifeform: ACHTUNG, panzers! zoolag going down for ~30min for test of experimental patch, in ~10m
asciilifeform: test unit here is 'n150' .
mircea_popescu: but yes, totally -- go to college, on entrance test basis, 100% free, luxurios accomodations even. however at the end of it, the bottom half straight to kolyma.
asciilifeform: aside from the test where you actually gotta explore the solution space to converge on solution, lol
mircea_popescu: and yet -- very intelligent by any test.
BingoBoingo: For some reason my tonsils got huge and they feel like they are trying to pull out the insides of my ears. Going to go test the socialized medicine.
a111: Logged on 2018-10-12 18:18 asciilifeform: the 1 'litmus test' i was able to think of , is the 'i pick a block hash and you gimme the block in <1s' algo.
asciilifeform: and instead each variant is a thing that its author is forced to actually test as-written.
mircea_popescu: fwiw i simply discard lots and lots and lots of "705babygirl 23F sub 1m What?" ; they come in at the rate of a few / minute, and who the fuck cares anyway. "think of it as an intelligence test -- that you failed", right.
mircea_popescu: i guess there's no way out of implementing some inverse-index or i guess ngram really, to test this out. mebbe my intuition is entirely off
asciilifeform: the 1 'litmus test' i was able to think of , is the 'i pick a block hash and you gimme the block in <1s' algo. ☟︎
mod6: btw... jurov and I are working on relocating the bitcoin.foundation website to the foundation's server in Pizarro. We're gonna get it set up, and test before we announce a cut-over date. Smooth transition, ftw.
ave1: diana_coman, I'll add my test and fix the typo and regrind
deedbot: Invoiced mircea_popescu 0.05786600 << S.MG Test Machine
BingoBoingo: !!invoice mircea_popescu 0.05786600 S.MG Test Machine
asciilifeform: Mocky: 'aggressive' trb vs regular, is just about 'night and day'. i also have ben_vulpes's 'super aggression' item lined up for test.
diana_coman: is uy3 the test s.mg server?
mircea_popescu: get the latter, why not. feel free to test it once on, also, give the girl a start.
asciilifeform: mircea_popescu: if he's stuck to the old gnudiff format, which per my test appears to be the case, names aint hashed at all ( they're covered by the gpg sig of given patch, but that's it )
asciilifeform: diana_coman evidently did not test this ( not really criticism of diana_coman , iirc we did decide that republican proggies ought not to include raw binariolade )
phf: diana_coman: kk, will test and upload
mircea_popescu: iirc 20kb packets made it over test
mircea_popescu: http://btcbase.org/log/2018-10-01#1856857 << these limits are everywhere. "we deliver" "no you don't" / "restaurants exist" "no they don't : http://trilema.com/2017/fake-news-are-just-one-tail-of-the-failed-female-state/ " and so on. the world essentially consists of the ~assumptions~ of existence and function of a large crowd of morons who never test these. ☝︎
Mocky: don't test me
a111: Logged on 2018-09-28 06:24 diana_coman: asciilifeform's published test data seems to match what I got on my initial tests with 1 second delay; my current plan is to collect first at least 1 week worth of data and then to repeat the experiment with a. smaller delays b. several senders perhaps
a111: Logged on 2018-09-28 00:58 mircea_popescu: but yes -- the test can (and likely will) be tightened. for starters we just wanted to get a sort of "absolute path limits". and THESE do indeed turn out to be further out than originally thought -- 2kb packets make it np unfragged and in order 100% of the time, and even 20-60kb packets made it.
diana_coman: asciilifeform's published test data seems to match what I got on my initial tests with 1 second delay; my current plan is to collect first at least 1 week worth of data and then to repeat the experiment with a. smaller delays b. several senders perhaps ☟︎
asciilifeform: mod6: i seem to recall a much sadder london test but that was with very heavy packets iirc
mod6: thats still 2 orders faster than the last test tho.
mircea_popescu: but yes -- the test can (and likely will) be tightened. for starters we just wanted to get a sort of "absolute path limits". and THESE do indeed turn out to be further out than originally thought -- 2kb packets make it np unfragged and in order 100% of the time, and even 20-60kb packets made it. ☟︎
mircea_popescu: early proto-test indicated most of the lossage happens before 2nd hop
asciilifeform: diana_coman's test jig ( i did not modify it except for the dest ip ) currently fires 1 / sec.
asciilifeform: going by the current empirical test, however, a packet that frags into 2 or even 3, typically goes. tho it remains to be seen whether they start falling down once you saturate.
asciilifeform: diana_coman: i'ma test & mirror that one when i get a chance
asciilifeform: i'ma test.
asciilifeform: afaik nobody's used esthlos's item in battlefield yet, it'll need good bit of test.
asciilifeform: aaand it loox like diana_coman already did my chore for me... i'ma do the elementary test on it nao, and sign/mirror.
asciilifeform: anyway i have no opposition to an experimental test
a111: Logged on 2018-09-25 14:57 asciilifeform: diana_coman: i'd like to see a test where there is only 1 size, across various sizes, if possible
a111: Logged on 2018-09-25 12:30 diana_coman: uhm, correction re above since I messed up the pilot test earlier (sender sent 100 as shown, but receiver waited for only 50 and then turned off so not much use); updated pilot test with receiver properly looping forever and sender sending 100 packets: 81 of those made it; sender log: http://p.bvulpes.com/pastes/zs4p5/?raw=true ; receiver log: http://p.bvulpes.com/pastes/4qiPb/?raw=true
asciilifeform: diana_coman: i'd like to see a test where there is only 1 size, across various sizes, if possible ☟︎
diana_coman: I can certainly run such a test too if we want to check that specific situation; now that I got ~everything in place, knobs can be easily turned for sure
asciilifeform: esp if all fragged ( e.g. test with size of 2048 strictly )
diana_coman: uhm, correction re above since I messed up the pilot test earlier (sender sent 100 as shown, but receiver waited for only 50 and then turned off so not much use); updated pilot test with receiver properly looping forever and sender sending 100 packets: 81 of those made it; sender log: http://p.bvulpes.com/pastes/zs4p5/?raw=true ; receiver log: http://p.bvulpes.com/pastes/4qiPb/?raw=true ☟︎
diana_coman: the pilot test is UK -> UY (same as yesterday's)
diana_coman: for completeness, the logs from the new pilot test: sender log is http://p.bvulpes.com/pastes/SdU8P/?raw=true and corresponding receiver log is http://p.bvulpes.com/pastes/OXUWa/?raw=true
diana_coman: http://btcbase.org/log/2018-09-25#1854219 -> tiny new pilot test seems to suggest it's not as bad as that: all 100 packets of a batch (sent in burst mode, no delays) made it, with pseudo-randomly chosen sizes between 6 and 2048; predictably though, order was messed up at times ☝︎
asciilifeform: ( i predict that any frag at all will give loss that cancels out the win from using heavy packets. but nobody should take my word for it, test, test )
a111: Logged on 2018-09-24 13:50 diana_coman: hm, a first tiny pilot test of the UDP send/receive looks quite dire (4 in 20 made it, when sent in batches of 4, random lengths); however, I don't know if it's not just overflowing the out buffer to start with (since default value in /proc/sys/net/core/wmem_default is 212992 so real would be half that iirc)
diana_coman: hm, a first tiny pilot test of the UDP send/receive looks quite dire (4 in 20 made it, when sent in batches of 4, random lengths); however, I don't know if it's not just overflowing the out buffer to start with (since default value in /proc/sys/net/core/wmem_default is 212992 so real would be half that iirc) ☟︎
asciilifeform looks forward to test
asciilifeform: trinque: neato. and asciilifeform loox forward to test.
BingoBoingo: "personas fisicas" also get to make THREE compras de exterior without taxes provided they are under 200 U$S and 20 kilos which can be tracked here https://aplicaciones.aduanas.gub.uy/LuciapubX/DECLARACIONES.Cargas.HWCantEncPostales.aspx and only shows an incoming test parcel I've got inbound.
diana_coman: asciilifeform, yes, atm I'm looking at it for the test harness, for which I think it fits great (since I wouldn't want to go writing 1-65535 explicitly if I can help it)
asciilifeform: but would rather avoid an ocean of os-specific items that i'm expected to sign despite never having tested and possibly not even owning the iron with which to test ( again see the #ifdef megathread )
diana_coman: asciilifeform, the timeout was the only one I hesitated on too (and the only one I'm using in the original test for that matter) but precisely like you, when I thought of it, I came up with "well, threading should be enough anyway " and uhm, no reason why it's *needed* there
diana_coman: asciilifeform, ah, I forgot to mention it explicitly but yes, my tests include ave1's gnat as well as adacore's gnat; this is pretty much for any ada code I test
asciilifeform: still needs a bigendian test tho.
diana_coman: asciilifeform, confirmed working nicely with its own tests + adapted client/server test as used previously
diana_coman: ave1, I adapted my udp test from http://ossasepia.com/2018/09/14/smgcomms-implementation-chapter-1/#selection-159.0-159.2325 to use your suckits and it works!
a111: Logged on 2018-09-14 18:09 phf: asciilifeform: i have a bigendian box and i also tested at least one call for bigendian test on it. i think it was diana_coman's keccak though
asciilifeform: ( good chance to test just how portable is the latter )
phf: asciilifeform: i have a bigendian box and i also tested at least one call for bigendian test on it. i think it was diana_coman's keccak though ☟︎
asciilifeform: this being said, still gotta come up with bigendian box to test with at some point ( iirc trinque had one somewhere, but afaik no gnat for it yet )
asciilifeform: diana_coman: currently i am lacking a bigendian iron (or a gnat for such) so never got a chance to truly and properly test endianism conversion (as for ffa, it is endian-insensitive, but this is simple where there is no networking)
asciilifeform: some serious load test would be neat
mircea_popescu: this isn't a critical application yet. experimenta, meaning -- if you want to use this to test 'em or w/e i don't mind.
asciilifeform: re sd card -- mircea_popescu may find it interesting that asciilifeform's initial (week before flight) rk test bed used a premium 'camera' 128G sd to boot ~and~ run from. but found that it is intolerably slow, the things have pretty sad randomaccess performance
asciilifeform: diana_coman: test anyffing you can think of.
diana_coman: asciilifeform, well yes, but this was more than a reset at any rate so I'd rather re-test anyway
asciilifeform: diana_coman: fwiw fg is not affected by resets; but i won't presume ever to tell anybody not to test theirs ( personally i only 'is it still there?' on machine boot, but thats just me )
asciilifeform: diana_coman also be sure to test yours
mod6 goes off to test the rockchip
a111: Logged on 2018-09-11 14:39 asciilifeform: mircea_popescu: i test'em before handing over to mod6 .
mod6: http://btcbase.org/log/2018-09-11#1849451 << asciilifeform handed them over to me to test they had been reset, and one could login via a temp password. This was done to douchebag's and mats' old rockchips. They both worked great after reset. I never touched the /etc/conf.d/net file. ☝︎
asciilifeform: mircea_popescu: i test'em before handing over to mod6 . ☟︎
deedbot: Invoiced mircea_popescu 0.05780660 << S.MG Test Machine (UY3) through September 2018
BingoBoingo: !!invoice mircea_popescu 0.05780660 S.MG Test Machine (UY3) through September 2018
BingoBoingo: They gotta constantly A/B test their copywriting
asciilifeform: mircea_popescu: what means 'test' here ? they'll put customer traffic through it instead of juniper ?
mircea_popescu: so, you're paying good money to have BingoBoingo on the ground there ; what "on the ground means" is that he can go to dc and say "hey guise, will you test this box we make ? no charge". IF they agree, and IF you have made the tmsruniper, and IF indeed they come out of it thinking they're better off with that, THEN you can sell it for credit, and not need to do more wires. yes ?
asciilifeform: seems like the only way to test is empirically
asciilifeform: there also appears to be a knob that switches off the usb console entirely, but i have not tried it yet, will test when i'm about to throw out the current mobo i have in there (with the damaged mouse connector) and put in new one (i have replacements)
lobbes: question for the V experts out there: I was test pressing a vpatch genesis I created and ran into an odd barf ('Pressed file hash did not match expected!') >> http://p.bvulpes.com/pastes/A5d1u/?raw=true
mircea_popescu: nah, sig test would have to be added.
asciilifeform: how to do : it charges cap ( the demo unit i have here from china, has 4 1F 2.2v caps ) and runs a boost converter from these, as well as lighting up 'power good' pin when 3.3v (from the latter) is available. so you put a test load on the 3.3v out ( say , 1W 100 ohm resistor ) and record events when 'power good' signal changes.
asciilifeform: mircea_popescu: in su there was legend of at the very least 1 test pilot who flew mig29 vertically into ground , when wife left
BingoBoingo: Apologies for the delay, I gave esthlos Vs some test drives
mod6: Alright, so I ran the script, and I got an error when it did the apache configuration test near the end of the script.
lobbes: still need to test the hell out of it, so I'm aiming for maybe a week (which, still gives me 2 months for auctionbot ETL. eh. will see)
trinque: sure, only thing I'd counter is the game by which they steal is the same test-taking nonsense seen elsewhere
asciilifeform: BingoBoingo: danke schon ! lemme test fire it, brb
Mocky: there are still ways to do it, e.g. with the count method, you have to test for count < 1.
mircea_popescu: though might give you history / lang test, i dun recall