log☇︎
1900+ entries in 0.174s
asciilifeform: any rng test can be rewritten as a producer of a predictor. or the test is rubbish.
asciilifeform: that way you do not have to wonder wtf the particular test consisted of. and can immediately replicate.
mod6: my second run on the 2nd FG out of my bunch has completed. results here: http://www.mod6.net/fg/fg-test/fg2.ent_run2.txt http://www.mod6.net/fg/fg-test/fg2.dieharder_run2.txt
mod6: shinohai: sure will do. it's just the eatblock test. once it's complete, i'll re-run all of the charts & metrics
mod6: In other news, my eatblock re-test is at 433K+ & into blk0042; not much more to go there...
mod6: also, eatblock test is no around 411K bocks and into blk0034... so should be done in a day or two.
mod6: Running another test against the same phyiscal fg right now (figure I'll do three per unit).
mod6: Ok update here... 2nd FG first entropy tests are in; collected ~1.2Gb. http://www.mod6.net/fg/fg-test/fg2.ent_run1.txt http://www.mod6.net/fg/fg-test/fg2.dieharder_run1.txt
asciilifeform: liquishit that doesn't pass the laugh test.
mod6: Eatblock test upto 363K+ blocks and into blk0019.
mod6: http://www.mod6.net/fg/fg-test/fg1.ent_run4.txt http://www.mod6.net/fg/fg-test/fg1.dieharder_run4.txt
asciilifeform: one of the items on asciilifeform's back burner recently has been to properly test various variants of this.
mod6: I'm doing that test next.
mod6: Alright asciilifeform, mircea_popescu, et. al. Here's the 3rd run of the dh & ent against the third collection of ~1.1Gb of entropy. http://www.mod6.net/fg/fg-test/fg1.dieharder_run3.txt http://www.mod6.net/fg/fg-test/fg1.ent_run3.txt
mod6: ok here's what I had when I copied it: http://www.mod6.net/eatblock-test/debug.log
mod6: alright, mircea_popescu & asciilifeform ; full offline eatblock test is running again
jurov: on the yes test, i got warning:
mod6: i'll try that next time i run a test.
mod6: is this kinda what you're looking for? http://www.mod6.net/eatblock-test/ProcessBlockTimeVDBWaitTime.png
a111: Logged on 2017-04-18 05:31 pete_dushenski: http://www.mod6.net/eatblock-test/charts/Memory.png << whoa
pete_dushenski: http://www.mod6.net/eatblock-test/charts/Memory.png << whoa ☟︎
mod6: mircea_popescu, asciilifeform, ben_vulpes, et. al: Eatblock test results & blog post: http://www.mod6.net/eatblock-test/
trinque: gives one fail on "address test" before it's gone for good.
trinque: x1 carbon's got bad ram, actually freezes its own firmware memory test.
mod6: I think I just reached the end of my eatblock test...
asciilifeform: the sad thing is, dh was mostly used historically to test prng. so the 'waste' was, i picture, not seen as problem.
asciilifeform: dh also does a very, imho, questionable thing, of throwing away most of the input (yes) on account of insisting on using new segment of the input for each test
asciilifeform: now at the rate dh eats bytes, a fully rewind-free test of, e.g., FG would take...
mod6: Here are my results from the second run from the 1st FG tested: http://www.mod6.net/fg/fg-test/fg1.ent_run2.txt http://www.mod6.net/fg/fg-test/fg1.dieharder_run2.txt
mod6: gonna see what those results look like... then will probably continue on and test the remainder of the fgs I've got.
asciilifeform: http://web1.bookoo.com << 'Hi. Welcome to TEST Bookoo! Our fantabulerrific online yard sale in TEST' >> looks like a faux alibaba
asciilifeform: usa was , at its root, a massive test of 'hey nations are bogus, nobody needs, let's all be Independent People Doing Own Thing'
mod6: asciilifeform: werd. anyway, it's exciting to test these.
asciilifeform: also i recommend to test every unit, aha
mod6: I'll test all of my other units in this same fashion too.
mod6: anyway, just wanted to be /sure/ that I wasn't getting some sort of varient test results from the dh.
asciilifeform: ftr no fg ships that reports a fail, or a 'weak' in any test consistently (2+ 1G runs.)
asciilifeform: mod6: certain % of the time, test 'weak' and , rarely, 'fail' ( simply means that author of dh decided that given result falls out of uniform distrib. range.)
mod6: mircea_popescu, asciilifeform, et. al.: My unboxing and testing of a randomly selected FG shipped to me: http://www.mod6.net/fg/fg-test/
mod6: i'll include the advanced test parts before publication.
mod6: So I collected 1.2Gb worth of data from FG. Which for the Advanced test I don't think would be a problem... about to do the `ent` & `dieharder` tests.
a111: Logged on 2017-04-14 14:57 mod6: with both of my RNG-TWs connected, it was collecting at about 7kB/s ... for that quick 10s test or so to ensure that it was a "full speed"
mod6: with both of my RNG-TWs connected, it was collecting at about 7kB/s ... for that quick 10s test or so to ensure that it was a "full speed" ☟︎
mod6: fg advanced test update: i've collected about 500Mb so far. ~half way there.
asciilifeform: and want to fist-pass test if this is in fact so.
mircea_popescu: yeah, but it has test weights. just as good.
mircea_popescu: and does the test measure at 10.000 ?
asciilifeform: the weight is a test standard (for some reason the box came with 2, and 1 lives in a dedicated hole in the chassis)
Framedragger tried fg last weekend, was all good, (very) small sample (2.7MB) had 7.999936 bits of entropy per byte. but yet to test more thoroughly, including removing shields, etc.
mod6: a good test either way.
mircea_popescu: !!deed for instance works fine by my eyes, but it does not pass the "monkey do, monkey see" test.
a111: Logged on 2017-04-09 13:48 BenBE: I'm the maintainer of the GeSHi syntax highlighter for PHP, which is used in e.g. Wikipedia for source highlighting of articles. Also working on several crypto-related projects like my own TLS/SSL test, a collection of publicly-known set of compromised keys, an OpenSource management software for handling X.509 certificate issuance for a certificate authority.
BenBE: I'm the maintainer of the GeSHi syntax highlighter for PHP, which is used in e.g. Wikipedia for source highlighting of articles. Also working on several crypto-related projects like my own TLS/SSL test, a collection of publicly-known set of compromised keys, an OpenSource management software for handling X.509 certificate issuance for a certificate authority. ☟︎
mircea_popescu: i can see the theoretical "benchmarks" online too. the problem is they're bs, "test file generated by seq 100mn". hurr. how about generate a file with lines of random length between 1kb and 1mb, then benchmark it.
phf: a sentence like "hello world this is test" might get an invocation like render("world this") followed by render("hello") followed by render("is test"), simply because higher level widget engine decided that's the order of exposure, or hierarchy, or whatever
mircea_popescu: oh, oh, for YOUR test beds. i see. lol
asciilifeform: for the test beds, noshit
asciilifeform: not speaking of machines here, i dun have a 'large comp that ain't a c machine' to even test with.
asciilifeform: (picture if two d00dz were sent into two separate dungeons , and promised impalement if they come out with c compilers that produce binaries for particular test program that differ EVEN IN ONE BIT. quite impossible for them to avoid the stake, because c is ~nonstandardized~, in the sense where the standard does NOT specify all cases)
asciilifeform: it is so as to make the deterministic test possible. turns out, the warm-up of crystal oscillator is NOT deterministic. so if you let the thing begin IMMEDIATELY when the clock is valid,
asciilifeform: (and the diff b/w a shield and an antenna is the ground. who is going to test the grounds? adlai ?)
mircea_popescu: yeah, right, one's the kinda-almost test harness
ben_vulpes: http://btc.yt/lxr/satoshi/source/src/test/test_bitcoin.cpp?v=wires_rev1#1088
asciilifeform: ben_vulpes: iirc the test target built a turd that exercised data structures, serialization, etc. rather than whole thing
ben_vulpes: currently attempting to reason about differences between targets bitcoind and test_bitcoin
ben_vulpes: anyways, hey asciilifeform have you ever built the trb test target?
Framedragger: asciilifeform: ack! got yer package after customs cleared it. visually looks intact and in good order. (won't be able to test it till ~weekend)
asciilifeform: ( it would be possible to build a behavioural test stand for it, if the pinout and timing characteristics were known. theoretically anyone with a 'macivory' and some spare time, could come up with these. )
shinohai: Bored with all those other apps? Examine thy semen! http://www.theverge.com/circuitbreaker/2017/3/22/15009960/male-fertility-device-smartphone-app-semen-test
Framedragger: test
asciilifeform: ben_vulpes et al : g5 is valuable , on top of other reasons, to test 'bigendian' behaviours of various proggies.
trinque: runs (for example) a test server for deedbot just fine
a111: Logged on 2017-03-20 01:04 Framedragger: sure. my internal alfatron says "may as well test for tape backup storage system viability" though; surely it's inefficient to test *all* things in a domain that one is not optimistic to begin with. but i see your rationale of course
Framedragger: sure. my internal alfatron says "may as well test for tape backup storage system viability" though; surely it's inefficient to test *all* things in a domain that one is not optimistic to begin with. but i see your rationale of course ☟︎
mircea_popescu: wtf is this. did it test out ok ? did the pos just die ? WHO THE FUCK DOES THIS
asciilifeform: phf: i've contemplated a 'cheat' : run macivory 'lifesupport' on patched m68k emulator (e.g., 'basilisk') that would bring nubus i/o out to a test stand, on which the card would sit
phf: aah, yeah, i'll test it in xl
asciilifeform: i was merely confounded by 'i'll test it in my ivory'
phf: i'm not sure if it's working, it's from a "possibly broken" box, but i can lend you that one after i test it in my ivory, and if you have a viable project of some sort, i can probably get you more
asciilifeform: mircea_popescu: demurraging coin fails the berlin wall test.
mircea_popescu: altogether not a bad idea to make a set of test chains as per the tsting discussion last week. this can surely be both part of it and the incentive to do it.
mod6: There has been an automated test added for this as well.
mod6: this change is pretty trivial. probably will update V99994 before it comes out (was about to release this weekend). will need to add at least an automated test too. not a big deal.
mircea_popescu: makefile bs. alright, well, i guess we'll need a test for file exists or some such hair.
mod6: let me dig it up -- bumeanwhile, it's regarding src/obj, src/obj/nogui, src/obj/test
asciilifeform: so this'll test the sync-from-0
mircea_popescu: a right, we get to test this too!
asciilifeform: meanwhile in the test room, 'Serial correlation coefficient is 0.000000 (totally uncorrelated = 0.0).' << i had nfi this was possible
mircea_popescu: this is no kind of argument against having test suite.
mircea_popescu: prolly but that's finnickyer. unless of course we bother to create 3rd item here, which would be a test chain.
ben_vulpes: should even be possible to test reorg behavior!
mircea_popescu: prolly should churn the chain as a test yeh. also a tx fuzzer would be great in general.
ben_vulpes: i went so far as to set up a solipsistic test net before bumping into the dumb-as-rocks "needs 2 nodes in order to mine" shit
asciilifeform: (it gets test-drained weekly, also.)
mircea_popescu: i've not seen any. would be more interesting to test for the hole manually by fuzzing on testnet
asciilifeform: mircea_popescu: could be interesting to test for double spends 'manually' (using specially written proggy that walks the blox)
Framedragger: hey, you wanted some fs test, i'm just reporting on levels of shittiness found
Framedragger: but on the other hand this'd benchmark/test directory traversal more.
asciilifeform: (would like to test 2way 'wiring' first on a non-infrastructural set of boxen )
Framedragger: will get a way to test real disk soon, didn't want to run on personal trashy PC, hence shitty server