343200+ entries in 0.227s
mircea_popescu: anyway, re
the other point : even if scheme does in
the end become an integral part of sane error management a) his work is very important for providing
the structure
that'll illuminate
that and b) will prolly be mostly reused strings in different wrapping anyway.
ascii_butugychag: (i.e. part 1 can be mechanically-litmused: 'this only adds
the
ts genesis material')
polarbeard: usually
there is a lot of blind
trust involved
mircea_popescu: i'm slowly discovering what an IMMENSE help
to debugging and code maintenance
this
thematic unity
thing is, incidentally.
polarbeard: mircea_popescu: yep, will come back with a
task-unified group of patches
mircea_popescu: specifically because "majority is better
than minority in every practical way" except
the one way
that actually matters, which
the majority is never willing
to confront.
mircea_popescu: and
this is how bitcoin will continue
to kick
the
teeth in of any sort of fiat, no matter how superficially bitcoinized.
mircea_popescu: THIS is how bitcoin,
the "minority coin" par excellence, kicked fiat's
teeth in.
mircea_popescu: thestringpuller minority coins may be "worse in every practical way", but
they are BETTER in
the only way
that matters : minority doesn't inflate, majority does.
mircea_popescu: the best way forward, imo, is
to restructure it as 1) one large patch consisting of nothing but
the return fixes and some smaller patches doing
the rest of
the stuff built on
top of
that.
mircea_popescu: polarbeard don't get
this wrong, you did a lot of very useful work in
that patch.
thestringpuller: "And minoritycoins will be worth much less
than majoritycoins, because
they are worse in every practical way.
They
take longer
to confirm, fewer people accept
them for payment,
they are protected by less hash power, and
they are much more difficult
to
transact safely." << Not if majority coin is sunk early on and money flees
to minority coin. What "actual" economists is
this d00d hanging out with
that "vetted his scaling propos
polarbeard: punkman: yes,
the rube-goldberg pipe was inteded
to be used as a nohup'd
tail on debug.log
ascii_butugychag: mircea_popescu: one of
the items i have in mind for shivafication is sane event logging
polarbeard: I'm open
to suggestions but now I'm a bit saturated of
this, I'll somewhere else for some days...
polarbeard: yeah, I understand people don't want
to run rube-goldberg machines
mircea_popescu: or whatevs,
tail logfile | mess for
teh realtime stuff
mircea_popescu: the point is
that people running bitcoind have a whole array of measures, which mostly work as cat logfile | mess
polarbeard: one field forward for
the category, if you want both
then yes, more piping
polarbeard: xargs -I'@' sh -c 'echo \"@\" |cut -d\" \" -f1,3- |
tee \$(echo \"@\" | cut -d\" \" -f2)'
punkman: polarbeard:
the
thing usually runs as a daemon
mircea_popescu: yes, if you just dump into console i'm going
to have
to pipe
the program
throuygh a bunch of different greps
mircea_popescu: note however you should dump
to file not
to console. because otherwise mess of
tee's
mircea_popescu: note
that
the only way i got my apparently very controversial /s /t patch
to even be considered was showing how it could be machined.
polarbeard: also I
thought you can dump
to console and grep until we reach consensus on how we want
to do
the actual log splitting and filtering
ascii_butugychag: so i can ask
the same question, line after line, for a whole patch
mircea_popescu: polarbeard yeah, you get
the right idea, but seriously, making a single patch consisting of JUST -return error +return error is
the right way
polarbeard: mircea_popescu:
the patch ended up being a bit verbose, yep, but it made sense
to me fixing
the messages as I was adding metadata,
that's why I didn't add filtering,
to not add actual logic
to an already pretty big patch
mircea_popescu: 1601+ return error(SBLK "failure wrting block file %d
to disk due invalid stream position", nFileRet);
mircea_popescu: there's a lot of stuff in
there
that will necessarily result in a huge patch (the guy is stuck, willy-nilly, doing a lot of
ascii_butugychag: mircea_popescu: it bothered me
that polarbeard apparently does not yet fully grasp
that every line he writes is a unit of ~work~ for other people
to do.
☟︎ mircea_popescu: punkman
the rotating log fix separate from
the rest, for one
thing.
mircea_popescu: you
two made me google
that damned
thing. seriously, a mouse ?
pete_dushenski: ascii_butugychag: nor i. rat9 looked spiffy but chewed
through swappable battery in ~5 hours. even when 'worked' it
twitched and jumped and spasmed and
triggered all kinds of inane shit. couldn't even highlight menu bar drop-downs.
mircea_popescu: . i'm with ben_v (and apparently ascii
too) in
that it's really best practice
to make patches small and issue-focused. it's not just a matter of review, it's
to ease future rebasing and so on. as antidesign it may seem
to you personally on
the basis of workflow,
this is one of
those
things where individual has
to
take a small step back
to allow society altogether
to exist.
mircea_popescu: . i'm with punkman in
that if you're going
to prefix in
that much detail,
there's going
to have
to be some sort of filtering scheme. i'm not with him
that flags are
the solution, it just seems an ad-hoc hack unix did
that got carried forward. i'm still partial
to my multi-logfiles idea, fwiw. altogether a lot more
thought must be put into "how
to structure"
this, but sadly it depends on a fully specified bitcoin, so it
ascii_butugychag: pete_dushenski: supposedly
there is an x11 patch
to make it go
ascii_butugychag: pete_dushenski: rat9 suxx, you can have mine for
the cost of postage
mircea_popescu: just, in a metaprogramming sense
that did not exist prior
to V putting light on
the issue.
assbot: Logged on 29-01-2016 16:06:03;
thestringpuller: how you gonna sell a mouse for 60 bucks
mircea_popescu: so in full
terms, i would say
that again, including both in
the same patch is both premature optimization and a kludge.
phf: a way
to indicate (a->a' b->b') would normally be
to include
them both in a single patch. i
think
that's how it's done now
mircea_popescu: nothing of course prevents you from
taking z and rebasing it as a first level node off x.
mircea_popescu: because people didn't make both y and z first level nodes of x, people are indicating
that whether you know why or not, you need a->a' for b->b'.
mircea_popescu: phf nothing in
the physical situation as you describe it indicates
that. nevertheless,
the implicit meaning in how
the
topo sort owrks is
that you need a
to a' for b
to b'
to be meaningful.
this could be because one is .h and
the other .cpp, fopr instance, or becauise other stuff is coming later or for any other reason
phf: right now it seems like an artifact of
topo sort
☟︎ phf: so in
the above example x y z are distinct patches,
their effect is
to
take files
through states, a, a', a". so if a patch
takes a->a',
then it needs a in genesis. if a patch
takes a'->a"
then it needs genesis and a->a' patch. i was
trying
to indicate
that y and z modify
two separate files, i'm
trying
to see what is
there
that indicates
that state (a' b') is desirable over (a' b)
mircea_popescu: im sorry, i mean : if you didn't need y,
then y should have been z'.
assbot: Logged on 29-01-2016 14:29:59; PeterL: my node is syncing, seemed
to be getting
tons of junk. I restarted it, using -addnode with
the nodes from
the wiki, seems
to be going good again.
mircea_popescu: phf in
the example you gave
the implication is you NEED z.
punkman: phf, my "shortest path" algo is
to recursively find all antecedents of
the patch I want
to press. it's useful.
phf: if i have files a b and patches x y z, x is genesis, y does a->a', z does b->b',
topo sort x z y. if i were
to ask for y press, i also get z, whether i need it or not. conversly if i press z, i don't get y.
mircea_popescu: that came in a few different flavours historically, and amusingly all unaware of
their communion and unadressing it.
mircea_popescu: (or even WHETHER it actually works. hence all
the - equally unexamined and inept - prb derpage about "blessed" chains)
mircea_popescu: it DOES work for bitcoin (in blockchain
terms it's "longest chain"), but altogether it's not clear WHY it works
mircea_popescu: the shortest path
thing is not only premature optimization but also a very mangled sort of heuristic.
mircea_popescu: and so... nobody could have foreseen all
the way
to chad.
mircea_popescu: you can't switch back. once you've decided
to
take
the hardwood floor and light your stove with it, you can't
then decide
to
take stove ash and make a floor.
mircea_popescu: BingoBoingo ironically, consider
the historical aspect. vc's strategic mistake
to NOT submit early and well ensured
that
they don't have a voice at
the one
time it'd have mattered. now
they're getting locked out of bitcoin, which means computing, and what'll mit do, switch back
to scheme because
THAT is what junior devs will need
to get jobs ?
phf: i'm confused about how v presses. i
thought it was along
the shortest path, but it actually does
topological sort and
then works with resulting list. so if my
topo is a b d c f and i ask
to press c,
then i get a b d c?
BingoBoingo: ascii_butugychag: I guess I got
to now. Shiva represents first serious mega-weaponized divergence of
trb from prb
ascii_butugychag: (if you have
the patience, watch
the films, even,
they are interesting in a number of ways incl. historically)
ascii_butugychag: BingoBoingo: once you understand how scheme's
tail call optimization works, you are well on your way
to actually grasping
the
thing
ascii_butugychag: doable yes. but in
the same sense as
the
thing i just did, was doable.
ascii_butugychag: also it will need some means of restarting it if
the repl is hosed
ascii_butugychag: whereas a persistent lisp will need a reconnectable pipe and
this will require surgery inside
tinyscheme's intestines
mircea_popescu: ascii_butugychag>
this will have
to be dealt with << doesn't seem
terribly hard.
ascii_butugychag: adlai:
to me it simply means 'i don't have
to leave my editor and FUCK NO I AM NEVER pasting code into a motherfucking shell 10000000
time a day'
adlai also wanders off
to eat
adlai: to me, and i imagine
to programmers who have done CL work recently, "slime" means a lot of
things
that you do NOT want
adlai: ascii_butugychag: sure, i'm
taking slime-the-~idea~ and
trying
to "take away everything
that doesn't look like an elephant"
adlai: the subset of "all slime functionality", which do seem
to me relevant for shiva, are
the repl (history, presentations),
the fancy-inspector (designed for CLOS objects, might be neutered+mutated deal with C++), and perhaps sldb
adlai: slime is a massive
turd, most of which is
totally irrelevant for shiva (unless bitcoin-os is planned
to serve as yet another example of Greenspun's 10th)