90500+ entries in 0.558s

mircea_popescu: the whole nonsense unpleasantly reminds me of that sf story about
a socialist future where people got "handicaps" to match their intelligence, lest they're smarter than their stupid parents / the runts in the litter.
mircea_popescu:
a) you see who deedbot trusts. that is called the l1. b) you see who l1 trusts. that is called the l2.
a111: Logged on 2016-08-22 13:48 asciilifeform: then again, he argued - imho very successfully - against
a canonical ~tree~, not against www which shows all known vpatches in tree form
mircea_popescu: so it attracts
a certain sort of stupid, not just in the sense of "
a certain sort of stupid people", but actually
a certain sort of stupid OUT OF people, even normally intelligent ones.
mircea_popescu: much like the bloviating bovines / paul biggars of the world perceive bitcoin to be "
a safe space for their creativity" or somesuch, similarly build process.
mircea_popescu: but i think
a large-ish part of hte problem is that build process per se is not actually specified.
a111: Logged on 2016-08-22 03:17 mircea_popescu: and trying to implement
a "better" autoconf, even "by hand", will not result in anything better
a111: Logged on 2016-08-18 22:32 asciilifeform: 'The situation is somewhat akin to
a retarded girlfriend trying to flood your apartment, that not only opens all the faucets and stops all the drains, but also takes the "extremely clever" measure of puncturing the water pipes, so she can then preciously inform you that "turning off the faucets won't help" and you must work with her to somehow create
a raft out of your widescreen TV so as to navigate the marshy terrain that used to b
a111: Logged on 2016-08-22 03:11 phf: i say it's
a chicken and egg problem, because you know you can get vendor versions of posix tools (and then still reduce the available "language" even more, by taking away some features that might be missing in nominally posix sh on some obscure system), but you can't really do anything else, until you established some truths about your environment
a111: Logged on 2016-08-22 03:02 mircea_popescu: basically, seems to me most, or at least
a good chunk of autoconf problems as described by alf come from the fact that it tries to parse, rather than compile. in the abstract sense of these terms.
a111: Logged on 2016-08-22 09:35 jurov: it can stay as
a historical record, but it's currently frozen with #b-
a wot and redoing wot synchronization is exactly such kind of nontrivial change
jurov: it can stay as
a historical record, but it's currently frozen with #b-
a wot and redoing wot synchronization is exactly such kind of nontrivial change
☟︎ diana_coman: I'll search the logs to find phf's pointer and have
a look at it
mircea_popescu: phf once management has put compatibility on your list of targets, you can say that all you want, best done in
a mutter under own beard.
phf: i think at best it could be trimmed down, but i think even that's doable by limiting the number of m4 macros used for includes. or perhaps being very specific with what you want. "i need gcc 4" will necessarily be less messy than "i need
a c compiler", considering that most of the time the program is not ready to deal with wide range of compilers anyway
mircea_popescu: except it will appear to, for
a briefd interval at first. then it will create friction between purists and inclusionists, which is just about as dumb drama as livejournal disputes.
mircea_popescu: and trying to implement
a "better" autoconf, even "by hand", will not result in anything better
☟︎ phf: i say it's
a chicken and egg problem, because you know you can get vendor versions of posix tools (and then still reduce the available "language" even more, by taking away some features that might be missing in nominally posix sh on some obscure system), but you can't really do anything else, until you established some truths about your environment
☟︎ phf: but fwiw even if openbsd, say, will give you
a similar compilation environment, it's not going to give you bash out of the box. it's ksh, so you're back to "least common sh denominator"
phf: neither were designed for programming, sh is rudimentary even by bash standards. on linux sh is linked to bash, but elsewhere you might actually encounter
a barely posix take on it.
phf: there are no arrays. you literally don't get better storage than "text in
a variable". there's very limited math. even expressing something simple "do this n times" is pain.
mircea_popescu: heck, gcc is not much better than
a three pass parser most of the time.
mircea_popescu: basically, seems to me most, or at least
a good chunk of autoconf problems as described by alf come from the fact that it tries to parse, rather than compile. in the abstract sense of these terms.
☟︎ mircea_popescu: and sure, maybe d1 = "need to compile c code" in which case "check for d1" means "do we have
a way to compile c code on this system ? ok, what is it ? and how is it invoked ? ok here we go then :"
mircea_popescu: let's work with an abstract example. suppose there's project X, which can for the purpose of configure be reduced to
a list of n lines, whereby each line produces
a dependency from the list of D1- D5 by the criterion that line# mod 5. so
a sane autoconf will read the whole list, produce
a list reading "d1, d2, d3, d4, d5" and then proceed to check these. once. five fucking checks, five lines of checking.
phf: well, so sh/m4 are dumb. one is
a very primitive programming language (no portable notion of arrays for example, so "read all lines" "create
a list" is not
a straightforward thing), the other one is
a templating language
phf: programming in sh/m4 combination is not
a sane thing
phf: so you either rely on those, or you have
a chicken and egg problem (how do i compile c without knowing how to compile c)
mircea_popescu: i don't even understand why does it parse rather than compile, for instance. seems
a very nutty design decision.
phf: automake is distinctly not magic. it's
a text interpolation macro system. i'm pretty sure there's not even
a tree shaker there (i.e. if you check for gcc, you check for gcc every time something says "i need gcc". only reason it's not done
a dozen of times is because there's also
a check "did i check for gcc already?". never the less all those dozen gcc checks end up in ./configure)
phf: there's also
a dozen of different "opengl"
mircea_popescu: this thing works. and it works in
a very messy environment - gfx drivers,
a raft of dependencies, etc.
phf: it doesn't satisfy our own requirements of fits in head, it goes through
a rube goldberg machine in order to produce
a specific build for
a specific system. "mac os x? fuck you. openbsd? fuck you". it doesn't succeed at own goal of producing bit identical builds.
mircea_popescu: so far the major problem is irc usage ; and gpg/deedbot
a very distant second.
mircea_popescu: actually ~every project of "battlefield" size seems to end up with
a half mb to mb configure script.
mircea_popescu: they hardwired
a bad path in there and i ran outof patience before finding the right magic knob.
mircea_popescu: what it does is exactly this : takes
a situation where "fuck 83% of people" and allows them to install w/e it is you're making.
mircea_popescu: what i had in mind phf is, automake runs on project, keeps track of every bit of itself that it actually called, then inserts
a pruned down version of itself somewhere, so you can ship the software with THAT instead of the whole automake.
phf: mircea_popescu: well, there's the primary file, where you say things "i need
a c compiler for this project"
mircea_popescu: yeah but it doesn't do it by itself. it's just
a macro merge thing neh ?
mircea_popescu: phf i wonder if
a "prune down to THIS codebase" function would be useful.
phf: automake solves
a bootstrapping problem in that common subtrate across all posix systems is "sh", so it's
a tcl-like macro language built on top of shell. biggest problem with it (besides the typical communal retardation) is that nobody actually spends any time trying to understand it, before dismissing it
mircea_popescu: phf yeah. i have
a problem in that i want to impose eulora as standard for republic code ; but nevertheless automake IS
a humongo pile of perl.
phf: trb builds on three systems altogether, and it's still
a pita
a111: Logged on 2016-08-22 01:02 asciilifeform: automake is
a work of evil, first thing i did when sawing off the mpi lib is to nuke it.
phf: there are some clearly meat-y parts, and then there's
a lot of gendered pronounce
phf: i like how they added
a stub for GOST
phf: estream-printf got
a whole blob for testing added including #ifdef TEST and emacs bindings for compile-command to do rapid testing from inside the file. how hacky
phf: that thing looks like an air, so i figured it was junk, internally. but it's quite reasonable. harddrive has
a rapid replacement design, etc.
mircea_popescu: i really wouldn't advise anyone to open
a li-ion shit. why the fuck would you do that, next step is go to africa root through garbage for
a living.
phf: not sure i can source
a custom battery though
a111: Logged on 2016-08-22 00:23 asciilifeform:
http://btcbase.org/log/2016-08-20#1526079 << i watched thread and wondered how long an (allegedly) cunt-enabled ninjashogun would last in the field, vs
a conventional one. seems like we have an answer.
phf: oh also i couldn't figure out how to get intel to work on x11, rather than radeon. it seems like it requires some boot time pokes, which you need hurd to execute (or else figure out how to run
a random code from lilo but before kernel is loaded, etc.)
phf: i'll give it another try. i started using it prematurely, and was basically running around looking for
a powerplug at the point where i really didn't want to deal with shit like that
phf: yeah, i think i figured out how to power down radeon and use intel, which solved some of the issues, but there's also
a bunch of other peripherals, that remain powered-up even not in use, bluetooth, gsm, etc.
phf: i'll boot the laptop in
a few and share the relevant details for interested parties
phf: asciilifeform: there's some other package like laptop-mode maybe, that has
a collection of shell scripts that live on top of apm and actually make do all the sensible things
a111: Logged on 2015-03-04 06:52 asciilifeform: BingoBoingo: does engl. have
a catch-term for 'fatal buggery' ?