Hide Idle (>14 d.) Chans


← 2022-01-26 | 2022-01-28 →
01:31 signpost http://logs.bitdash.io/pest/2022-01-26#1003094 << when in doubt, get older vintage. 2.6.2 worx fine. must've snagged 2.6.4 when carving off individual vpatches
01:31 bitbot Logged on 2022-01-26 22:50:27 signpost[asciilifeform]: farting around with a segfault in the flex build, details here https://developers.redhat.com/blog/2019/04/22/implicit-function-declarations-flexs-use-of-reallocarray
01:31 * signpost rebuilds all, sacrifices virgin chicken
~ 36 minutes ~
02:08 signpost asciilifeform: sometime, perhaps tomorrow, it'd be great to give the v "manifest" thing another look.
02:09 signpost still feels like a kluge to me.
02:10 signpost to restate the problem for the logs, extant v-trons walk backward from a given specified leaf node to genesis, then press forward along that path.
02:11 signpost this is to avoid the fact that two given leaf nodes may both mutate the same "from" hash, and it's not decidable in current spec which takes precedence.
02:16 signpost http://trinque.org/2018/06/02/v-manifest-specification/ << relevant write-up
02:20 signpost questions I think are worth revisiting: 1) do we in practice have any use for multiple press heads? 2) is the lack of v-tron enforcement of manifest contents matching the computed patch flow prone to unnecessary operator error? 3) do we care if we preserve vpatch compat with traditional posix patch?
02:24 signpost (also ignore the GNS mention in the blog post. I was macroexpanding that to a distributed item, not a reboot of DNS with replaced crown)
02:26 * signpost would like operators to have maximal control over which patches they choose to use, rather than an author-dictated ordering, where/if possible
02:29 signpost asciilifeform originally envisioned v pressing all paths that can be pressed, iirc, or in other words, if the operator doesn't want a patch pressed why'd he place it in `patches` dir?
02:30 signpost in that view, having same antecedent in two patches simply fouls the press, and operator is invited to regrind one of them.
02:31 * signpost sees very little problem with this approach, but invites reminders if forgot other problems.
02:37 signpost another approach would involve allowing patches to declare antecedents without requiring that those antecedents be modified.
02:38 signpost perhaps this can even be done today with a no-op hunk. I'll have to try that.
02:40 signpost hm, dropped a line fighting with mega-slow ping. will pick this up tomorrow. cya!
~ 11 hours 59 minutes ~
14:40 u0_a185 test message from cli
14:43 asciilifeform http://logs.bitdash.io/pest/2022-01-27#1003201 << correct, at least for asciilifeform's orig.
14:43 bitbot Logged on 2022-01-27 02:10:30 signpost[asciilifeform|billymg]: to restate the problem for the logs, extant v-trons walk backward from a given specified leaf node to genesis, then press forward along that path.
14:45 asciilifeform http://logs.bitdash.io/pest/2022-01-27#1003204 << orig. v.py took an explicit press head for 1) make explicit what yer intending to do; 2) save some labour of moving patches in/outta the patch dir when pressing variants. (arguably one could dispense with the press head arg, yes.)
14:45 bitbot Logged on 2022-01-27 02:20:39 signpost[asciilifeform]: questions I think are worth revisiting: 1) do we in practice have any use for multiple press heads? 2) is the lack of v-tron enforcement of manifest contents matching the computed patch flow prone to unnecessary operator error? 3) do we care if we preserve vpatch compat with traditional posix patch?
14:48 asciilifeform http://logs.bitdash.io/pest/2022-01-27#1003204 << 2 -- yes; 3 -- asciilifeform likes that vpatches can be pressed 'by hand' if req'd (say, on a n00b's box where no vtron, or as field expedient)
14:48 bitbot Logged on 2022-01-27 02:20:39 signpost[asciilifeform]: questions I think are worth revisiting: 1) do we in practice have any use for multiple press heads? 2) is the lack of v-tron enforcement of manifest contents matching the computed patch flow prone to unnecessary operator error? 3) do we care if we preserve vpatch compat with traditional posix patch?
14:49 asciilifeform http://logs.bitdash.io/pest/2022-01-27#1003210 << you'd have to break the classical patch syntax, and aint clear to asciilifeform what is gained by doing so in this case
14:49 bitbot Logged on 2022-01-27 02:37:41 signpost[asciilifeform]: another approach would involve allowing patches to declare antecedents without requiring that those antecedents be modified.
14:50 asciilifeform ( yes unix patch permits noop hunks, but traditional vtron dun permit identical 'in' and 'out' hashes, because such a pair cannot be placed in order by the tree walker, it could fit in >1 spot)
~ 32 minutes ~
15:23 PeterL http://logs.bitdash.io/pest/2022-01-27#1003202 << wasn't the bigger problem that patches touching different files would get missed in the flow and left out?
15:23 bitbot Logged on 2022-01-27 02:11:47 signpost[billymg|asciilifeform]: this is to avoid the fact that two given leaf nodes may both mutate the same "from" hash, and it's not decidable in current spec which takes precedence.
~ 20 minutes ~
15:43 shinohai testing android logging flush times
15:57 asciilifeform PeterL: orig. yes. was why 'manifests'
15:57 shinohai $vwap
15:57 u0_a185 The 24-Hour VWAP for BTC is $ 36605.63 USD
15:58 * asciilifeform in orig. v did not include any such demand, relied on brains of operators to produce valid fine-grained patches if they feel like.
16:03 signpost PeterL: both are consequences of dependency information not included in the vpatch.
16:04 shinohai $vwap
16:04 u0_a185 The 24-Hour VWAP for BTC is $ 36605.63 USD
16:04 asciilifeform 'manifest' fwiw was orig. a response to an 'own goal' by asciilifeform , where produced a v-valid patch that pressed to an unbuildable item on acct. of not cementing a #include dep.
16:05 signpost yeah, practically speaking I wish to lop of a lot of caked, gnarled hair-and-shit off pentacle after initial release.
16:05 signpost rip the fucking autotools out of it, the perl deps, etc etc.
16:06 signpost might as well consider whether current v is as simple as can be before this, where cutting *unrelated* wads off the patient.
16:06 asciilifeform makes sense, and worth the sweat defo imho
16:08 signpost http://logs.bitdash.io/pest/2022-01-27#1003216 << yeah, not dissuaded of the "press all that can be pressed" notion just yet, and dispense with HEAD, or make HEAD optional arg. at least want to bump into something solid down that line of inquiry.
16:08 bitbot Logged on 2022-01-27 14:45:27 asciilifeform: http://logs.bitdash.io/pest/2022-01-27#1003204 << orig. v.py took an explicit press head for 1) make explicit what yer intending to do; 2) save some labour of moving patches in/outta the patch dir when pressing variants. (arguably one could dispense with the press head arg, yes.)
16:09 signpost http://logs.bitdash.io/pest/2022-01-27#1003218 << agreed, I use also to bootstrap pentacle on a system which doesn't yet have v.
16:09 bitbot Logged on 2022-01-27 14:48:16 asciilifeform: http://logs.bitdash.io/pest/2022-01-27#1003204 << 2 -- yes; 3 -- asciilifeform likes that vpatches can be pressed 'by hand' if req'd (say, on a n00b's box where no vtron, or as field expedient)
16:09 signpost http://logs.bitdash.io/pest/2022-01-27#1003220 << I think this one can be dismissed on grounds of breaking trad. patch, but I included for completeness.
16:09 bitbot Logged on 2022-01-27 14:49:50 asciilifeform: http://logs.bitdash.io/pest/2022-01-27#1003210 << you'd have to break the classical patch syntax, and aint clear to asciilifeform what is gained by doing so in this case
16:10 asciilifeform signpost: it isn't even that it's utterly impossible to improve on unix diff format (or the vpatch glue on top of same); simply that imho there is much value in stable format and particularly that erryone in fact has already support for it outta-the-box
16:10 signpost http://logs.bitdash.io/pest/2022-01-27#1003222 << I'm possibly dense, but wouldn't the >1 spot be disambiguated by the inclusion of other antecedent hashes which do mutate?
16:10 bitbot Logged on 2022-01-27 14:50:49 asciilifeform: ( yes unix patch permits noop hunks, but traditional vtron dun permit identical 'in' and 'out' hashes, because such a pair cannot be placed in order by the tree walker, it could fit in >1 spot)
16:11 signpost but yeah, it brings in "then you have to require that patch includes disambiguating hashes" and that spiders out.
16:11 signpost asciilifeform: yep agreed, I like preserving traditional patch support.
16:11 asciilifeform signpost: in principle yes. asciilifeform was under the impression that you were contemplating a 'per hunks' fine-grained walker in place of the original 'grouped by patch' tho
16:11 signpost busybox patch eats happily, for example.
16:12 asciilifeform aaha
16:14 * asciilifeform many times historically contemplated replacements for trad. diff that would e.g. not spew entire line if you changed 1 char. but it would subtract from human-readability, which is an explicit goal of vtronics.
16:15 asciilifeform that problem is prolly best handled by e.g. colourizers for reading patches a la phf's
16:17 signpost yep, one can go all the way to "aaaa why not AST-level-diff" with that.
16:17 signpost (to which, certainly, but on a better planet or in 300yrs)
16:18 signpost to argue in favor of the manifest, a prime benefit is that the line-of-history intent of the author is made explicit and part of the authored content, rather than keistered away in another representation in e.g. .git/
16:20 asciilifeform aaha, abolishing metadata liquishit was anuther explicit aim of vtronics
16:20 * asciilifeform fucking hates '.git' and similar disk litter shat by heathen versionatrons
16:21 signpost indeed, git's meant to maximize code vomit - by volume - of desk slaves
16:21 shinohai git out!
16:22 signpost I think I'd be satisfied by v telling the operator when manifest diverges from press flow, and I'm still interested in perhaps making head optional.
16:22 signpost though that latter bit isn't urgent.
16:23 asciilifeform 'head optional' is comparatively trivial, simply let it pick the outermost leaf in the flow (and eggog if there's 2+ equal branches there)
16:23 asciilifeform ^ famously in asciilifeform's original v.py the last part was omitted
16:23 signpost a use-case in my mind is keeping a pile of experimental patches without needing to re-grind them.
16:24 signpost but vpatching these by hand is not so much work
16:24 signpost and have to regrind anyway once ready to publish to peers, to add a manifest entry.
16:27 * signpost pretty satisfied actually. will take a run at v from "vtools" barking when manifest is stale.
16:38 PeterL so I was thinking, it seems messy to have things like empty files/folders with just a comment in them (#this file is empty), wouldn't it be better to include a "setup" script that went through and touched all the needed empty files/dirs?
16:44 signpost PeterL in what?
16:45 * signpost dealt with this in pentacle, but considers software that "requires" empty files broken
16:45 signpost I just did a for f in $(find -size 0); do echo '' > $f; done
16:45 signpost (for now)
16:47 PeterL like in blatta, lib/__init__.py
16:49 signpost imho the line of reasoning you're following becomes "autotools"-esque over time
16:49 PeterL is that bad?
16:49 signpost i.e. src that mutates itself in all manners after press
16:50 signpost python has a particularly stupid way of denoting what directories are considered modules
16:50 signpost seems fine that the burden of this falls on python src to have a return in those files.
16:50 PeterL and don't you also need to make some files executable, the setup script could do that too?
16:51 signpost yeah, that is reasonable, and in fact this is what "pentacle" is
16:51 signpost lemme snag you an example build script.
16:52 signpost http://paste.deedbot.org/?id=fZSk
16:53 * signpost has taken the view that these scripts do not belong in blatta's own src, because this is how you end up with a sad piece of software befouled with 1000 systems' compatibility tweaks, which again, straight to autotools hell.
16:54 signpost folks might find this build script reminiscent of gentoo's "ebuild", but simpler.
16:54 * signpost even has a basic use-flag mechanism in there.
16:54 * signpost bbl, appointment
~ 15 minutes ~
17:09 signpost btw that's the build script for gnu make.
~ 3 hours 21 minutes ~
20:31 awt I could release a version of blatta that wouldn't die on GetData packets right now, but unless anyone is in a particular hurry, I'll go ahead and finish up GetData support prior to the next release (currently only sends out handles GetData requests, doesn't know how to handle GetData responses).
20:32 awt Pest supported version will be 0xFC
20:33 awt BTW trying out PyCharm. Pretty nifty.
20:34 whaack awt: not too bad eh?
20:34 whaack view -> active editor -> show whitespaces is nice for checking to make sure your tas/whitespaces are done correctly
20:35 awt whaack: indeed. Would be cool if I could figure out how to get it to do a build where it starts up 3 instances and allows attaching the debugger to them.
20:35 whaack awt: you had a few unused imports, strange whitespaces, etc. that the ide makes very obvious
20:35 whaack awt: there is a part where you can write a script inside it iirc
20:35 awt whaack: yeah helped fix all that
20:36 whaack and you can write a deploy script and hotkey it, it's nicely organized
~ 32 minutes ~
21:09 signpost they have a pretty good deal on buying intellij ultimate these days, includes the py-ide and many others.
21:09 * signpost uses daily
21:09 signpost their php support is even nice when dealing with the shitwad known as wordpress.
~ 30 minutes ~
21:40 whaack signpost: what is the antecedent of 'they' in 'they have a pretty good...'
21:40 whaack are you selling they are selling at a good price?
21:42 signpost yeah, saying the company sells it pretty cheap per year.
21:42 signpost "jetbrains"
21:44 whaack pycharm has a community edition that is free
21:45 whaack but i have paid for jetbrains products before and would happily do it again
21:46 * asciilifeform recalls their java thingie
21:46 * asciilifeform served time in a java mine, used
21:48 signpost I don't recall what they leave out of the free ones, but might be entirely sufficient.
21:50 signpost the refactoring tools are god-tier.
← 2022-01-26 | 2022-01-28 →