**** BEGIN LOGGING AT Sun Aug 28 02:59:58 2016
Aug 28 09:03:37 <alfatau> zmatt: hi! "<zmatt> especially since I'm just talking to myself apparently" -> I'm really sorry, yesterday I wasn't able to continue chatting due to force majeure (boss). Anyway I really appreciate your hints and your help and I've just read all your last messages.
Aug 28 09:06:16 <alfatau> zmatt: anyway, I'll check your suggestions tomorrow, so just thank you again for your attention,
Aug 28 09:35:35 <KotH> alfatau: reading the backlog, i think you are using the wrong tool
Aug 28 09:36:01 <KotH> alfatau: if you'd use an cortex-m4, then getting to <80us for udb2spi would be a piece of cake
Aug 28 09:36:08 <KotH> alfatau: probably even <10us
Aug 28 09:37:35 <veremit> KotH: if we had a dollar/euro/pick-currency for every time -that- happened .... :]
Aug 28 09:38:02 <KotH> veremit: and we'd be poor
Aug 28 09:38:30 <KotH> veremit: dont forget, most of us make money with customer stupidity
Aug 28 09:38:32 <veremit> still?! -sigh-
Aug 28 09:38:45 <veremit> hmm .. fair point
Aug 28 09:39:08 <veremit> some do :] mostly a step away from that nightmare
Aug 28 09:39:46 <KotH> jupp
Aug 28 09:39:55 * KotH left that game a while ago for greener pastures
Aug 28 09:40:11 <veremit> wise man
Aug 28 09:40:19 <KotH> probably not
Aug 28 09:40:25 <KotH> i'm a lot poorer now
Aug 28 09:40:37 <veremit> lol . well .. pick ya poison :P
Aug 28 09:40:41 <KotH> have barely enough money to pay food and rent
Aug 28 09:40:47 <KotH> but then, i travel also much more
Aug 28 09:41:22 * tbr glares at KotH, lucky bastard
Aug 28 09:41:30 <veremit> you musta made a mistake there somewhere then ..
Aug 28 09:41:59 <veremit> travelling is nice .. but so is coming home :)
Aug 28 09:43:23 <KotH> tbr: nothing to do with luck
Aug 28 09:43:36 <KotH> tbr: just need to punch the right guy at the right time
Aug 28 09:43:55 <tbr> mhm
Aug 28 09:43:58 <veremit> ^ +1
Aug 28 09:44:07 <KotH> veremit: no mistake. conscious decision
Aug 28 09:44:16 <KotH> veremit: stopped working and became poor student
Aug 28 09:44:28 <veremit> mixed blessing that
Aug 28 09:44:42 <KotH> at least i'm not getting more gray hair
Aug 28 09:44:51 <veremit> but I'm getting ready to punch a guy ;)
Aug 28 11:47:55 <alfatau> KotH: so you think 80usec is a reasonable deadline for udp2spi?
Aug 28 11:50:33 <alfatau> KotH: so why are you saying I'm using the wrong tool? I want to use linux kernel to have an already implemented tcp/ip stack. Where are you pointing your finger?
Aug 28 11:51:37 <alfatau> KotH: ah, I just noticed you mentioned Cortex-M4
Aug 28 14:38:00 <zmatt> KotH: regardless of processor, using 100BASE-TX ethernet here is a problem
Aug 28 15:04:24 <kodur_k> Please suggest some IoT cloud platforms. Essentially what i want to do is: there is a BBB connected to internet, i could make a server and send commands to it, but i want to do it remotely, as i wont have direct IP access and i don't wanna use dyndns. can i do it with these cloud platforms?
Aug 28 15:12:20 <zmatt> you can do it if you have the ability/rights to run some sort of central server / proxy / whatever
Aug 28 15:15:25 <kodur_k> okay, can you please suggest such cloud platforms which would help me doing it?
Aug 28 22:06:21 <KotH> zmatt: well, he wrote _after_ reception, so the time needed to receive the packet has been substracted
Aug 28 22:57:39 <zmatt> KotH: well I'm not worried about the time to receive the packet itself (though I'm not sure what latency CPSW adds also) but rather other packets queued in front, since he mentioned there would still be other traffic like ssh
Aug 28 23:02:49 <KotH> zmatt: true that
Aug 29 00:55:55 <zmatt> holy fuck I should have installed distcc waaaaaay sooner
Aug 29 00:57:03 <zmatt> some test compile of a library went from 2 minutes on a beaglebone to 10 seconds
Aug 29 00:58:31 <veremit> you fool .. hahahaha
Aug 29 00:59:00 <zmatt> (distcc-pump, one quad-core Xeon E5420 @ 2.5GHz)
Aug 29 00:59:15 <veremit> I wouldn't compile anything > about 10k on an ARM
Aug 29 00:59:33 <veremit> that's why I don't (yet) run gentoo on any ..
Aug 29 00:59:51 <veremit> although I do have cross-compilers built .. just gotta fix up the symlinks for distcc
Aug 29 00:59:58 <zmatt> well I cross-compiled stuff often enough, but dunno... I guess I was vaguely aware of distcc, but in my mind it was something complicated to setup and I never even realized it could work transparently for distcc
Aug 29 01:00:06 <zmatt> I didn't have to fix up any symlinks
Aug 29 01:00:08 <veremit> no its quite easy
Aug 29 01:00:15 <zmatt> the debian package made those already
Aug 29 01:00:20 <veremit> ah debian probably does it all for you
Aug 29 01:00:45 <zmatt> I did fix up the distcc-pump script a bit to actually put the dir containing those symlinks in PATH :P
Aug 29 01:00:59 <zmatt> little oversight
Aug 29 01:01:00 <veremit> well .. if you want to use pump mode .. that might help ..
Aug 29 01:01:38 <zmatt> yes, and if you setup your hosts file for pump mode then you'll get a stream of warnings if you compile without pump
Aug 29 01:01:49 <veremit> but yeah I been using distcc with x64 fir years
Aug 29 01:01:52 <zmatt> also, aforementioned lib was 40 seconds without pump
Aug 29 01:01:58 <zmatt> nowait
Aug 29 01:02:02 <zmatt> 28 seconds
Aug 29 01:03:10 <zmatt> pump still needs to be modernized a bit... a systemd user service or something
Aug 29 01:03:15 <zmatt> socket-activated
Aug 29 01:07:35 <zmatt> I wonder if pump has any security whatsoever or whether you're basically serving your whole disk on the network (read-only I'd hope, but still)
Aug 29 01:26:03 <zmatt> veremit: what happens if there's a gcc version mismatch between server and client
Aug 29 01:26:06 <zmatt> ?
Aug 29 01:26:15 <veremit> trouble :D
Aug 29 01:26:48 <zmatt> guessed as much... so surely it'll detect that and complain? .. right?
Aug 29 01:26:49 <veremit> major and minor should match .. ie. 4.9.x are ok .. 4.8.x etc
Aug 29 01:26:53 <veremit> no...
Aug 29 01:26:57 <veremit> it just builds crap lol
Aug 29 01:27:02 <zmatt> lovely
Aug 29 01:27:10 <veremit> yup .. it has no way of knowing
Aug 29 01:27:25 <veremit> because it doesn't pass version informatino over .. just the 'flavour' of compiler
Aug 29 01:28:01 <veremit> its the one limitation .. on gentoo we have a utility called 'gcc-config' so that you can select the active gcc if you have multiple installed
Aug 29 01:28:04 <zmatt> and it's impossible to fix that braindamage in the protocol because.... ?
Aug 29 01:28:13 <veremit> its Google.
Aug 29 01:28:20 <veremit> iirc.
Aug 29 01:28:22 <zmatt> I mean they managed to add pump, surely that's a lot more invasive
Aug 29 01:28:44 <veremit> naa .. I suspect they just changed the flow of data slightly
Aug 29 01:28:54 <veremit> but .. it can'y br impoddinlr
Aug 29 01:28:56 <veremit> wtf
Aug 29 01:29:05 <veremit> It can't be impossible*
Aug 29 01:29:09 <veremit> keyboard registration error
Aug 29 01:29:24 <zmatt> argh, what's with gmane.org being down
Aug 29 01:29:31 <zmatt> soooo many broken links on the interwebs now
Aug 29 01:31:34 <zmatt> ah some people had fun problems: the need to change which build servers are used depending on which compiler (e.g. native or cross) is used
Aug 29 01:32:16 <zmatt> and no way to just set it once at the start of compiling since the thing being compiled was a cross-toolchain
Aug 29 01:35:34 <veremit> yikes
Aug 29 01:35:42 <veremit> yes, you need to watch what you're doing :)
Aug 29 01:35:59 <zmatt> though really it's silly noone has fixed this yet
Aug 29 01:36:01 <veremit> but its not all that hard once you get the idea
Aug 29 01:36:16 <veremit> yea its a bit of a half-assed utility
Aug 29 01:36:25 <zmatt> typical case of "works well enough that everybody is too lazy to fix it"
Aug 29 01:36:29 <veremit> but I guess it .. right
Aug 29 01:36:32 <veremit> exactly
Aug 29 01:36:40 <zmatt> (including me)
Aug 29 01:36:45 <veremit> needs some hefty patching :P
Aug 29 01:36:52 <zmatt> not really
Aug 29 01:37:21 <zmatt> the zeroconf mechanism (if it works, it didn't for me iirc) could advertise the toolchains installed (incl version number)
Aug 29 01:37:30 <veremit> eww zeroconf ..
Aug 29 01:37:38 <veremit> why add the bloat
Aug 29 01:37:44 <zmatt> yes, not having to configure shit, how terrible
Aug 29 01:37:57 <zmatt> well avahi is already on all machines anyway
Aug 29 01:38:02 <zmatt> (all our machines)
Aug 29 01:38:20 <veremit> well .. from the "do-it-all-for-me-d" solujtion .. why aren't you just running windoze already?!
Aug 29 01:38:36 <zmatt> avahi has been a lifesaver often enough for me
Aug 29 01:38:53 <veremit> only ever caused problems in my experience
Aug 29 01:39:04 <zmatt> large network, dhcp, dns server that's occasionally amnesiac
Aug 29 01:39:07 <veremit> 'discovery' never works
Aug 29 01:39:20 <zmatt> well not large network, but large enough you don't want to search manually
Aug 29 01:39:53 <zmatt> don't get me wrong, avahi is a crappy implementation
Aug 29 01:40:03 <veremit> its an apple-derived functino no?
Aug 29 01:40:05 <zmatt> even though I thought mDNS was pretty straightforward
Aug 29 01:40:09 <zmatt> it's RFC
Aug 29 01:40:27 <zmatt> https://tools.ietf.org/html/rfc6762
Aug 29 01:40:34 <zmatt> but yeah originally from apple
Aug 29 01:40:50 <zmatt> microsoft has the similar but incompatible LLMNR
Aug 29 01:41:21 <veremit> gotta reinvent the wheel ofc.
Aug 29 01:41:31 <zmatt> I wonder what the state of both is in systemd... I know a while ago it was already a (very basic, hostname only) LLMNR responder
Aug 29 01:41:50 <zmatt> systemd-networkd I mean
Aug 29 01:42:00 <zmatt> or -resolved I guess maybe
Aug 29 01:42:28 <veremit> just download systemd-fixitformenaod
Aug 29 01:43:01 <veremit> it'll surely spawn a gettherightd-downloader to get a package which solves everything :)
Aug 29 01:43:08 <zmatt> but really, with mDNS and IPv6 you could have networking that *just works*
Aug 29 01:43:48 <veremit> about to open the ipv6 can o worms
Aug 29 01:43:59 <zmatt> no router needed for local networking, no (re)configuration needed in case a router does show up, no "what's your IP?"
Aug 29 01:44:05 <veremit> could be an interesting journey :)
Aug 29 01:45:27 <zmatt> we might finally have networking that, for normal users, works out of the box just like AppleTalk did over 30 years ago
Aug 29 01:46:55 <zmatt> (except lacking a consistent UI and with many legacy apps that will manage to fail in this environment because it wasn't like this from the start)
Aug 29 01:48:11 <zmatt> when using AppleTalk not even developers ever used or cared about things like network IDs or node IDs, even though they did of course exist under the hood
Aug 29 01:51:07 <zmatt> ohh, looks like more complete LLMNR and mDNS support has landed in resolved
Aug 29 01:58:16 <zmatt> the only thing I don't like is that it introduced dnssec
Aug 29 01:59:25 <zmatt> that's really one of the most pointless developments in internet "security" that provides no benefit, slows things down, and may even open new vulnerabilities
Aug 29 02:07:33 <zmatt> oh LLMNR is actually an RFC too, but an "Informational" one rather than "Proposed Standard"
Aug 29 02:12:47 <veremit> lol
Aug 29 02:37:44 <kremlin> zmatt: attempting PRU/L3 test today
**** ENDING LOGGING AT Mon Aug 29 02:59:58 2016