#milkymist IRC log for Thursday, 2012-02-09

wpwrakxiangfu: good news: i built softusb with your patches and my riitek rf combo keybaord is still happy00:57
xiangfuwpwrak, great.01:00
kristianpauldamn, there is something wrong with the jtag-pod or..02:55
kristianpauli can flash NOR02:56
kristianpauldetect all OK02:56
kristianpauli see led activity on the pod as soon system boots02:56
kristianpaulbut i dont see nothing with flterm/gtkterm..02:56
kristianpaulno boot usual messages..03:01
kristianpaulah yes i also ran m1 relflash script03:01
Action: kristianpaul brings laptop03:03
wpwrakmaybe gateware got out of sync ?03:03
wpwrakno serial sounds either like a jtag board problem or a uart incompatibility03:04
kristianpauli'm using right now M1's local osc if you meant that03:04
kristianpaulbut the leds are blinking..03:04
wpwrakgateware = the stuff the fpga "runs"03:04
kristianpaulah03:04
wpwrakblinking leds ... what else could you expect of life ? ;-)03:05
kristianpaulno i reflahsed already from xiangfu script03:05
kristianpauli mean leds in the jtag pod03:05
kristianpauli'll remove it and shor rx and tx and see03:05
kristianpauls/shor/short03:06
kristianpaulnice, i have a loop03:08
kristianpaulmay be bad ground between m1 and the jtag-pod?03:08
wpwrakare you sure the M1 is okay ?03:09
kristianpauli'll try other laptop, lets see03:09
wpwraki mean, the sw on it03:09
kristianpauli wasnt03:09
wpwrakif you install the regular firmware, does it render ?03:09
kristianpaulbut i reflashed "factory" image03:09
kristianpaulhavent try render on it since more than six months :-)03:09
wolfspraulfactory :-)03:09
Action: kristianpaul remenber bring monitor too03:10
wpwrakah, no screen. that's a good way for complicating your life :)03:10
kristianpaulmy desk is small, adding screen complicate my life03:11
kristianpaul;)03:11
wpwrakceiling + wires ? :)03:11
kristianpaulnot exactly03:12
kristianpaulis more a problem of where i get space for put screen on plus wiring the vga cable where M1 is located...03:13
wpwrakyup. hang it from the ceiling, above whatever else you have03:15
wpwrakassuming you haven't exercised that option yet, of course03:15
kristianpaulceiling is 3mts over me !03:15
wpwrakor maybe hang it on a wall03:15
kristianpauli usually use a small chair :)03:16
wpwrak(3m) there's a recent invention called "ladder" :)03:16
kristianpaulok..03:16
kristianpauldont have one :(03:16
wpwraksigh03:17
kristianpaulYES i see FB menu03:17
wpwrakcan i tie you shoelaces for you as well ?03:17
kristianpaulfpga is okat then..03:17
wpwrakwhee ! :)03:17
kristianpaulnow lets try another usb port from another computer03:18
kristianpaulshoelaces, no thanks !03:18
kristianpaulladder are expensive, 100usd the chepaer one ..03:19
wpwrakif you live in an apartment building, the "encargado" may have one03:19
wpwrakif you don't, perhaps you have neighbours ...03:20
kristianpaulyes i do03:20
kristianpaulbut lest not talk about social interaction with then right now ;)03:20
kristianpaulok it renders too03:20
wpwrakmaybe indeed a uart problem then03:24
wpwrakperhaps try m1nor'ing these: http://downloads.qi-hardware.com/people/werner/tmp/m1-20120125/03:24
wpwrakthat may not be what you want in the end, but it's at least a set that's consistent within itself03:25
wpwrakand the uart should work :)03:25
kristianpaulit works on my laptop03:26
kristianpauldoes it matter i recently connected another jtag-pod?03:26
wpwraknow you lost me03:27
kristianpaulor there were two pods connected at same time03:27
wpwrakoh yes, that would confuse things :)03:27
wpwrakonly one at a time :)03:27
kristianpaulwhy!!03:27
kristianpauli connected two jtag-pod03:28
wpwrakbecause few if any of the tools know how to deal with > 1 ;-)03:28
kristianpaulall worked fine was last time i remenber..03:28
kristianpaulhmm03:28
wpwrakthey will do something. 50% chance it's what you expect. 50% it ain't :)03:28
kristianpaul:-|03:29
kristianpaulnow i need make back to work the first jtag-pod...03:29
kristianpaulwhat hell happened.. :\03:29
Action: kristianpaul refuses to reboot its dekstop computer03:30
kristianpaulha03:32
kristianpaulindeed flterm wasnt showing the truth, neither screen03:32
kristianpaulcheck this from gtkterm03:32
kristianpaulhttp://pastebin.com/ScfdLkbx03:32
kristianpaulis my m1 booting ;)03:32
kristianpaulahh may be was that linux upgrade i aplied on sunday..03:34
kristianpauloh well lets reboot03:34
Action: kristianpaul dislike reboot computer after update linux03:35
kristianpaulargh, it works now03:38
kristianpaul:)03:39
kristianpauloh  migen is in opencores03:40
Fallenou_nice to see that Lattice guys are reading Milkymist ML08:41
wolfspraulyes but not so nice if you know that it's mostly lip service :-)08:42
Fallenou_lip ?08:42
wolfsprauljust talk08:44
Fallenouoh right, ok :)08:44
wolfspraulbut I had a lot of ideas how to respond to the "remember that it's us who are the *official" maintainer..." line08:45
wolfspraulha ha08:45
wolfspraulyes08:45
wolfspraulI remember that every day :-)08:45
Fallenou"we duly noted your concern and will let you know when the issue will be fixed"08:45
wolfspraulthat must be why it's dead08:45
wolfspraulunmaintained08:45
wolfspraulbuggy08:45
Fallenouahah08:45
wolfspraul...08:45
wolfspraulanyway08:45
Fallenouis this true anyway ? I thought lekernel was the maintainer now08:45
wolfspraulI replied on the list and in private, and I don't expect an answer or any real help.08:45
Fallenouok, really too bad08:46
wolfspraulup in corporatetown they haven't understood some of those foss principles yet08:46
FallenouI was thinking they were starting to notice they have bug and that someone actually use their crap08:46
wolfspraulbut it's enough to claim 'official ownership' once in a while ;-)08:46
wolfspraulfrom the meeting room...08:46
wolfspraulif I would count on them, I would have long left the Milkymist project08:47
wolfspraullet's see what they contribute, Sebastien has had the best results in getting some small things from them, I believe08:47
wolfspraulfor now I remember about 'official'08:47
wolfspraulha ha08:47
Fallenou=)08:47
wolfspraulwe should send them a big shiny badge08:48
wolfspraulmaybe the fact that they are the official maintainer (which they reminded us about) is the reason that there is no progress?09:16
wolfspraulit's probably closer to the truth than anything else - a more responsive maintainer who could deal with issues more timely would make all the difference09:17
FallenouI guess you're right09:29
Fallenoulet's hope for the best : they realize they have work to do and they do it09:29
Fallenoubut don't depend on it09:29
wolfspraulto be fair to them, as a business it is very hard for them to justify just 'improving' something and spending money on engineering09:30
wolfspraulwe sometimes forget what luxury we have in just doing whatever we like ;-)09:31
wolfspraulwhereas a business needs to understand why, when and how much money is coming back when they spend that 1000 USD09:31
wolfspraulr&d is 5-10% of a healthy high-tech business, so if fixing a GCC bug costs 10,000 USD (with all their overhead every little thing they do is very expensive to them), they need to make 100-200k USD back - that's a lot of chips to sell...09:32
Fallenou:/09:33
wolfspraulthat's the reality on the other side, so my expectations are accordingly09:33
Fallenouok got it09:33
rohre09:35
bkerowolfspraul: That's if the GCC bug was the goal, rather than the expected result of some R&D10:09
wolfspraulsorry don't understand10:10
bkeroWhy so many companies are drawn to open source software is that they're not going to pay 10k for initial software licenses.  The gcc patch was contributed as an asidebecause there were no more appropriate tools out there.10:10
wolfsprauldo you think Lattice will get serious about supporting gcc-lm32 ?10:11
wolfspraulhow can we encourage and support them, if at all?10:11
bkerodocumentation, other things to encourage adoption of the platform10:11
bkerowidespread usage10:11
bkerocasual users10:12
wolfspraulsorry I cannot follow you. there are specific bugs right now, how are they going to be fixed?10:13
wolfspraulLattice is the 'official' maintainer in order to ignore bugs?10:13
wolfspraulmaybe it's easier to assume that there is no active maintainer, and then compare whether it's easier to get llvm or gcc fixed. which is essentially what mwalle, sebastien and others are doing.10:14
bkeroThat's a good way to approach it.10:14
wolfspraulwe do need gcc in the long run for Linux, but that just means another task, because waiting for Lattice to respond to bugs assigned to them hasn't been working so far :-)10:14
lekernelwhy not compile Linux with clang?10:15
wolfsprauljust needs trying, but my gut feeling tells me until you have a full rootfs you will have fixed those gcc bugs several times...10:15
lekernelhttp://lwn.net/Articles/411654/10:16
wolfspraulunless Lattice actually does something, their 'reminder' is noise10:16
wpwraktheir "reminder" brought them out of hiding, though10:17
wpwrakmaybe they mistakenly thought everything was well10:17
wpwraki mean, lm32-gcc so far has worked perfectly for me. so it's not10:18
wpwrakbroken in very obvious ways10:18
lekernelwell, as long as you only have C code and use 4.5.x, there are no big problems10:19
wolfspraulthey are probably passing a perfect binary in a .zip file around inside Lattice, and summer 'early friday' time is coming soon as well :-)10:19
lekernelbut C++ or 4.6/4.7 are broken10:20
wolfspraulI think the simple truth is that Lattice has no resources to fix gcc bugs10:21
roh:)10:22
wolfspraulso that's an extra item somewhere on someone's "keep an eye on this" list, but they can win absolutely nothing inside Lattice with that10:22
wolfspraulso it's just kept at the level that it doesn't blow up :-)10:22
wpwrakmaybe the current discussion can help with blowing it up :)10:23
rohi dont think so10:23
roh'hey, there even is gcc 4.x for that platform'10:24
Action: roh has seen much worse in sdks for various platforms10:24
wolfspraulsince they care about 'official' so much, one way to make it blow up is to convince upstream to hand over the 'official' maintainership to someone who meets minimum responsiveness and bug fixing standards10:25
rohthats my point10:25
wolfspraulbut even if that works, it only means all actual work has to happen outside Lattice10:25
rohregularly vendors do not 'maintain' their sdks for thei chips10:25
rohits usual to do this 'one-shot' and only fix major showstoppers for paying customers10:26
wolfspraulyes10:26
rohsometimes even that is more workadound that fixing10:26
wolfsprauland whining foss people are not among them10:26
wolfspraulthe skill there is just to buy time and keep them quiet :-)10:26
rohin my pov i think maintainance level and quality is much better in the foss industry than the closed-world vendor world10:27
rohsimply because most times i get a glimpse behind the curtain ist abyss10:27
roheh its10:27
wpwraklekernel: if C++ is broken, that's clearly bad. but do you run into a lot of things that need > 4.5 ?10:27
rohi dont wonder.. such code i also would not want to show around ;)10:27
wpwrak(i just checked and found that my workstation has 4.4.3, from 2009. for the ben, i even have 4.3.3 or 4.4.2)10:30
rohgcc version 4.4.1 (Ubuntu 4.4.1-4ubuntu9)10:30
rohold stuff10:31
wpwrakseems to do the job :)10:31
lekernelI'm thinking about the long term here10:31
lekernelso you're advocating that obscure, unmaintainable spaghetti code should be kept around?10:31
rohack. we have enough stones on the way to break the wheels we just reinvented. lets not add more if we are not in real trouble10:32
wpwrakthe current release is 4.6.2. so if they're at 4.5.3, it means they haven't kept of for maybe half a year. that's not catastrophic.10:32
rohlekernel: as long as we do not have a drop-in replacement which doesnt cost uncalculatable much developer time...10:32
wpwraklekernel: gcc maintainability seems to be a non-issue for almost anyone who even knows what gcc is :)10:33
lekernelhaha, with this attitude we'd still be writing COBOL :)10:34
rohgcc is the evil mother in law of linux10:34
rohyou cant live without it, but you cant really get to like her (if you arent somebody with an evil grin yourself)10:35
wpwraklekernel: so unless you're looking for a new hobby and have becoming the official maintainer for lm32 gcc in mind for it, it shouldn't matter to you. provided that lm32 gcc is being maintained.10:35
lekernelroh: you know, clang can compile a linux kernel ...10:35
wpwraka lot of people seem to get along with gcc well enough10:35
rohwpwrak: i always thought maintainership in gcc is decided by who ever lost the poker game with the maintainer last..10:35
wpwraki mean, gcc developers10:35
wolfspraulwpwrak: c++ is broken10:36
rohlekernel: clang isnt avail in the same quality for all platforms yet.10:36
wolfsprauland I don't think you can consider lm32 gcc maintained, it's fake maintained, which is worse10:36
lekerneltrue, and I just gave it a try for LM3210:36
lekernelactually most of it works10:36
wpwrakwolfspraul: (C++) yes, that's something that needs fixing in the long run. once we have linux, we'll also be exposed to C++ code10:37
wolfspraulI agree that it seems the gcc-lm32 situation does not seem bad or in a state of emergency10:38
rohlekernel: well.. if you can get it to compile all you need..10:38
wolfspraulbut it is sad that Lattice pretends to maintain it, but actually ignores everything. that is demotivating to people that are actively running into bugs, so lattice should not be surprised about what they do next...10:38
wpwrakwolfspraul: (fake maintained) perhaps it was. now lattice publicly stated it isn't. i would view this as an opportunity to solve any communication problems that may have been causal in the lack of support10:38
wolfspraulthey have done this before, what you see is the fake10:38
wolfspraulas mwalle also said on the list, in more drastic terms :-) (empty promises)10:39
wolfspraulall they care about is being the 'official maintainer'10:39
rohbusiness as usual to me10:39
wolfspraulcat inmail > /dev/null10:39
wpwrakhave they spoken up in public about the gcc situation before ?10:39
wolfspraulpublic/private mails not so sure10:41
wolfspraulthe situation is like this for a while10:41
wpwraki would attach some significance to this being a public statement. much harder to go back on that10:41
wolfspraulyou can email bugs to that @latticesemi.com address, and nothing will happen :-)10:41
wolfspraultry try10:41
wolfspraulI say: fake10:41
lekernelquite funny how you advocate GCC but complain about the Xilinx tools :-) they're about at the same level for me10:42
wpwrakyes, the "corporate bug desk" doesn't sound very encouraging10:42
wolfspraulmy 'thanks for offering help' is slightly poisoned, because I have a history of mails towards Lattice (including Jonathan I believe) which are being 100% ignored10:42
wolfspraulI will just continue like this :-)10:43
wolfspraulothers have gotten at least empty promises back (which then were not held)10:43
wpwraklekernel: i wouldn't worry about code quality in gcc. my main concern would be algorithmic complexity. if you want to understand a modern compiler, you have to be up to date on your algorithms. and there's a bunch of them that aren't trivial.10:43
wolfspraulthe good news is that we don't depend on Lattice, so they either move or not, whatever. they can only embarass themselves.10:43
wolfspraula bit sad considering that they opened-sourced the nice lm32 core...10:44
wpwraklekernel: i think for anyone who makes the effort of actually being able to understand what's going on in the compiler, code ugliness is probably a minor concern.10:44
wpwraklekernel: and yes, i know how ugly GNU code can be ;-)10:44
wpwrakwolfspraul: do you know what role this Johnathan has ? management ? developer ? PR ?10:45
lekernel...and about making that effort, LLVM has quite some documentation, while GCC doesn't have any (according to rms policy until recently)10:46
wolfspraulhe never replied to me, I don't know10:46
wpwraki think out objective with gcc should be to get lattice to take care of it. help them where we can, but not try to add large portions of a compiler to the list of things we maintain, be it officially or de facto.10:56
wpwrakif lattice are unwilling or unable to do it, that's of course a problem. but i wouldn't take that for granted just yet.10:57
wpwrakalso, the milkymist project is growing. this should also increase its importance for lattice.10:57
lekernel...and perhaps the resources that we can use to make GCC join COBOL in the museum of computing aberrations :-)10:58
wpwrakwolfspraul: and they aren't completely unresponsive. remember the odd licensing terms that they fixed. so somebody there is paying attention.10:59
wpwraklekernel: gcc gets rejuvenate every once in a while. i wouldn't worry too much about it. rumor has it that its internals were much uglier a few years ago :)10:59
lekernelbut yeah, from a more pragmatic pov I agree with you11:00
lekernelotoh if JP Bonn is able to continue his nice LLVM work ...11:01
wpwrakgoogle for "lm32 core" and milkymist comes up as #2. very good. that's something lattice management should find relevant when considering the importance of keeping us happy about architecture support ;-)11:03
lekernelas I understand it, this is for an automatic "adaptative CPU" venture11:03
lekernelof some sort... I don't know all the details11:03
wpwrak(llvm work) i think it's great to have an alternative. not only as a backup but also to provide competition to the leader.11:05
lekernelbtw, what can gforth be good for? the freebsd bootloader uses it?11:07
wpwraki'm a bit puzzled about the number of forth enthusiasta we have in the qi-hw universe. that's the first place i came across any at all in the last 20+ years :)11:09
wpwrakforth comes more from an age where you know every byte of your SRAM on a first name basis, where your PC had BASIC in its ROM, and where compilers were something very expensive only professionals used11:11
wpwraks/know/knew/11:11
wpwrakit's a cute concept, though. very minimalistic.11:16
kristianpaulAnd thats good milkymist still allo you to know every byte of your SRAM :)11:38
kristianpauleven more if you embeded it ! :-)11:38
cladamwwpwrak, lekernel R143 yes is still open issue on dissipation. just saw email. ;-) so we are considering to do like two 1206 for reaching 50 Ohm in parallel confirmly ?12:53
cladamwhmm wpwrak just replied. ;-)12:55
wpwrakyeah. i don't know whether heat would be an issue there12:57
cladamw(HPD pin) wpwrak, lekernel  btw, I added Q4 MOSFET for voltage level shift from 5V to 3.3V in latest sch. could you see that when you are available ?12:58
cladamwhttp://downloads.qi-hardware.com/hardware/milkymist_one/sch/tmp/MILKYMISTONE.pdf12:58
cladamwwpwrak, thanks about replying my question about ledm currents. ;-)13:03
lekernelcladamw: you do not need the bidirectional shifter circuit here13:03
lekernela simple common-source circuit is enough13:04
lekernelhttp://en.wikipedia.org/wiki/File:N-channel_JFET_common_source_degeneration.svg13:04
lekernelwithout Rs13:04
lekernelso that would be http://en.wikipedia.org/wiki/File:N-channel_JFET_common_source.svg13:05
lekerneland with a pull down13:05
lekernelon the gate13:05
cladamwVout is inverted to Vin ? Should they be not inverted ?13:12
wpwraksoftware can handle that ;-)13:14
lekernelthe fpga will take care of that13:14
cladamwI was thought PHD needs un-inverted though, no ? or you meant V+ to J17.HPD, and Q4.drain to fpga pin ?13:14
cladamwaha...s/w handles.13:15
wpwrakor, in our case, even gateware ;-)13:15
cladamwso i edit again now. ;-)13:16
wpwraka 1/4 W through-hole resistor heats to about 75 C at 107 mA, according to my not very reliable thermometer13:19
cladamwwait... i confusing again. so Vin is HPD pin, and Vout is to fpga pin, no Rs, and R232 keeps 4.7K, Q4.gate say using like a 1K pulldown ? correct ?13:19
wpwrakhas a bit of heat dissipation through the leads, though. and i suppose into the thermocouple.13:20
lekernelcladamw: use a 10K pulldown and you can connect the drain directly to the FPGA without a pull up - we can enable the internal pull up of the FPGA13:21
wpwraklow = not connected, right ? so R232 being a pull-up doesn't make sense13:21
wpwrakand yes, 10 kOhm should be plenty :)13:22
cladamwlekernel, wpwrak okay, edit it now. :)13:24
cladamwpls reload latest link again. :)13:34
cladamwwpwrak, lekernel http://downloads.qi-hardware.com/hardware/milkymist_one/sch/tmp/MILKYMISTONE.pdf13:39
larschm... if you brick your milkymist-one will it become a milkymi-stone?13:41
Thihi....13:42
wolfspraulThihi: hi :-)13:49
wolfspraultoo bad I have to logout, too tired. will read the backlog.13:49
ThihiHi.13:52
wolfspraulwe still don't have an official update you can flash onto your m1 ;-)14:01
wolfspraulbut good things are coming, if you are patient...14:01
wolfspraullike support for usb-midi controllers, images, etc.14:01
wolfspraulwhat's your impression so far? is anything publishable right now?14:02
lekernelimages are there already wolfspraul ...14:19
Fallenou14:46 < larsc> hm... if you brick your milkymist-one will it become a milkymi-stone?  < nice one =)15:03
kristianpaullol15:10
Thihiwolfspraul, well, my impression is that MM1 is way too hard to use for most people. But it's a nice kit for people who are into tinkering and coding and bright lights and good parties, I guess.15:12
kristianpaulThihi: what you think is too hard?15:14
lekernelThihi: what do you compare it to? have you tried, e.g. modul8, resolume, processing, puredata, ...?15:16
Thihikristianpaul, well, for instance, the ui is tedious way beyond what most people would consider accessible in any way. Yeah, anyone who really reads everything they see on a screen and can analyze what they see learn to use it, but in a day and age where people are used to really easy and simple uis, the one in mm1 is obscenely nerdy. I have no problem with it, but "I wouldn't make my friends use it", I guess.15:31
FallenouI think you say this because it has windows and things that looks like a traditional windowed window manager (gnome/kde/windows 7) but looks like an old one15:34
Fallenouif you remove windows I think you would not say that15:34
Fallenoua totally different UI with no windows15:35
Fallenousay for example just a full screen app with buttons etc15:35
Fallenouand you cannot compare it to gnome/kde/windows 7 and it is still usable15:35
Fallenoubut you don't think it's nerdy15:35
lekernelThihi: ever seen puredata? http://crab.files.wordpress.com/2008/05/regen-pd.png15:36
ThihiWell, there is certainly nothing wrong with the hardware itself, and it does produce nice enough visuals, so yeah, with a different ui it would be much more usable :)15:36
FallenouI'm pretty convinced that FN would look really less "old style UI"/nerdy without any windows and just a full screen app15:37
lekernelbased on good ol' Tk unchanged from 1991 http://en.wikipedia.org/wiki/Tk_(framework)15:37
ThihiAnd then a lot of the ui stuff seems counterintuitive for people I've dabbled with it. Like the populating of patches to match keys is very tedious to do by hand, but autopopulating just sucks.15:38
Thihilekernel, and I'm not trying to compare it with anything now, I'm just talking about the level of accessibility. The concept and hardware of mm1 seem completely sound to me.15:39
FallenouI think a few things in the UI are "first versions" which can be improved15:39
ThihiYeah15:39
Fallenouusually in a project the feature is just developped, the UI is done quickly to make the new feature available15:40
Fallenoubut often the UI is not perfect the first time15:40
ThihiI'm sure the product will get better as time progresses.15:40
Fallenouit's only there to show the feature15:40
Fallenouso there is always room for improvement15:40
ThihiBut wolfspraul asked me what I think of it now.15:40
Fallenouok15:40
kristianpaulThihi: what guy you imagine? can you  provide a sketch for a friendly gui?15:56
kristianpaulor a reference15:56
wpwraklekernel: hmm, 32 bit word collision probability ... collision between what and what ? i.e., could it be subject to the birthday paradox ?18:31
lekernelit's just a magic word to check that you're indeed reading a wishbone bus description18:49
lekernelso no birthday paradox18:49
wpwrakah, good18:53
wpwrakit's funny, when reading the presentation, all of a sudden you realize how simple it all could be if you'd drop "Shouldn't leave proprietary blocks out in the cold" and "Should not force the use of external sources for metadata"18:55
wpwraki translate the latter to "support closed source drivers"18:56
larscfor me the whole stuff just reads like 'distributed devicetree'18:57
wpwrakyeah18:58
larscand the distributed is only in there because they want to support proprietary blocks19:02
wpwrak"playing nice" with proprietary stuff has a non-negligible cost19:05
wpwraklekernel: btw, one thing that bothers me with llvm is that this project seems to depend quite a bit on Apple. doesn't that worry you at all ?19:06
whitequarkwpwrak: oh please. don't go the gcc way.19:20
wpwrakgcc's been a reliable companion for me for the last ~25 years. you'll have a hard time convincing me it's bad :)19:25
wpwrakit has also suffered and survived a number of crises in maintainership. community support is very strong around it.19:26
wpwrakof course, things are more difficult with you're on a fringe platform like lm32. but you'll have that problem everywhere.19:27
wpwraks/with/when/19:27
whitequarkI mean the part about gcc not supporting plugins because it would make it easy to use for proprietary developers.19:28
whitequarkllvm's main strength is modularity, and making a modular compiler is the one way to make a good compiler19:28
whitequarkgenerally, you don't make a good (technically) project if your goal is to make something free19:29
whitequarkyou make a good project if your goal is to make something awesome.19:29
whitequarkif your goal, as a milkymist dev, is to make a free SoC, you will fail19:30
whitequarkif you want to make a good _and_ free SoC, through, you have some chances to succeed19:30
whitequarkopen-source exists not because of some stallman-like shit or whatever. OSS and FOSS exist because knowledge wants to be free19:31
whitequarkit's more efficient this way19:31
whitequarkscience is basically open-source for centuries19:31
wpwrakusually the two go hand in hand. those who only talk about freedom but have no idea about technical quality rarely get beyond hello.c :)19:31
whitequarkwhen they do, they make arcane stuff like gcc.19:31
whitequarkif it would die tomorrow, I won't drop a single tear19:31
whitequarkthe only way to make a good product is to make a product which is good for everyone19:35
whitequarknot my making a product which is bad for whoever wants to use it in a proprietary way19:36
whitequarkthe same applies to GPL vs. MIT19:36
wpwraki'll remind my butcher of that the next time i buy meat for a barbecue ;-)19:36
whitequarker19:37
whitequarkwhat? :)19:37
wpwrakgood for everyone -> the meat must be acceptable for vegans too19:37
larscyou'd want to make a good product for your target group19:37
larscnot for everybody19:37
wpwrakotherwise it's not a good product, according to whitequark19:37
whitequarkwpwrak: don't nitpick on details. you understand what do I mean19:38
whitequarkmeat isn't deliberately made bad for vegans19:38
wpwraklarsc: precisely. and the target group isn't necessarily self-selecting.19:38
whitequarkhm19:40
whitequarknow that we speak about it. what's the license of M1 SoC?19:40
larscgpl19:43
wpwrakit's a mix of GPLv3, 3-clause BSD, LSCOSLA, probably more19:45
wpwrakmost of the stuff we do is GPLv319:45
whitequarkLSCOSLA?19:55
wpwrakLattice Semiconductor Corporation Open Source License Agreement19:55
lekernelwpwrak: I don't know which is worse, Apple or RMS23:45
wpwrak;-)23:46
lekernelLLVM is actually more independent, since its code is more readable and documented and therefore relies less on the obscure knowledge of some elite23:46
wpwrakhave you heard of EGCS ? that was a gcc fork when development there stalled. it later became the official gcc.23:47
lekernelyes, I have23:48
wpwrakso there's provably enough of that "obscure knowledge" around that the community can work around maintenance issues23:48
wpwrakand again, i think the main difficulty when working on a compiler is understanding enough about compilers in the first place23:48
wolfspraulah, llvm still :-)23:49
wpwrakyeah :)23:49
wolfspraulas expected there was no further response from Jonathan on the list or to (my) private mail23:49
wolfspraullife as usual at Lattice Corp :-)23:50
wpwrakit's a bit sad that lekernel can't control his anger better. i don't think attacking lattice will increase their willingness to provide better support23:50
wolfspraulI agree, of course23:50
wolfspraulI am sad that we take from Lattice (the lm32 core), but don't find smarter ways to give back23:50
wolfspraullike promoting their chips or services or whatever. or understanding their lm32 strategy, their gcc support strategy, and so on23:51
wolfspraulbut amazingly, except for rare lip service mails, there's only silence...23:51
wolfspraulit's just unbelievable that we cannot have a simple few-line exchange with Lattice about their real resource commitments to gcc, or how to setup a better maintenance and bugfixing workflow23:51
wolfspraulbut it seems we cannot...23:52
wpwrakthey are kinda obscure. i wonder what keeps them in business23:52
wolfspraulmaybe some govt contracts :-)23:52
wolfspraulthere's a lot of money to be made in many ways23:52
wolfspraulI just finally want to find someone to have an intelligent short conversation with.23:53
wolfsprauleven if it's 100% negative, then at least that's clear23:53
wpwrakyes. if it's a secrecy-heavy domain, like military, then that could explain some of the communication problems23:53
wolfspraulif you read Jonathan's mail carefully, it's written as if to please some higher-up23:54
wpwrakmaybe you have to find an FPGA conference and try to arrange a meeting there. some people function only face to face.23:54
wpwrakcould be23:54
lekernelwpwrak: where did I "attack" lattice? I think my wasn't even the most aggressive23:55
wolfspraulas if he wants to repeat how it should be (and is actually not), so that he can tick off a checklist and point to his own mails in defense later (internally)23:55
lekernelmy mail23:55
wolfspraullekernel: have you had back and forth communication with someone from Lattice?23:56
wolfspraulwith Jonathan?23:56
wolfspraulis he an engineer? customer support? engineering manager? other?23:56
wpwraklekernel: "What about Lattice funding JP so he can more easily continue his nice LLVM work?" wasn't nice. they just said they're standing behind gcc, which you thus summarily reject.23:56
wolfsprauluntil we understand more about *their* business, and *their* needs, we will cluelessly guess around23:58
wpwrakthe other negative responses were about past frustrations. that's a different category. it leaves open a path for better cooperation in the future.23:58
wolfsprauland unfortunately they don't tell us, which is something I rarely see23:58
wolfspraulso yeah, maybe they have a 10-year govt contract on whatever project, and the 'outward' functions die out a little over the years :-)23:58
wpwrakone important thing in negotiation is to always leave your opponents golden bridges to go where you want them to go :)23:58
--- Fri Feb 10 201200:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!