#qi-hardware IRC log for Tuesday, 2013-02-12

xiangfulindi-: Hi you patch on toped have merged in http://code.google.com/p/toped/source/detail?r=236312:07
lindi-xiangfu: that was fast :)12:07
xiangfufor 2) Vcs-Git: I think I will just remove that lines. that is simple and fast for me. 12:08
lindi-xiangfu: that's totally ok12:08
wpwrakxiangfu: btw, a while ago, you made a few changes to integrate atben/atusb in more recent openwrt kernels. in which tree are these changes ?12:11
xiangfufor the 3) lintian complains. I always just ingore those warnings. since I found that the toped libs are not fit to party executables.12:11
xiangfuwpwrak: let me check. 12:11
lindi-xiangfu: party executables?12:12
xiangfulindi-: I mean those libs are not fit to link12:13
lindi-xiangfu: yep, so maybe they shouldn't be in that directory?12:13
lindi-xiangfu: maybe /usr/lib/toped/ ?12:13
lindi-xiangfu: with    find  /usr/lib -type f -mindepth  1 -name "*.so"     you can see a lot of such private libraries12:15
xiangfuwpwrak: a small change is http://pastebin.com/hyUBCgnr for fix atusb cannot build as modular12:16
xiangfuwpwrak: other changes I think some functions rename for sync with linux-
xiangfulindi-: is there a easy way change those .so files to /usr/lib/toped without touch the upstream code?12:22
lindi-xiangfu: ./configure --libdir might do the trick if you are lucky12:23
xiangfulindi-: thansk. I will try that.12:23
wpwrakthanks ! 12:24
xiangfulindi-: are you a DD?12:24
lindi-xiangfu: yep12:24
lindi-xiangfu: http://codesearch.debian.net/search?q=--libdir%3D%2Fusr%2Flib%2F[a-z]12:24
xiangfuI can apply that after several month later. :(12:25
lindi-xiangfu: the nagios and lighttpd examples might match your scenario ^12:26
xiangfuyes. I just trigger the build. 12:26
lindi-xiangfu: don't let that slow you down, if you package useful software and do it well it is usually easy to get somebody to sponsor it12:27
lindi-xiangfu: mostly people have problems finding sponsors when they package something totally esoteric with zero users12:27
xiangfuyes. I tried to ask help in [Debian Electronics Packaging Team] but no one reply me. I am working on fpgatools and toped.12:29
xiangfuwhich si all about electronics. :-)12:30
xiangfuthe toped upstream authors already use toped libs in their chip design.12:30
xiangfubtw: Fedora already include toped12:31
lindi-xiangfu: team mailing lists are bit difficult since nobody feels like taking the responsibility :/12:32
xiangfulindi-: too many details on debian packaging :)12:32
lindi-xiangfu: toped looks ok to me, fpgatools is probably not ready to be included in debian in current state however?12:32
lindi-xiangfu: for me the most confusing part was how git is used with debian packaging in some teams12:34
wolfspraullindi-: just for functionality, in today's state fpgatools can absolutely be used to program (one particular) fpga chip, the xc6slx912:35
wolfspraulversioning wise, it's 0.01 :-)12:35
lindi-wolfspraul: yes it's great but not for the masses yet :)12:35
wolfspraulsure, and never will be12:36
wolfspraulwhat is the goal?12:36
wolfspraulwe need to find more exotic tools/libs that are included as a proof that it is worthy of inclusion? :-)12:36
lindi-wolfspraul: heh12:36
wolfspraulit works today, and is the most advanced free tool to program that kind of chip12:36
wolfspraulby far, since I am not aware of anything even close12:37
lindi-wolfspraul: well personally I try to only upload stuff that I have tested. would need to buy that chip first :)12:37
wolfspraulah ok, that makes sense12:37
wolfspraulthat's at least a strong argument :-)12:37
lekernelhi wolfspraul 12:37
wolfspraulactually I don't care that much whether there are strong reasons or not for inclusion or not12:37
lindi-wolfspraul: but don't take this wrong, we have a growing fan group in our monthly free software meetings12:37
wolfspraulI'm definitely not going to spend time arguing about that. BUT ... I am very much interested in quality feedback that helps improving the tool12:37
lindi-wolfspraul: in fact, next thursday we are probably going to browse your git log at least briefly12:37
wolfsprauland that's why I like the process of Debianization12:38
lindi-I guess if I had time to play with it I could buy the chip and learn to use it12:38
wolfspraulso the argument "I can only talk about things I use" makes 100% sense to me12:38
wolfspraulthe argument "it's not ready" not so much12:38
lekernelwolfspraul, did you get my email about JTAG pods?12:38
lindi-wolfspraul: yeah I don't know how functional fpgatools is, I understood the led blinking example was done just last year12:39
wolfspraulthat's about as far as it gets12:39
wolfspraulbut keep in mind - it's a library12:39
wolfspraulI am working on making the library more capable, but it's still and for the time being will be a -dev thing12:39
wolfspraullike a programming language12:40
wolfspraulso when is a programming language "ready"?12:40
lindi-wolfspraul: good question :)12:40
wolfspraulI'd say when the syntax is kind of stable, makes sense, etc.12:40
wolfspraulso I rather work on better documentation, more entry points, more examples, etc.12:40
whitequarkwolfspraul: no no no no, don't bring your flaky measures of my perfect syntax here :D12:40
wolfspraulthat I will do, on my own. plus things from feedback that 'make sense' to me12:40
lindi-wolfspraul: yeah please don't take me wrong12:40
wolfspraulbut as for 'when it's ready' - you need to define that12:41
wolfspraulno, I totally don't. I appreciate your feedback!12:41
wolfspraulit's an obscure exotic thing, for sure12:41
wolfspraulbut it needs to be by definition, the only way out of obscurity is the very path we are on now12:41
wolfspraulincluding your very correct observations12:42
wolfspraulI would version it 0.01 today, for sure12:42
wolfspraulin good old humble foss style12:42
whitequarkwolfspraul: on the serious side; I'm not sure if it is a good idea to bring young software to Debian12:42
lindi-I have some fpga-knowledgeable friends, if I could get them play with that chip maybe they could help12:42
whitequarkit then tends to stick around in very obsolete version12:42
whitequarkjust my 2ยข12:42
xiangfulindi-: --libdir works fine. uploading...12:43
lindi-xiangfu: did you run lintian before upload? ;)12:44
lindi-whitequark: yeah especially if it is easy to compile from git12:44
xiangfulindi-: yes. I think those warnings are not important.: http://pastebin.com/JfE0h26m12:44
lindi-whitequark: however, we can upload it debian but never release it as part of any stable release12:44
wolfspraullindi-: if you follow sources, there's not that much going on lately12:49
wolfspraulI am slowly digging away at the distributed memory stuff lately, filling in bits and pieces everywhere12:49
wolfspraulmaybe another week or two in dist-mem12:49
wolfspraulafter that, another big block, will be block memory12:49
wolfspraulthat will be at least 1 month or so, LOTS of pieces there, for sure12:49
wolfspraulafter that some more infrastructure to improve programmability, more simple routing etc.12:50
wolfspraulbetter examples12:50
wolfspraulthen I want to start with a small core, J112:50
wolfspraulI will try whether it makes sense to program the chip via a llvm backend12:50
wolfspraulthat's roughly what you can expect in the next few months12:50
wolfspraulbut progress is slower now mostly because I am dealing with the effects of a growing codebase, so it gets harder to make changes12:51
lindi-whitequark: I have done work with LLVM but mostly with IR encoding12:51
wolfspraulbut good news is that I'm not doing this the first time, and I think so far the codebase remains manageable and valuable for the changes I need to apply as I'm going forward12:51
wolfspraulthat doesn't rule out a complete rewrite one day though, maybe when I switch to artix-7 (totally not on the radar right now)12:52
wolfspraulso much just as background noise if you check the sources again :-)12:52
xiangfulindi-: just fyi. a new toped upload to  http://mentors.debian.net/debian/pool/main/t/toped/toped_0.0+r2366-1.dsc12:55
xiangfulindi-: you can take a look when you have time.13:00
wolfspraullekernel: no sorry, I don't have those pods right now13:00
wolfspraullindi-: please keep the fpgatools feedback coming13:00
lekernelwolfspraul, who has them?13:00
wolfspraulif you want hw to play with, maybe we can build some little something for you :-)13:00
wpwrakwhitequark: (old stuff stays round forever) i think that's a very good point, especially in this case where still lots of interface changes are likely13:05
wpwrakwolfspraul: i think getting it into debian should be an extremely low if not negative priority at the moment. since it's a library, any users will have to know how to compile things anyway, and there are no applications that depend on it. so having it in debian give you very little practical value.13:07
wpwrakwolfspraul: however, what still seems to be missing is to get the word out. get people to experiment with it, provide feedback. that way, you can stabilize the interfaces.13:08
lindi-wpwrak: yep you need to reach the people who have the chip or who want to buy it13:11
wolfspraulit's a programming language, it will always stay like this13:11
wolfspraulif someone wants to make use of it, they will find it13:11
wpwrakwolfspraul: naw, what you have now is a very low-level interface. when other things may be layered on top. and you may want to spread out at the bottom too, for example to different chip sizes or different chip generations.13:13
wpwrakwolfspraul: how you package all this is a largely orthogonal problem13:13
wpwraklindi-: exactly. there has to be a board people can go and buy. can be custom made or something that already exists. but there must be an easy way that also leads through the integration quagmire.13:15
wpwrakwolfspraul: and of course, documentation :) seems that you're working on that. i wish i could help there, but i'm running out of time :-(13:16
wpwrakso the items needed for success are 1) a cute demo to catch people, 2) documentation to give them an idea of what to expect, 3) a board they can get easily, for their own experiments, 4) documentation to tell them how to use the library/language, 5) a reasonably simple way to generate and run their "hello world"13:18
wpwrakof course, 4) should not only be about the syntax but also the semantics. e.g., with a brief overview of the FPGA organization and the structures things refer to. part of that could be taken from xilinx (or other) documentation, but it may be helpful to have at least an overview, so that people don't have to jump around among dozens of documents13:21
wpwrakregarding the hw, if you choose to make a custom board, then, well, you have to make that and produce a number of them. or find someone who will produce them.13:22
wpwrakif you choose to use an off-the-shelf board, then you may need to adapt the library for its chip, and test the critter. in the process, also integration of the tools needed to operate the board will happen13:23
wpwrakyou can also try a multi-stage approach: just make a demo and dump whatever you have, then hope that someone else picks up tasks like finding/making hardware for it13:24
wpwrakthat's a risky approach, though, since we already know that people are very reluctant to get involved at such tasks. also, this will unavoidably cause a longer delay between first demo and actual hardware, so you'll not be able to "ride the hype"13:25
wolfspraulno no :-) my plan is entirely different. if someone wants to use this for anything, they need to drive the use cases forward13:27
wolfspraulsome things are on the slow cooker, this one will slow-cook...13:28
wpwrakhow do you expect anyone who could drive it forward to find it ? right now, you're reaching basically a small group of hobbyists who can't commit to any effort of that scale13:29
wpwrak(even though they may be able to contribute a lot with time, a piece at a time)13:29
lindi-if we got one such chip to our hacklab and a hello world I'm sure we might get some contributors :)13:30
lekernelwolfspraul, ping13:35
--- Wed Feb 13 201300:00

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