#milkymist IRC log for Thursday, 2011-09-01

xiangfukristianpaul, have you copy Xlinx.lic to fidelio.qi-hardware.com?02:17
kristianpaulxiangfu: yes, why?02:46
xiangfukristianpaul, I meet a error when compile standby.bin. says like miss license file.02:47
xiangfukristianpaul, I just copied one to /home/xiangfu/.Xilinx/Xilinx.lic02:47
xiangfukristianpaul, check : http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-08312011-1923/ the last 100 lines shows like:02:47
kristianpaulyeah may be because the user02:48
kristianpaulas i ran it as root..02:48
kristianpaulactually first timne i isntalled ise that way :)02:48
kristianpaulwow 27Mb log ;-)02:49
xiangfukristianpaul, the error message is like: http://pastebin.com/JjXgtEbH02:51
xiangfukristianpaul, so I copy one my laptop Xilinx.lic to fidelio.qi-hardware.com://home/xiangfu/.Xilinx/Xilinx.lic02:51
xiangfukristianpaul, /home/xiangfu/.Xilinx/*.lic <---- :)02:52
xiangfukristianpaul, http://fidelio.qi-hardware.com/~xiangfu/bin/compile-milkymist-firmware.sh here is the script file02:53
xiangfukristianpaul, much small then openwrt ben nanonote :)  it's ~220M02:54
kristianpaulxiangfu: so problem is gone now?02:59
kristianpauli'm wondering if you have to run xlcm again..03:00
xiangfukristianpaul, not sure. wait for next trigger by cron job03:01
kristianpaulcant you push it now? :)03:01
kristianpaulor too busy server?03:01
xiangfunow? that is bad. the build order is like openwrt-xburst, openwrt-milkymist, milkymist-firmware.03:02
kristianpaulhe, i tought that03:02
kristianpaulno problem03:02
xiangfumay cause some problem. so wait is better maybe next 3 hours it will start build(there are errors on openwrt-xburst, so it will not build 36 hours :)03:03
kristianpaulbtw a grep sys_clk that match an (*) asterisk just in case timining is not met could be considered03:05
wpwrakspeaking of easing access for developers, do we actually have any decent instructions for writing M1 patches/presets ? or is M1 really close enough to milkdrop that anything that works there works on M1 too, and vice versa ? (i'm still worming my way through the milkdrop instructions)05:23
wpwrakoh, excellent. thanks a lot !05:27
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-openwrt.minimal-09012011-1006/09:01
lekernelrejon__, check the log11:02
rejon__oh, i missed some11:03
rejon__chinese internet!!!!11:03
rejon__the logs look not complete11:04
rejon__looked at wrong logs11:05
kristianpaulximian: (patching) hey great page, i havent noticed it:)11:18
kristianpaulergg was xiangfu :)14:22
errordev1loperhey ..19:02
errordev1loperanyone used Modelsim+VPI ?19:02
errordev1loperwell, generally anyone with the knowledge of Modelsim cli would be able to help19:03
lekernelwhy not use gplcver or icarus?19:03
lekernelboth support VPI19:04
errordev1loperlekernel: yes, I used Icarus19:35
errordev1loperbut I want to try my to learn how to use VPI with modelsim :)19:35
errordeveloperwhat I had been trying to do is this:19:38
lekernelwhy bother with modelsim? HDL simulation is perhaps the only EDA area where free software doesn't suck19:38
errordeveloper https://github.com/errordeveloper/sftb/blob/master/vpi/test-modelsim.csh19:38
lekernelthe only important thing which isn't possible with free software but modelsim supports is mixed language simulation19:38
errordevelopersure ... well, but I would like to come up with a cross-toolchain framework19:39
lekernelthere's even verilator, which is often _faster_ than modelsim19:39
errordevelopersure, I'll get to verilator too :)19:39
errordeveloperand looks like you don't even need DPI or VPI with verilator19:40
errordeveloperjust call the library from main()19:40
lekernelhttp://vcomp.sourceforge.net/ and http://veriwell.sourceforge.net/ are also interesting, though a bit buggy and no longer maintained19:41
lekernelyeah, verilator converts your verilog into a c++ model19:41
lekernelso VPI no longer makes sense there19:41
errordeveloperbut as I said, I am working towords cross-tool framwork19:43
errordeveloperso you could do 'make USE=MyTool' on you project19:44
errordeveloperthat's very quick explanation though19:44
errordeveloperI found hdlmake and boldport19:45
errordeveloperhdlmake is okay, but still very immature and after a discussion on their mailing list I somewhat figured that I'd rather roll my own framework19:46
lekernelhmm... why not try improving hdlmake instead?19:46
errordeveloperboldport looks kindda cool, but the author seems to have no intention to make it a command line tool or at least a gui19:47
lekernelI guess Tomasz would welcome your contributions :-)19:47
errordeveloperboldport is web-based ...19:47
lekernelplus hdlmake has a skilled user base already19:47
lekernelboldport is SaaS I wouldn't touch19:47
errordeveloperyeah ...19:48
lekernelI won't even discuss it :-)19:48
lekernelbut hdlmake is nice and properly licensed19:48
errordeveloperwell, I thought to kind of see how far I'll get and then backport into hdlmake19:48
lekernelhdlmake is used for some CERN projects19:49
errordeveloperand I didn't fill like learning python right now19:49
lekernelyou should. python is one of the best languages out there.19:49
errordeveloperI'd do it in Ruby and then once I have a clear methodology I can see how to re-implement it in python19:50
errordeveloperthat was an idea somewhere at the back of my mind19:51
errordeveloperwell, hdlmake in it's current state needs a lot of refactoring19:52
errordeveloperwhich I couldn't take on with my little knowledge of python19:52
errordeveloperand the goddamn 'meaningful whitespaces19:53
errordeveloperwould kill me again!19:53
lekernel:-) I don't know why so many people dislike that idea19:53
lekernelI find it elegant19:53
errordeveloperit's not a bad thing, just gets me mad a little bit19:54
wpwraklekernel: the whitespaces get inconvenient if you move around code for debugging. also, python has that unspeakably retarded requirement of continuing many lines with a backslash. that's perhaps the #1 weakness CPP inherited from the mainframe days, and i find it rather absurd that python kept that. even the bourne shell is a little smarter :)19:57
wpwrak(moving around code for debugging) e.g., add an if to skip a block or maybe add a print to check a variable. in C, i can put all these things in the first column, so they stick out and i won't forget to remove them when i'm done. in python, Guido doesn't trust me to have the moral strength to write clean code, so i'm forced to make even such very temporary constructs look "nice". and of course, that also makes is harder to spot them late20:03
wpwrakr, when the bug is dead and it's time to clean out the debugging stuff.20:03
mwallewpwrak: most \ can be replaced with parenthesis, and most editors can indent blocks/lines :)20:44
mwalleif there are many backslashes for line contiuation, the code will need some refactoring imho :)20:48
mwalleand i like it that theres only one coding style :)20:50
wpwraksometimes, others force long identifiers on you. then your lines overflow very quickly. also, the class - function - body structure means that almost all of your code is already two tab stops indented. so you lose one level of indentation to OO.20:52
wpwrakthe uniform coding style is nice, though. i just wish a little more thought had gone into the scanner. you could have the "clean" and uniform style and a better syntax, without too much effort.20:53
wpwrak(\ -> parentheses) yes, but then your code gets less readable. lots of parentheses are fine for your Lithp machine, but for humans ?20:54
wpwrak(long identifiers) and sometimes you don't have a good alternative either. particularly if doing geometry. lots of things that need qualifiers. of course, you can just use x0, y2, a13, and so on, but that kinda defeats the purpose :)20:55
wpwrak(autoindent) yes, but that still hides your debugging code :) i _want_ it to stick out. forgetting anything from it is bad. and sometimes, you need lots of stuff scattered all over the place to track down a particularly pesky problem. and you want it there until after you've implemented and confirmed your fix.21:00
wpwrakof course, you rcs will tell you what changed, but you still need to decide which is debugging and which is fix, for they'll all look alike. making it easier to make mistakes is a bad idea in my opinion.21:01
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-openwrt.minimal-09012011-2240/21:36
--- Fri Sep 2 201100:00

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