sb0 changed the topic of #m-labs to: ARTIQ, Migen, MiSoC, Mixxeo & other M-Labs projects :: fka #milkymist :: Logs http://irclog.whitequark.org/m-labs
sandeepkr__ has quit [Read error: Connection reset by peer]
sandeepkr__ has joined #m-labs
<sb0> mithro, what part of clock domain crossing?
<mithro> sb0: Correct and efficient ways to do it in FPGA fabrics and represent that in Verilog (and thus migen). I think I understand big things like using FIFOs (for doing whole bytes / buses) but not really how to do thing like single bit signals (especially efficiently).
<cr1901_modern> mithro: Two flip flops in series clock in the source domain should be sufficient
<cr1901_modern> clocked*
<cr1901_modern> Idk why two are needed; it has something to do with metastability
<sb0> look for e.g. double stage synchronizer. there is a component for that in migen as well
<cr1901_modern> The output of the second flip-flop is "seen" by the dest domain
<mithro> cr1901_modern / sb0: I want to understand it properly, IE How and why. Everything I've started reading causes my eyes to glaze over after a couple of paragraphs.
<cr1901_modern> mithro: Presumably, you've read fpga4fun's page on the topic? That's what I used to understand it in the past, but it obviously hasn't stuck
FelixVi has quit [Quit: Leaving]
<GitHub158> [artiq] jordens pushed 2 new commits to master: https://git.io/vVSwd
<GitHub158> artiq/master 7f7750e Robert Jordens: installing.rst: triple colons
<GitHub158> artiq/master 62eae5e Robert Jordens: installing.rst: adapters renaming from ed1c368
<GitHub133> [artiq] jordens pushed 1 new commit to release-1: https://git.io/vVSwF
<GitHub133> artiq/release-1 a82f042 Robert Jordens: installing.rst: triple colons
<GitHub121> [misoc] sbourdeauducq pushed 2 new commits to master: https://git.io/vVSrT
<GitHub121> misoc/master ed0e451 Florent Kermarrec: cores/sdram_settings: simplify sdram modules
<GitHub121> misoc/master 452cbc3 Florent Kermarrec: cores/sdram_settings: use phy rate to convert ns values to sys_clk cycles (required to compute margin)
<bb-m-labs> build #93 of misoc is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/misoc/builds/93
<bb-m-labs> build #322 of artiq-kc705-nist_clock is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-kc705-nist_clock/builds/322
<bb-m-labs> build #105 of artiq-win64-test is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/105
<bb-m-labs> build #581 of artiq is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/581
<bb-m-labs> build #323 of artiq-kc705-nist_clock is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-kc705-nist_clock/builds/323
<bb-m-labs> build #106 of artiq-win64-test is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/106
<bb-m-labs> build #582 of artiq is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/582
mumptai has joined #m-labs
<cr1901_modern> sb0: Has the minimum gcc version required for MiSoC been bumped since the API change? Getting a "cc1.exe: error: -Werror=incompatible-pointer-types: no option -Wincompatible-pointer-types" error.
<sb0> cr1901_modern, I suppose.
<sb0> when was this option introduced in gcc, if ever?
<cr1901_modern> I have no idea. Seems to be between 4.8 and 5.2, accoding to this: https://github.com/Grawp/rtl8812au_rtl8821au/issues/1
<cr1901_modern> Is "-Werror=incompatible-pointer-types" a Clang option (I don't really use Clang)?
jaeckel has quit [Ping timeout: 276 seconds]
<sb0> rjo, did you do some force-pushes or something? after pulling, your manual changes appeared in my "git diff" uncommitted changes somehow
rohitksingh has joined #m-labs
<whitequark> that doesn't happen after force-pushes
<sb0> and this sort of accident is precisely why I don't trust things like "merging release-1 into master"
<rjo> sb0: no. i didn't do anything.
<rjo> sb0: that sounds more like a mess up on your side.
<whitequark> yes.
<whitequark> not understanding git is different from git working incorrectly.
<sb0> yes, somehow I have the old version of doc/manual/installing.rst, except I did nothing that could result in that
<rjo> sb0: if you pull (resulting in a merge) and do a bit of saving in your editor at the right time i can see that happening
<sb0> hm, yes, it could be that
<sb0> so that would no be git's fault
<whitequark> "but conda will
<whitequark> faithfully follow Murphy's law and won't downgrade llvmlite-artiq to update
<whitequark> artiq, unless you use the command above"
<whitequark> amazing. so it won't downgrade when you want it, but will downgrade when you don't
<GitHub16> [artiq] sbourdeauducq pushed 1 new commit to master: https://git.io/vVSDM
<GitHub16> artiq/master 3a0c206 Sebastien Bourdeauducq: do not open console windows for GUI applications on Windows
<rjo> sb0: do you also get https://github.com/m-labs/artiq/issues/379#issuecomment-206442353 now when closing a gui app?
<rjo> or do you have a newer python that shows which exception is lost?
<sb0> rjo, I get some garbage, but not that
<sb0> on Windows. i still have old quamash on linux
<rjo> i get it since the new quamash but that change seems completely unrelated to me.
<sb0> maybe windows prints the error message correctly
<rjo> yes. my guess is also that it probably was there before.
<rjo> it seemed to be related to the fds for ipc per applet.
<sb0> happens intermittently
<sb0> like any good bug does
jaeckel has joined #m-labs
<rjo> sb0: hmm. that one looks very windowsy
<sb0> rjo, when the browser is started, the right panel will show directories, not just hdf5 files
sandeepkr_ has joined #m-labs
<bb-m-labs> build #324 of artiq-kc705-nist_clock is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-kc705-nist_clock/builds/324
sandeepkr__ has quit [Ping timeout: 248 seconds]
kuldeep has quit [Ping timeout: 260 seconds]
sandeepkr__ has joined #m-labs
kuldeep has joined #m-labs
<bb-m-labs> build #107 of artiq-win64-test is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/107
<bb-m-labs> build #583 of artiq is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/583
sandeepkr_ has quit [Ping timeout: 260 seconds]
<cr1901_modern> sb0: What is the rationale behind these two lines? w/ the new MiSoC, the new "--gateware-toolchain-path" in misoc's builder.py does not have any effect: https://github.com/m-labs/migen/blob/master/migen/build/xilinx/ise.py#L148-L149
<cr1901_modern> on Windows*
<cr1901_modern> (Of course, I can just add Xilinx to my PATH. I'm guessing you assume Windows users would do that?)
<sb0> whitequark, any clue about this? https://github.com/m-labs/artiq/issues/386
<whitequark> what DRM?
<whitequark> that's nonsense
<sb0> login to win7-experimental. it's showing it right now.
<whitequark> I can't, your DNS broke
<whitequark> router.lab.m-labs.hk or win7-experimental.lab.m-labs.hk do not resolve
<sb0> works here
<whitequark> are you in the lab?
<sb0> lab.m-labs.hk has address 59.152.213.90
<sb0> no
<whitequark> no
<whitequark> that's the IPv4
<sb0> 2002:3b98:d55a:0:a60:6eff:fe44:4295
<whitequark> as I said
<whitequark> win7-experimental.lab.m-labs.hk does not resolve
<whitequark> I don't care about lab.m-labs.hk
<sb0> win7-experimental.lab.m-labs.hk is not supposed to resolve on the internet
<whitequark> it is
<whitequark> that was the whole point of introducing IPv6
<sb0> are you saying this machine is visible to the whole internet? it's not properly secured for that
<whitequark> no, it isn't *accessible*, that's what the firewall is for
<whitequark> (it's set to drop forwarded packets by default)
<whitequark> okay, I've connected
<sb0> cr1901_modern, there is no source command on windows. is there an equivalent xilinx script?
<whitequark> there is an equivalent of source command
<whitequark> doing > a.bat is like "source a.sh"
<whitequark> whereas > call a.bat is like "./a.sh"
<sb0> okay, but is the xilinx script supposed to be executed like that?
<sb0> _florent_, ^
<whitequark> well, look at it...
<whitequark> if it does a bunch of sets then yes
<sb0> I don't have xilinx tools installed on windows, this is all _florent_'s code
<cr1901_modern> sb0: Yes, the Xilinx script should be executed like that. Like source on bash, it will propogate the environment variables to the parent cmd.exe
<whitequark> btw, VS 2013 Community is free, it just needs a login to microsoft.com
<cr1901_modern> I'm not sure the best way to handle it, tbh
<sb0> _florent_, did the call command not work for some reason?
<cr1901_modern> whitequark: Are you sure you don't have it backwards? I think call is the equivalent of source on Windows
<whitequark> I am sure, yes
<cr1901_modern> source "runs the script, and then replaces the child shell with the current, saving all new environment state?"
<sb0> cr1901_modern, afaict, those two lines you mentioned should be removed, and source_cmd should be "" on windows. can you test and submit a patch?
<sb0> and source should default to True instead of None in build()
<sb0> xilinx says the call command should be used though
<cr1901_modern> Will do. source was originally False, IIRC. But I doubt it'll really mess with any existing code to make it true.
<whitequark> hm, interesting
<cr1901_modern> sb0: According to git blame, I added those lines
<sb0> look at what the shortcuts say
<cr1901_modern> So I think they should be kept
<sb0> *do
<sb0> cr1901_modern, well, please do your research and your tests, and sort this out
<sb0> and the default behavior should be sourcing on linux
<whitequark> amazing, both vs2013 and vs2015 refuse to login because of some unspecified error
<cr1901_modern> Ack.
<cr1901_modern> Okay, the call stays; past-me was correct for once.
<_florent_> sb0: this time it seems I'm not responsible for that: https://github.com/m-labs/migen/commit/6370acd9685184eff8698320c35278afe33c8f7b cr1901_modern can you look at that?
<_florent_> from what I remember I was not sourcing the environment on windows
<cr1901_modern> _florent_: Yes, this is my fault. I am now vaguely recalling that on your Windows machine you have Xilinx tools already set up when cmd.exe opens
<cr1901_modern> I wanted to use Migen without assuming a bash install, so I did the closest equivalent to the Linux commands. I don't have the Xilinx tools on my path, and the old default was to *not* source.
<cr1901_modern> When MiSoC got the API change, for whatever reason, the default on Windows only was not to source the Xilinx scripts.
<cr1901_modern> Incidentally, I need to fix the generated batch file to die more gracefully. So there will be two patches today.
FabM has quit [Quit: ChatZilla 0.9.92 [Firefox 45.0.1/20160315153207]]
rohitksingh has quit [Quit: Leaving.]
fengling has joined #m-labs
<GitHub129> [artiq] jordens pushed 2 new commits to master: https://git.io/vVSAy
<GitHub129> artiq/master 6c200b1 Robert Jordens: browser: make results a dock again
<GitHub129> artiq/master 5e5f94a Robert Jordens: browser: WIP for experiment replay
<sb0> rjo, artiq-kc7005-nist_qc2
<sb0> in doc/manual/installing.rst
<bb-m-labs> build #325 of artiq-kc705-nist_clock is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-kc705-nist_clock/builds/325
<bb-m-labs> build #108 of artiq-win64-test is complete: Failure [failed python_unittest] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/108 blamelist: Robert Jordens <rj@m-labs.hk>
<bb-m-labs> build #584 of artiq is complete: Failure [failed] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/584 blamelist: Robert Jordens <rj@m-labs.hk>
_rht has joined #m-labs
<whitequark> hmpf
<whitequark> this analyzer problem again.
<whitequark> sb0: I think I'm going to insert back the device reset to work around this bug
<GitHub10> [artiq] jordens pushed 1 new commit to release-1: https://git.io/vVShx
<GitHub10> artiq/release-1 2617b9d Robert Jordens: installing.rst: typo
<rjo> sb0: artiq-hal9000
<bb-m-labs> build #326 of artiq-kc705-nist_clock is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-kc705-nist_clock/builds/326
<whitequark> heh
<rjo> "I'm sorry, Dave. I'm afraid I can't do that."
fengling has quit [Quit: WeeChat 1.4]
<bb-m-labs> build #202 of artiq-pipistrello-nist_qc1 is complete: Exception [exception interrupted] Build details are at http://buildbot.m-labs.hk/builders/artiq-pipistrello-nist_qc1/builds/202
<sb0> whitequark, leave it as it is now. this collects some data on how frequent this bug is.
<bb-m-labs> build #203 of artiq-pipistrello-nist_qc1 is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-pipistrello-nist_qc1/builds/203
<bb-m-labs> build #109 of artiq-win64-test is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/109
<bb-m-labs> build #585 of artiq is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/585
<sb0> whitequark, and is resetting the device making the bug go away?
<sb0> whitequark, by the way, I don't understand why *my* DNS is causing trouble to resolve win7-experimental.lab.m-labs.hk. isn't that resolution happening on the openwrt router in the lab, and only there?
<whitequark> it works in the lab
<whitequark> and it ought to work elsewhere too
<rjo> whitequark: how would it? there is neither an ns entry nor an aaaa.
<sb0> I don't see how, nor why that would be desirable. all the public DNS has is A(AAA)-records for lab.m-labs.hk
<whitequark> rjo: I vaguely remembered that it existed, apparently wrongly
sb0 has quit [Quit: Leaving]
sb0 has joined #m-labs
<sb0> https://twitter.com/whitequark/status/719094878437638144 << yup, that's what I referred to as "DRM bullshit"
<whitequark> I mean, DNS changes little, I can still address it directly by the IP
<whitequark> ... if DHCPv6 actually worked on Windows...
<whitequark> or RA
sandeepkr_ has joined #m-labs
sandeepkr__ has quit [Ping timeout: 260 seconds]
kuldeep has quit [Ping timeout: 276 seconds]
kuldeep has joined #m-labs
sandeepkr__ has joined #m-labs
kuldeep has quit [Ping timeout: 246 seconds]
sandeepkr_ has quit [Ping timeout: 260 seconds]
kuldeep has joined #m-labs
evilspirit has quit [Ping timeout: 240 seconds]
<cr1901_modern> Hmmm... I wonder what the best way to debug MiSoC crashes is...
mumptai has quit [Remote host closed the connection]
_rht has quit [Quit: Connection closed for inactivity]