Wednesday, 2017-01-11

*** tpb has joined #timvideos00:00
*** mithro has quit IRC00:19
*** cyrozap has quit IRC00:20
*** shenki has quit IRC00:20
*** shenki has joined #timvideos00:21
*** ChanServ sets mode: +v shenki00:21
*** cyrozap has joined #timvideos00:25
*** mithro has joined #timvideos00:27
*** ChanServ sets mode: +o mithro00:27
techman83mithro: I had a dig into it, I'm probably lacking available time/energy to do it justice.00:32
techman83mithro: Were at the tail end of finishing a codebase overhaul at work, so a little drained!00:33
CarlFKxfxf: (11:00:40 AM) `justin: he looks like he's scaring the birds00:56
mithro_florent_: Looking at those coefficients and the xapp931, the CC and CD values appeared swapped?01:06
_florent_mithro: I'm looking at that01:11
_florent_mithro: you should take this:
tpbTitle: encoder/core: move output fifo to block ram(should ease P) · enjoy-digital/[email protected] · GitHub (at
_florent_mithro: CC and CD seems to be swapped indeed01:17
CarlFKthaytan: how up are you on the gst declink plugin?    or really how can I see details of what is currently plugged in?  (trying to confirm that the card will not read from the hdmi2usb, likely because dvi != hdmi01:19
mithro_florent_: thaytan said "<thaytan> I suspicious about the way it scales coefficients to 0..255 and truncates to integers immediately though" when I asked about the white saturation issue on #gstreamer01:25
tpbTitle: HDMI2USB-litex-firmware/ at nextgen · mithro/HDMI2USB-litex-firmware · GitHub (at
_florent_ok thanks01:34
thaytanCarlFK, I'm not really at all01:39
thaytanslomo knows it better, and I don't have access to one01:39
*** tims-extra-lapto has quit IRC01:49
CarlFKthaytan: no prob - we kinda think we know what is going on and we know how to work around it01:49
*** hyades has quit IRC02:01
*** sb0 has quit IRC02:05
tpbTitle: UNITEK (Y-6332) DisplayPort to HDMI Female (Support 4K) Converter - TAS Glenorchy (at
mithroxfxf: I'd get 5 or 10 of them if they are in stock05:09
xfxfthey're a 1h15m round trip though and not open today05:09
mithro_florent_: Any idea why on the IFCLK from the FX2 we don't use an IBUFG object to buffer the incoming clock?05:18
mithrohyadez: Ping me when you are around and want to do more work05:18
_florent_mithro: I probably reused some old code05:19
_florent_mithro: and VHDL code is using falling edge of the clock05:19
mithro_florent_: Yeah, it looks like it.... Would Xilinx infer that?05:19
mithro_florent_: We should replace that VHDL file...05:20
mithro_florent_: Seems like a pretty simple statemachine...05:21
mithro_florent_: I'm sure I understood this code at some point ago :-P05:22
mithrojea: Can you take a look at the EDID data here -> -- Does that decode look right?05:23
tpbTitle: Fix HDMI2USB output with Blackmagic gear · Issue #7 · xfxf/lca2017-av · GitHub (at
mithroBlock has broken checksum05:24
mithro_florent_: Adding an IBUFG makes the encoder soc work05:32
_florent_mithro: good05:33
mithro_florent_: Also allows the timing constraint to be applied to that signal too05:33
_florent_mithro: new feature:05:33
tpbTitle: output/core: add underflow detection · enjoy-digital/[email protected] · GitHub (at
tpbTitle: firmware: add underflow detection · enjoy-digital/[email protected] · GitHub (at
tpbTitle: Snippet | IRCCloud (at
mithroThoses are not good numbers :-P05:35
_florent_no but this a case that is not working05:35
mithroDo you get zero underflows in the case where it does work?05:37
mithro_florent_: Enabling the I2C stuff seems to cause the firmware to lockup, I'm currently trying to debug05:43
mithro_florent_: Is that feature ready to be cherry-picked across?05:43
_florent_mithro: yes underflow is zero when working05:43
_florent_so at least the underflow detection is working05:43
_florent_I'm trying to integrate that in a better way05:44
*** hyades has joined #timvideos05:44
mithro_florent_: Okay05:44
mithrohey hyades05:44
mithro_florent_: The encoder only soc seems to take about ~15 minutes for a full run05:46
tpbTitle: Snippet | IRCCloud (at
_florent_mithro: you can pick up the underflow feature, not going to touch it05:48
mithro_florent_: Will do that as soon as I finish the I2C reading05:51
mithrojea: Take a look under third_party/litex/litex/soc/software/ for our "stdlib" and other stuff05:59
tpbTitle: newlib/strcasecmp.c at master · eblot/newlib · GitHub (at
tpbTitle: Google Sheets - create and edit spreadsheets online, for free. (at
xfxfTLDR; most rooms have a fast laptop for mixing + camera input (via HDMI2USB), then a slower laptop connected via gigE at the front pushing in another HDMI2USB (presenter)06:20
xfxfone of the capture scripts we were using before (note audio-delay option):06:21
tpbTitle: video-scripts/ at master · xfxf/video-scripts · GitHub (at turned into voctomix-outcasts, should have similar params, was based off the same script06:22
*** aps has joined #timvideos07:15
hyadesmithro: morning!07:16
CarlFK title: The School for Sysadmins Who Can’t Timesync Good and Wanna Learn To Do Other Stuff Good Too07:24
CarlFKfun title07:24
CarlFKmithro: is there somewhere that mentions the opsis output being dvi not hdmi?08:35
mithroCarlFK: What do you mean?08:36
CarlFKmithro: your theory on why the black magic gear doesn't see anything08:37
mithroCarlFK: Sure - why do you care?08:37
CarlFKto pass onto someone who does blackmagic stuff08:38
CarlFK(07:10:33 PM) JP Viljoen: 720 x 480 59.94Hz, 720 x 576 50Hz, 1280 x 720 59.94Hz 50Hz and 1920 x 1080 59.94Hz 50Hz.08:38
CarlFK(07:13:34 PM) JP Viljoen: possibly the frequency on your output is mismatch08:38
mithroCarlFK: Yeah, I'm working on output the exact frequency they are requesting08:39
mithroCarlFK: Our signal is a basic DVI 1.0 signal, no HDMI data islands or anything like that08:39
mithroCarlFK: Can you ask him if this EDID information looks right?
tpbTitle: Fix HDMI2USB output with Blackmagic gear · Issue #7 · xfxf/lca2017-av · GitHub (at
mithroCarlFK: That is the EDID I dumped from Ryan's mixer08:41
mithroCarlFK: And RGB pixels, not YUV08:42
CarlFKis there a URL I can point him to that says this stuff?08:42
mithroCarlFK: no08:43
*** nrossi has joined #timvideos09:53
*** hyades has quit IRC10:01
*** aps has quit IRC10:06
mithro_florent_: How many CSR register locations can we have again? 16 or 32?10:09
xfxfmithro: i can't read some of your notes10:11
mithroxfxf: Just in the middle of debugging something, do your best and I'll come have a look shortly10:11
mithroxfxf: We haven't tested the new firmware10:14
xfxfmithro: OK, surely that's a good idea before james leaves?10:14
mithroxfxf: Need to add the pixel clock detector back in and a few other small things, it can't hurt but we'll need him back again in a day or two anyway10:15
CarlFKreplace his mac with one of those clip boards and send him home10:16
mithroxfxf: It might be worth seeing if you reproduce my results10:16
xfxfi'd prefer to get on with other testing - it's starting to get awfully near the weekend and I still need to do work on the monitoring stuff10:19
xfxfyour testing seemed thorough?10:19
xfxfam trying to make sense of your spreadsheet - when you're done come grab me10:19
mithro_florent_: It seems like the ethernet CSR registers must be in the first 8?10:22
_florent_mithro: not sure, why are you saying that?10:26
mithro_florent_: The reset behaviour seems to be coming from NetSoC and it occurs when I add too many csr registers to BaseSoC10:27
mithro_florent_: well, that is the current theory - still trying to prove that is true10:27
tpbTitle: debconf-video/ansible - Ansible playbooks for the DebConf Video Team's machines (at
tumbleweedfor a fun time, visit http://jvc:[email protected]/11:42
CarlFKtumbleweed: I have an uncommitted change: -    p.add_argument('--file', type=argparse.FileType('w'),12:10
CarlFKI never really tested it, but I guess I should commit it and do something with tags?12:10
tpbTitle: Release lca release 1 with first fixes · CarlFK/voctomix-outcasts · GitHub (at
CarlFKmithro: did you get it, and does it look ok enough that I can blast out all 17912:32
mithroCarlFK: I got it, replied with feedback, I would change a bunch of things12:34
CarlFKmithro: you think the start time should come before projector hookup?12:43
mithroCarlFK: When you have to be where is the most important information I think12:52
CarlFKmithro: send revised.  I am not going to bother trying to explain the review bit, most people don't care so I don't want to waste there time with too many details12:54
CarlFKer, /send/sent/12:55
mithroCarlFK: I would leave the review bit out then12:59
CarlFKmithro: meh, not worth messing with tonight.  did you see the revised with the when/where at the top?13:03
CarlFKand really if you are doing stuff again, you shouldn't be looking at this13:04
mithroCarlFK: ?13:06
CarlFKmithro: I'm leaving the review bit as is.   and I just noticed I put the when/where under hdmi.. leme fix that...13:08
*** CarlFK has quit IRC13:32
xfxfjea: micolous arrives sometime today (Thu), so you'll need to switch gears into streaming stuff later on tomorrow i'd imagine13:41
xfxfi guess this is somewhat contingent on the state of the machine imaging, though13:42
xfxfbut perhaps a good time to do it given the streaming scripts should be baked into the ansible stuff13:42
jeaI'm hoping to move to monitoring and streaming tomorrow13:42
jeayes, I hadn't thought about the streaming being in ansible, but that makes sense13:42
xfxfrad.  deeprave has been hacking on hdmi2usbd, should hopefully have some last bugs ironed, which will make the monitoring easier13:43
jeaok, that sounds good. then it will just be sorting out the data we want to capture, and pushing it somewhere13:44
xfxfat an absolute minimum even logging locally + using logrotate every hour or whatever would be good13:45
xfxfa nice feature might be to have it connect to vocto + split log files when a cut happens13:45
xfxfit'll make matching a log file to a video file super easy then13:45
xfxfi'm happy handling the logging/sending stuff somewhere side of things if you want to investigate the monitoring13:46
*** hyadez has quit IRC13:46
xfxfand the data normalisation13:46
xfxflogstash seems interesting13:46
jeasplitting log files, or at least logging a cut somewhere in the overall system would be useful13:47
xfxfi can see for mithro having a set of log files that directly correspond with our recording files being excellent for debugging13:47
jealogstash sounds good, but i haven't looked at it enough to say what it is like. i guess i'll find out soon!13:47
xfxfyou'll be able to see in the video when people are futzing around + match that easily to the right log file13:47
jeayes, i think that would be easiest for mithro13:47
jeai'm going to write a generic data gathering tool, and allow it to output to multiple things (file, time series, etc)13:48
jeathat way we can easily push the data to where we want13:48
xfxfdown the track i can see a ton of useful metrics we'd want in AV HQ for future events beyond the opsis monitoring13:49
xfxfplus the way the recording machines are being built is everything is systemd13:49
xfxfso i'd imagine we can do a bunch of stuff there too in terms of monitoring events13:49
jeai'll model it on what eventstreamer used to do, but make it more extensible13:49
jea(and no perl...)13:49
xfxfmy biggest concern with eventstreamr too was it became _the_ way things were started/invoked13:50
jeayes, i expect systemd to help a lot with that actually13:50
xfxfi'd really want the monitoring decoupled13:50
xfxfnot that it shouldn't be able to restart services, but in a decoupled way13:50
jeai want eventstreamer 2.0 to be lots of individual components that can be deployed in any way - monitoring, service management, etc13:50
jeaand there should actually be a streaming component to it for once13:51
jeainstead of a mostly empty folder13:51
* jea sleep13:54
*** hyades has joined #timvideos14:09
*** rohitksingh has joined #timvideos15:52
*** hyades has quit IRC16:21
*** rohitksingh has quit IRC17:43
*** sb0 has joined #timvideos18:24
*** [d__d] has quit IRC18:32
*** hyades has joined #timvideos18:33
*** [d__d] has joined #timvideos18:34
*** nrossi has quit IRC20:30
*** CarlFK has joined #timvideos20:54
*** ChanServ sets mode: +v CarlFK20:54
CarlFKanyone awake?21:04
tumbleweedCarlFK: the door is open, I'm hopping in the shower21:19
CarlFKyay - thanks21:20
*** hozer has joined #timvideos22:16
*** hyades has quit IRC22:21
tpbTitle: veyepar/ at master · CarlFK/veyepar · GitHub (at
CarlFK   class Log(models.Model):                           episode = models.ForeignKey(Episode)22:51
tpbTitle: veyepar/ at master · CarlFK/veyepar · GitHub (at
tumbleweedCarlFK: entirely untested
tpbTitle: debian Pastezone (at
tumbleweedI think it's the correct syntax...22:58
CarlFKtumbleweed: django.core.exceptions.FieldError: Unsupported lookup 'ge' for AutoField or join on the field not permitted.23:33
CarlFKdjango.core.exceptions.FieldError: Unsupported lookup 'ge' for DateTimeField or join on the field not permitted.23:36
CarlFK 873     clients=Client.objects.annotate(   max_date=Max('show__episode__start')) \23:39

Generated by 2.13.1 by Marius Gedminas - find it at!