10:50 <_whitenotifier-3> [glasgow] whitequark commented on pull request #671: U13 function description updated. U13 has same function as U20 (was same as U21). - https://github.com/GlasgowEmbedded/glasgow/pull/671#issuecomment-2298559643

11:26 <altracer[m]> when the alternative is desoldering a SOP8-208, clips don't suck as much -- that's how I was able to unintrusively test `bmpflash` against WeAct Pico and some routers. SI was good enough to push 5-10 MHz SPI (15?)

11:26 <altracer[m]> (BMD had a problem with Synopsys SSI of RP2040 for a while, where it was actually faster to clip onto XIP and reflash that directly, rather than manipulate it over SWD)

11:34 <whitequark[m]> I really don't mind desoldering SOP8-208...

11:34 <whitequark[m]> the problem with clips in my experience isn't SI so much as (a) poor contact and (b) boards not designed for in-circuit programming

11:34 <whitequark[m]> while (a) is solvable by holding the clip in just the right position (and even then you have to remember it at all times), (b) is a complete shot in the dark

11:38 <whitequark[cis]> it's mainly two things

11:38 <WilfriedKlaebe[m> is there some how-to design boards for in-circuit programming?

11:38 <whitequark[cis]> (a) you need to add a diode and some resistors so that if you connect a clip to the flash, it's not trying to power up the entire board (which the Glasgow won't be able to do for example)

11:39 <whitequark[cis]> (b) if it's meant to be programmed while powered on, which is useful in some circumstances, you need to add some way to stop whatever's on the board from talking to the flash, as that would totally corrupt the data

11:40 <q3k[cis]> i also have had terrible luck with clips, almost always end up desoldering the chip anyway (or at least soldering wires to leads if there's a diode)

11:40 <whitequark[cis]> I've used clips so many times and basically every time I do I regret it

11:40 <q3k[cis]> yep

11:40 <whitequark[cis]> "oh hey I can use a clip to save myself some soldering, which I really don't feel like doing" (25 minutes of suffering later) "fuck it, I'll just solder"

11:41 <q3k[cis]> which takes 3 minutes and actually works and you know it works and you stop having to second guess your tools

11:41 <whitequark[cis]> yes

11:45 <WilfriedKlaebe[m> doesn't sound too complicated...

11:45 <q3k[cis]> well, it's effort for whoever is doing the capture and layout

11:45 <WilfriedKlaebe[m> but I guess hardware vendors want to save the diode and the resistors...

11:45 <WilfriedKlaebe[m> would it hurt them to put footprints for them on the board and a cuttable trace bypassing them? one could then solder the parts when needed and cut the trace...

11:46 <whitequark[cis]> you also have to think about it

11:46 <whitequark[cis]> and these devices are not intended to be user serviceable

11:46 <whitequark[cis]> you assemble it, put it in a box, ship it, and forget about it forever

11:50 <WilfriedKlaebe[m> mmh.

11:50 <whitequark[cis]> which is probably not going to change without legislation

12:07 <altracer[m]> and then we also have abominations like RP2354 and most of ESP32-## where XIP program flash is not even pinned out... Even on Pico 1 the QSPI package is pretty tiny. Is there a way to talk to them via Glasgow or otherwise directly without desoldering?

12:10 <altracer[m]> Basically I am tangentially interested in 25x memory support. You can also read it in Quad-I/O mode now?

12:15 <whitequark[m]> you actually can't read it in quad I/O mode yet

12:15 <whitequark[m]> the underlying controller core knows how to do it, but the command set hasn't been extended yet

12:15 <whitequark[m]> for erase, there is no benefit, and for writes, maybe there is a small one? I'm unsure

12:16 <altracer[m]> okay, thanks for clarifying;

12:17 <whitequark[m]> it should be pretty trivial to extend reads, I may do this in a bit

12:29 <altracer[m]> before I crawl entire docs website and repo, can you tell me whether instead of $145 custom unshippable board with iCE40HX8k+FX2LP and GPIO frontends I can buy a $30 shippable Icebreaker 1.0e with iCE40UP5k+FT2232H and unbuffered GPIOs then use some of Glasgow applets as a cheaper devtool?

12:30 <altracer[m]> Generally I'd need a few digital PWM generators and a 1PPS generator, a few DAC channels at 19200 Hz or better; maybe a dual/quad UART with control lines and a SPI sniffer at 60 MHz (SDR, not DDR). Rare case 24c16 EEPROM, 25q64 NOR SPI Flash read/writer. I may solder some buffers and external DACs to custom PMODs for this. Otherwise the signal gen is running on WeAct STM32G474 and we use separate ft4232h & fx2la for capture (maybe

12:30 <altracer[m]> Kingston LA2016).

12:33 <whitequark[m]> the answer to this is "I actively do not wish to support any use that exists outside of the vertically integrated Glasgow stack"

12:33 <whitequark[m]> maybe you can use it, maybe you can't, you're completely on your own

12:36 <whitequark[m]> I don't get a single cent from sales one way or another, so from my perspective, more diverse use cases = more support burden and nothing else

12:41 <whitequark[m]> I'm especially frustrated with people complaining about how it's "unshippable", because (a) this is not and has never been my responsibility (it's OSHW, you can build your own if you're that unhappy about timelines), and (b) the cause is the worldwide semiconductor shortage that affected literally every industry

12:43 <q3k[cis]> (can confirm, built a few glasgows at jlcpcb)

12:48 <smeding> plus maybe the whole infrastructure required to acquire any niche product within days isn't exactly sustainable... and i gotta admit, the idea of not having so many projects that i can't do other things while i wait is pretty alien to me anyway

12:48 <smeding> uh, weird combination of negatives there, hopefully you get the idea

12:54 <altracer[m]> Understandable, have a nice day

12:57 <whitequark[cis]> for those on IRC/Matrix side, we still have a netsplit (mainly because matrix-appservice-discord is kind of garbage and I had to deal with other things recently)

12:57 <whitequark[cis]> I'll try to fix it as soon as I can

13:16 <altracer[m]> (I can't even build an orbtrace-mini locally, and it's been a year)

13:17 <whitequark[m]> oh, you're from Russia

13:18 <whitequark[m]> have you tried JLCPCB? q3k built a few Glasgow boards via JLCPCB, and I think they should still ship to Russia

13:18 <whitequark[m]> (is it "alt" for "altlinux" or just a coincidence?)

13:20 <whitequark[m]> in that case, JLC would buy the components, build the PCB, and assemble it, so you don't actually have to do much

14:45 <whitequark[cis]> why would they? they're a chinese company

14:45 <whitequark[cis]> ... oh, it looks like they actually do

14:45 <whitequark[cis]> https://www.reddit.com/r/PCB/comments/1d37su8/jlcpcb_denied_pcba_service_because_im_associated/

14:45 <whitequark[cis]> pieces of shit

14:48 <whitequark[m]> so, it looks like JLCPCB does not ship to Russia, but https://www.jlc-pcb.ru/ does

14:49 <whitequark[m]> (for those who are unaware, sanctions are in effect a regressive tax that doesn't change much for the wealthy or the MIC)

14:52 <whitequark[m]> consider:

14:52 <whitequark[m]> everyone involved in the supply chain knows this, and closes eyes on it as long as they can because it's profitable. capitalism, baby

15:05 <q3k[cis]> sanctions are such a capitalism brain move. when the worst punishment you can think of is preventing someone from buying things, because you've been telling yourself that being able to buy stuff is the peak of human experience.

15:06 <whitequark[cis]> yeah

15:07 <whitequark[cis]> having heard (second-hand, but from someone reliable) how people work around sanctions i find it that the only people they actually genuinely affect is the civilian population trying to go around their miserable daily life

15:38 <whitequark[cis]> it's a question of cost

15:39 <whitequark[cis]> if you can hire some guy to desolder them from washing machines (which by the way happens in peacetime too, and usually in China, due to lower labor costs) and you can sell them for even $0.01 more, you do it

15:40 <whitequark[cis]> buying sanctioned goods via the normal supply chain has the intermediary charge a 200-300% or something price for the effort and risk, so if you can get it for less, you do

16:41 <Jarrett[m]> boo capitalism

17:40 <adistuder[m]> I hope they are separated enough from the real JLCPCB. Would be a bit disruptive if they’d end up on the sanction list.

17:57 <whitequark[m]> I think it would be very educational, actually

17:57 <whitequark[m]> but you don't need to worry: the two are unrelated entities

17:57 <whitequark[m]> the russian one is just "borrowing" JLCPCB's name

17:58 <whitequark[m]> I'm sure the chinese JLCPCB is aware that this sort of thing is going on, but there's not much they can do except for extreme, KYC level vetting of their customers, and they obviously and appropriately don't want to do that; they do the exact bare minimum they need to not end up on sanction lists themselves

21:21 <itsmk> whitequark[m]: in a former job we were told customers sometimes sold their products without our chips, and if their customers then noticed it didn't work they'd solder in the missing controller.

21:40 <whitequark[cis]> itsmk: yeah i've heard stories like that too