Silicon Valley Redneck

Another day of debugging

The hung-MCU problem?  Turned out to be some weird issue with the UART and/or its (vendor-supplied) device driver.  Now I'm detecting when UART reads fail with "receive busy" status, and de-initializing and re-initializing the UART at the driver level to clear the problem.  It seems to be working.

There remain at least two significant problems with the demon on the Linux machine, but I may have just fixed one of them.  Now I need to generate some fault conditions to see if the fix actually works.

Meanwhile, I'll run my test script overnight.  Maybe this time I'll get zero errors, instead of many errors and various wedged MCUs.

Still bashing on the debugging

Not to be confused with debugging bash(1).

Yeah... wrapped the "check for ready to do something" code with a semaphore check, and most of the errors went away.

But I'm still getting some of the weird packet-corruption (with apparently valid CRC) errors.  And I'm still getting a problem in which one of the little satellite MCUs will hang.

Both of these are rare.  The hung-MCU thing seems the most serious, as well as the most common, so I've hooked up the debugger to one of them and now have a script bashing on it to try to trigger the problem.

Update:  I hate Heisenbugs.  The communication-bashing script ran overnight, and so far has wedged all but two of the satellite units... guess which one has the debugger connected.

But at least I'm picking up a hint of Clue, as (1) there's still corrupted communication going on, (2) the corrupted communication only seems to happen when there's a script sending commands, not when the demon is just doing background polling, and (3) the satellite units only seem to get wedged when there's a script sending commands.  This suggests that (a) there's still a lurking multi-threading issue in the demon, and (b) the wedgie is caused by a corrupted message, so maybe I'd better be looking at buffer overruns or some such.

Switching among languages

Making changes to a PHP script gets frustrating when one spaces out and forgets that it isn't Ruby.

Then, remembering that it's PHP, one can become further frustrated because PHP syntax looks a lot like C, but differs in significant ways.

But it's running now.  And, judging by the status messages I'm seeing (at infrequent intervals), I need to dig, initially, into the C code for the little MCUs to see why I'm getting invalid-command errors instead of dropped packets resulting from CRC errors.

And I need to do a whole bunch of buffer-flushing somewhere in the C code for the server process on the Linux module, maybe.  Or the client code needs to sync up better.

Whoops.  Script just finished running; time to ponder the output and then go have a look at the code on the yonder side of the connection.

(It's a good thing we postponed the trash-hauling session; not only do I have a stack of work to do today, but I'm a bit under the weather.  I hope it's not some sort of virus.)

Update:  I mentioned dropped packets and CRC errors?  Well... the CRC-checking code is definitely there, and if the CRC-16 (of a message much shorter than 256 bytes) doesn't match, the code definitely ought to be dropping the packet.  Which suggests that the invalid command results from the packet being corrupted before the CRC was calculated for transmission...?  Which, basically, makes no sense, unless there's some weird memory-manglement issue and two threads are somehow writing to the same buffer.

And trying to log all the packets on all the lines... ugh.  Might be possible; I suppose I could duplicate all outgoing and incoming packets via UDP and have a listening program on the workstation, but that's yet another piece of software/firmware test framework I hadn't planned on.  (Wait, maybe I could use UDP broadcast and capture the packets using Wireshark or something?)

Come to think of it, I should really be running the test scripts on the workstation, not on the embedded system.  But the low-level logging would have to happen from the demon on the embedded system (which has limited storage, which is why I'm not just dumping everything to a log file and FTPing it back later).

This test setup currently involves 18 processors, 2 of them multi-core.  So, yes, complicated.

Update 2:  Yes indeedy, after sifting through an Imperial boatload of UDP log packets I can confirm that the packet was corrupted before transmission.  Which points to some manner of thread-synchronization issue.  I'd been thinking that was unpossible, but in fact I recently optimized things such that transmissions can be triggered by the receive threads, so I need to revisit the conditions and the thread synchronization.

Trash hauling postponed

By another week; while I've been making great progress on the sheds and whatnot, there's a lot yet to be done, and if we're trying to get all the trash hauled in one go, tomorrow isn't the day.

Besides which, Joy has been busily sifting through the trash pile and taking large amounts of stuff to recycling; the more that gets recycled, the less we have to pay to get hauled.

No really great discoveries today, unless you count a new-in-box 8" floppy drive.  And a portrait of my grandmother, which I guess I'd had in some desk drawer at IPT back in the day.  Various still-usable things went in the Goodwill pile, and an end-table (a hand-me-down from a friend in the 90s, moved from living room to shed during a reorganization probably 2004 or something) went to the "Free!" pile on the driveway.

Real Work re-appeared today, on two fronts, so I guess I'd better settle in and do some of it.  (OK, so I already looked up the licenses for Ruby and Ruby-Gnome2, both of which are compatible with what we're doing.)

Stuff-sorting for the rest of the week had better focus on clearing out the house so that remodeling can get underway as soon as the contractor is ready to start.  This implies unplugging the DSL and then the phone service... maybe converting the voice number to VOIP, just in case anyone who only has the landline number wants to get in touch with me.  More stuff to plan.  Great.

"Sexist air conditioning"

Observation: men like office temperatures kept cooler, on average, than women do.

Suggestion: in many office settings, men are required to wear standardized (and horribly hot) suits; women have more flexibility in attire.  Maybe this has something to do with it?

Solution: let men wear dresses in the office.  Heck, all they have to do is identify as women during working hours!

Semirandomness

Hearing about that new unMuppet movie, what's it called, The Murdersville Happiness or something, has me humming Heather Alexander's "Storyteller."

Well... it's a bit on the dark side and concerns Jim Henson, so obviously.  Hadn't heard it in a good many years.

Found the foo bar

'Twas in a box of IPT stuff; apparently I didn't have it on my desk at Cyras/Ciena.

Those who knew me back in the day will know.  Those who didn't, don't need to.  But have fun in the comments section (not that I'll confirm nor deny any guesses, as that would take all the mystery out of life).

I also found the LOKI prototype (NS32016-based, wire-wrap).  More nigh-forgotten career history!  I'll tell the story someday, noting that I had nothing to do with StarDOS.

Puzzling term

Been hearing lately: "It girl".

Is that something like "Bond girl"?  The one the spy, or the sewer clown, seduces at the end of the movie?

I'm not current on this Hollywood stuff.

Found my tribble!

It was in the back shed, in a box of office stuff.  With a very dead 9V battery.  Methinks it'll be needing a new battery clip, but such are easily acquired.

Also in boxes of office stuff in that shed: many books I'd thought long since misplaced.  Including a box of vintage CPU-architecture manuals; perhaps someday I'll make good on my threat to implement a PDP-10 in Verilog.  (Implement in an FPGA, with an SDRAM chip emulating many mobies of core, and an SD card emulating a room full of washing machines... anyone got an installation CD for TOPS-10?)

And my Wile E. Coyote doll, and many other things, some of which I'll keep.  Lots of stuff that gets tossed, though.

Still haven't found the foo bar, which ought to be in some office box or other.  Probably in a Ciena box, in the side shed.  When did I last see it?

Let's see... RP06 washing machine holds 178MB, formatted.  So, 5.6 washing machines to the gigabyte; a 16GB flash card holds about 90 washing machines of data.  Peak transfer rate on the RP06 is 36 bits per 5.6μs, or 6.43 Mbits/sec; a Class 10 SDHC card has a sustained write speed of at least 80 Mbits/sec, which isn't usefully comparable, really, but I suspect that in practice a moderately capable SD card would actually outperform a fair-sized array of vintage spinning-rust units.  How many disk drives was it possible to connect to a single PDP-10, anyway?

Update:  Hello!  ITS source code is on Github!  So if some fool goes and builds a PDP-10 datacenter-on-an-index-card, the software for it is out there.  Might take a kinda high density connector to bring all the terminal ports off the board, if 'tis to be used for timesharing.

Update 2:  Hmmm.  Initial thought is that I want an FPGA in a QFP, because BGAs are such a pain.  I did once design a board (painfully) with a 1mm pitch, 256-ball package (a Spartan 3E, in fact), using 6 layers, 5 mil trace & space, and 10 mil (0.25mm) minimum drill.  If memory serves, that was an expensive board to get fabbed (but the client handled that part of the project).

Decently capable FPGAs are still largely BGA things, alas, but possibly-suitable Spartan 6 parts are at least available in 1mm pitch packages, not just the fine- and superfine-pitch ones.

And... getting prototype boards done has gotten cheaper!  Seeeeeeeed quotes $433.89 for 10 pieces of a 100mm square board with 6 layers, 4 mil trace & space, and 0.2mm (8 mil) minimum drill.  Vastly more expensive than a 4-layer board, but not absurdly out of reach, and the unit price isn't too bad, in case there turned out to be a market for such a dingus.

But it's still a stupid idea.  Not that that's ever stopped me before.

Update 3:  Just for giggles, I also checked JLCPCB's pricing for a prototype board run (10 pieces) with approximately the above specs.  Ninety.  Six.  Bucks.  Wut?  That's getting into "what the heck, make the thing just because" territory.  But I don't have time right now, I'm not up to speed on DipTrace yet (and such a complex board seems excessive as a learning project), and by the time I get around to it prices will likely be even lower.  Ah: I see JLCPCB has broken out pricing for >10 runs, with (for my 6-layer example) a $159 or something setup charge plus a pittance per piece.  For 10-piece orders, it's the special prototype pricing without a setup charge as such.

(Probably can't get away with using these cut-rate furrin suppliers for ITAR stuff, especially for mil-spec deliverables, but for most of my own projects, no problem.)

Making progress on the sheds

Cleared out several boxes from the side shed, having progressed to the area wherefrom various boxes could be discarded entire.  (Not as many as I'd thought, though; a few of what I'd though of as probably-trash boxes need at least some degree of sorting through.  Exotic fabrics mixed with mundane, and suchlike.)

Also cleared out some stuff from the back shed.  I found my box of servomotors!  I kept thinking there were some capstan motors from Digi-Data tape drives around, but could only find hub motors.  Turns out the capstan motors were packed separately, along with various other servomotors and some largish steppers.

And the box with the transfer punches and T-handle hex keys.  Also the archives from the camping-lantern project.  (On which, a report someday, possibly with video.  With advances in LED and lithium-battery technology since the 1990s, the concept now seems obsolete, but 'twas a cool idea at the time.)

Defying understanding

If you think the Jewish dietary laws are simple and comprehensible, and are looking for a true challenge....

Try coming to an understanding of waste management in the Bay Area.

The rules for what goes in which bin vary wildly from one city to the next.  They also change drastically over time, sometimes with fanfare and sometimes without.

A few years back, the edict was proclaimed: anything of an electronic nature must go to e-waste recycling; putting it in the trash is forbidden.

Now?  The e-waste drop-off places are few, far between, and very selective about what they'll accept.  Transformers and car stereos are apparently now officially(?) trash, to be disposed of in the regular household garbage can.  Which got tiny earlier this year, because household trash/garbage output is now expected to be 30% food scraps by volume, and who throws out that much food/trimmings/bones?

Household HAZMAT disposal has also gotten unreasonably complicated of late.  We think we have a plan for disposing of everything all legal and proper, but I wouldn't be at all surprised to see a lot of random dumping* of paint, pesticides, and suchlike along the roads, just because it's easier and maybe less risky than trying to do it right.

Heck, even disposing of garden stepping stones is turning out to be a major headache!

... If you have a 1950s-era Junior Chem-O-Kit, do they send out the bomb squad and counterterrorism unit nowadays?

* Clarification: in general, not by us.

And yet more sorting

Mostly stuff from the side shed today.

Three good-sized boxes that turned out to contain various papers brought home from the office... except that the papers were a mixture of 1990s stuff brought home from IPT, early-2000s stuff brought home from Cyras and/or Ciena, and 2004/2005 stuff relating to JCM and/or Humball.  Some of the IPT-era stuff was more Sierra Club than work, possibly because I was (back then) working on Sierra Club stuff after hours at the office.

Most of that stuff could go away; the "keep" stuff takes up a bit over half of a small moving box (and includes some work history that I figure I ought to preserve).

Then, on to the heavy things!  A large, defunct and non-repairable oscilloscope went on the e-waste pile, along with a few lesser oddities.  (I'll need to take the dead 'scope off my capital-equipment list come tax time.)  And, several more boxes of tools'n'stuff needed to be redistributed into new moving boxes or storage totes with some reasonable weight distribution.

A well-stuffed carload of packed boxes hauled off to storage, then a trip to Home Depot for more 5-gallon storage totes, whereof I've been using rather a lot.  Hey, look!  They're on sale!  But... no matching lids?  WTF?  Wander to the area near the back of the store where the category normally belongs: proper-looking lids, but no 5-gallon bins.  A few trips back and forth, and I've established that the lids do in fact match the bins and gathered up eight of each, which, aside from being most of their stock, is the maximum number of lids that can be jammed into a bin for carrying around.

So, a busy and tiring afternoon, with a bunch of stuff gotten out of the way.  There's more to be done this weekend.  And in the coming week.  And there's still stuff that ought to be sold or given away.

Meanwhile, the house next door is now on the market, with an asking price that's kind of two years ago, and about in line with what I'd expect a fixer-upper to fetch this summer.  Is there something horribly wrong with it, that's not apparent from the outside?  Will there be a bidding war?  Stay tuned!

Ant bait

Now comes in two kinds.

Type A attracts ants, and they gather it up to feed their colonies, but it isn't usefully poisonous.

Type B may or may not be usefully poisonous, but the ants want nothing to do with it.

Ah, for the good old days when ant bait was sweet goo with arsenic.  Borax just doesn't get the job done, and the non-borax ant baits fall down on the "bait" side of things.

Afterthought:  I'd thought of mentioning the DIY option, with some silliness like "Now, where'd I leave the polonium barium nitrate?" but thought better of it.  But now, hm.  I seem to recall that my father used to have a bottle of some arsenic compound, filed under pesticides.  We haven't gotten around to dealing with that cabinet yet (probably need a combined HAZMAT run for stuff from his garage, stuff from my garage, and stuff from my bicycle garden shed).  Even if he dealt with the arsenic at some long-ago disposal opportunity, he's probably still got some good stuff hoarded out there.  (Lacking arsenic... well, methinks I saw some cadmium and selenium recently.  I wonder if some compound of one of those would work on ant colonies.  Hmph; I already disposed of my nitric acid, but I might still have some hydrochloric acid around.  And all editions of the Rubber Bible are already packed and stored.  Maybe I could come up with some lead acetate; how's that work on ants?)

Hm.  Proper pet-resistant bait container is simply an old-timey glass baby-food jar with small holes punched in the metal lid, right?

Oof.

Did more sorting and packing today.  Distributed a few hundred pounds of dense stuff among 5-gallon sturdy bins, then topped them off with lots of less-dense stuff.

What I thought I'd be doing today was mostly hauling not-so-heavy stuff out of the sheds and adding it to the trash pile, but somehow I ended up dealing with a bunch of "keep" stuff first.

Now let's see if I can get my brain in gear, 'cause there's some real work I should be doing.

Culturally transgressive lunch!

Pita bread stuffed with orange chicken.

Because leftovers.

And because the leftover rice had been in the refrigerator just that little bit too long, and I was in too much of a hurry to cook up a new pot.

Same code, different meaning

So there's this possibly-of-interest property, out there in Tennessee, that had dropped out of sight and then reappeared.  Not at all the optimal property for our purposes, but awesome in its own way, and might fit the budget and serve the purpose... even if the house is awfully fancy.

Taking a closer look at the listing, I note that it's zoned R-1.  Er.  As in Residential, 1-each single-family residence allowed?  Outside the city limits, and more'n twelve acres?  I wonder what implications that has for home-based business, monster workshop, hobby livestock and poultry, and suchlike.

So I do a little digging, and find the Blount County zoning regulations.  Pages and pages of grandfathering of things prior to 2000, mustn't upset the old-timers, fine.  Family businesses, maybe OK, but shoehorning my business into any of the approved categories gets a bit weird unless I just call it "other computer-related."

Finally, I get down to specifics.  Definitions!

R-1    Rural District 1        for medium to low density development

Oh.  Not the same as R-1 out here.  Rural District?  I can live with that.  And it doesn't sound like county officials would come around making a fuss about a few non-commercial chickens.  Though it does sound like one needs permission from the county before setting up a private airport (presumably for full-size planes, not RC models).

Wherever we think we'll be settling, I'd probably better have a chat with the Chamber of Commerce before we go making offers on property.  And maybe with the zoning authorities, if any.

Latest bad argument for Net Neutering

Making the rounds, variously: the Santa Clara County Central Fire Protection District found its ISP, Verizon, throttling its bandwidth during the massive wildfires!  And Verizon admitted having done so!

The account here is hopelessly biased, but at least does include one critical detail: the service in question is wireless Internet.  Which makes two huge differences.

First, wireless connectivity does make sense for emergency services; when I first heard the story, I was trying to understand how any of it related to Internet service in the general sense.

Second, wireless Internet service, in any given place and time, is very much a finite, shared resource.  Even without explicit throttling when users exceed some monthly soft-quota, if there are too many people trying to use the service, performance goes to crap, as I personally have observed on multiple occasions when, in unusually crowded areas, despite having a strong signal, I could get little or no bandwidth.  The mere presence of a crowd with smartphones suffices to cause the problem, given the amount of background communication done by all the bloatware on phones nowadays.

Given the second point, using regular wireless Internet service for emergency communications is an exceedingly bad idea, as an actual emergency will lead to increased traffic and severe degradation of available bandwidth (and that's assuming that the infrastructure hasn't been damaged by whatever the emergency is).

What the fire department needs is not Net Neutrality.  What it needs is exactly the opposite of Net Neutrality: it needs the ability to have emergency traffic prioritized, and everyone else in the area disconnected.

But the usual suspects are out braying that everyone should have unlimited bandwidth at all times, and no one should have to pay for all of this.  Which sounds like a lot of other things promoted by people with no comprehension of basic economics (including innumerates with economics degrees).

Additional:  The emergencies in question were out in the boonies.  For those of you who have never contemplated a map of North America, there's a lot of it, mostly very sparsely populated.  The wildfires in question were where wildfires happen, i.e., where the wild things are.  In those big areas between the urban centers, that is.  Where infrastructure is as thin on the ground as population, and even finding a cellphone signal can be problematic.

But, sure: those who have never visited the rural areas of this continent, and have no conception of the scale of things, can happily dream of uniform network buildout, giving the forests the same level of coverage as the cities.  All of this without increasing your cellphone bill to pay for it.  Yup.

Hmmm.

In today's Frazz, a suggestion for some upcoming movie: an autonomous vehicle chase scene.

Which, hmmm.

There's already been a radio-controlled-car chase scene, in one of the Dirty Harry movies.

Add some modern tech, and you've got a self-guided, hunter-seeker road torpedo!  Just needs the ability to identify the target vehicle (as well as navigating through, and under, surrounding traffic).

Update:  En route to morning walkies, I refined the road-torpedo concept into something that would make an awesome movie prop, if only I knew someone who made action-adventure movies.

On walkies, I came up with a concept for using the road-torpedo prop as the central gimmick in a tinfoil-hat video with layers of gimmickry and meta-gimmickry.  But if I currently don't have time and resources to build a proof-of-concept road torpedo, I certainly don't have the time and resources to round up people and make and distribute a video.

Ergo, filed away for maybe-next-year.

Accountability

Traditionally, corporations are accountable to their shareholders, being as how the shareholders are the owners.

Elizabeth Warren is pushing the notion that corporations should be accountable to the politicians, on behalf of various fancied "stakeholders" who aren't, technically, stakeholders.  (See also: some long-forgotten dude named "Benito".  Political control of economic activity without outright expropriation isn't exactly a new idea.)

Curiously, though, she doesn't seem upset that some of the big modern corporations aren't even accountable to their shareholders; duckduckgo "google voting shares" for an example.  Sure, they're happy to have the public at large put money in, but actual voting power must be reserved to the Insiders.  (This is also a handy way of avoiding shareholder activism.  Remember that, from the 1980s or something?)

Having one of those moments

Pondering a hypothetical project for next spring.

Comes the issue of connecting a moderately fancy inverter ($75-ish) to the corresponding motor ($60-ish) and maybe a temperature sensor.  I can visualize the sort of connector: professional-looking, environment-resistant, all that sort of good stuff.

So I start looking.  Filtering it down by connector category, data and power contact counts, and current rating on the power contacts.

The cheapest option comes in at a bit over $450.  Eep!  And that's just for the bulkhead connector; I hadn't even started looking at the cable side.

Maybe I should consider a different approach.  Like using a $16 terminal block for the power connections, and dealing with the environmental considerations in some other manner.

(The project in question keeps morphing.  Started out as simple, goofy, and totally redneck, and somehow turned into a slick technology testbed.)