Here's my first attempt at a workaround for KB5127936.
To install, extract this ZIP and copy the 3 EXE files to C:\Users\{username}\AppData\Local\Arduino15\packages\teensy\tools\teensy-tools\1.62.0\, replacing the copies from 1.62.
I changed...
All four of my computers are now working, all are running Windows 11 Pro, versions 24H2, 25H2 and 26H2. "KB5127936 workaround used" is the message on them all. Thats a big thumbs up, thank you Paul.
I'm going to add a comment without knowing whether it's relevant or not: when listening on a socket, would adding the SO_REUSEADDR flag help? (Shot in the dark.)
I meet the same yesterday, both arduino 2.0 and 1.8 can not work with loader,
then i removed KB5127936, KB5127694 - the windows patch,
it can fix the problem.
This question on Microsoft's site has an answer to uninstall those 2 updates, but also has an accepted answer that just updating Windows Defender may also fix the problem.
Oh my, looks like Microsoft may have published a buggy update that breaks localhost. :(
https://learn.microsoft.com/en-us/answers/questions/5585563/localhost-not-working-anymore-after-2025-10-cumula
People in that conversation are reporting...
Figured I'd share this long-running and finally finished project based on the Teensy 4.1 + Audio Shield. Making V1 was my first Teensy project ever, and I'm happy with how this one turned out. Hope you dig!
The writeup in the list of links at...
It works fine as a host with an OTG adapter (especially handy for using USB devices with the T4.0).
Using a USB-C to USB-A adapter would probably work around the pulldown problem, but half the attraction of having a USB-C port directly on the...
If a client is sending only fixed-sized messages and available() shows there is enough data available to fill one, I should be able to pass a pointer to a message directly to read(). A cast is necessary to uint8_t* but never required for void*.
I'm curious what types of memory you're reading into and how you handle partial reads into something that isn't an unsigned char, i.e. the element type is bigger? Wouldn't you need to cast anyway? Do you think you could give me a code example...
Here are the measurements you asked for, all on the same board: Teensy 4.1, 2 x APS6404L (8 MB + 8 MB = 16 MB), core 1.62.0 with your startup.c v2 installed (at boot FLSHA1CR0=0x2000, FLSHA2CR0=0x8000, MPU region 0x70000000 = 32 MB). F_CPU 720...
There was a reason I added the void* output methods but not equivalent input methods. Now what was it again... maybe I just felt at the time that the semantic difference was important or something.
I ran jmarsh's alternating PLD test on my board (Teensy 4.1, 2 x 8 MB APS6404L = 16 MB, core 1.62.0, 720 MHz). It reproduces the 2^30-cycle stall with the stock core, and the three fixes remove it. The MPU variant does not fault.
Test
jmarsh's...
I've added this already (locally), however, I discovered an existing race condition during testing between peek, read, and the remote connection closing. Working on solving that and adding some tests for the condition.
More details:
There are...
Found it: it is the same speculative-read problem, and in my case the trigger is micros()
Thanks to everyone, in particular to the thread by SamiH (the AHBBUSWAIT explanation: 0xFFFF x 1024 cycles x 16 = 2^30 cycles) which made me look at the...
Looking at the startup code I see the reason for this - when only one chip is present, the FLEXSPI2_FLSHA2CR0 register is never touched and keeps its default value (64MB). So in that case FlexSPI doesn't think the memory is "unbacked" and never...
Update: the MPU region size is not the cause.
I ran the same sketch object (the sketch was not recompiled, only the core) with the stock SIZE_32M and with SIZE_16M for the FlexSPI2 region. I confirmed the active setting by reading the MPU back...
I meant: Could you check the timing with that branch? I wasn't asking for a comprehensive test; just timing with the SimpleHTTPClient example. :)
I want to see if some of the memory stuff and fixes I've been doing change your measurements.
@wwatson could you try out the branch I just pushed, for your wired Ethernet tests? `fix-memory-and-stuff` (39ff9dece4). It has some memory usage fixes and improvements (eg. churn) (along with some other fixes and improvements — see the...
Remember that you're also testing traffic that goes through the Internet, so a high data rate isn't guaranteed.
But yes, it is interesting how the Wi-Fi seems always double the speed of the wired ethernet, in your tests.
A basic test:
Setup: Teensy 4.1 plugged via Ethernet into a non-gateway eero node (I.e. the node is wireless)
Versions: Teensyduino 1.62, QNEthernet 0.37.0
Result: Received 93090 bytes in 0.148662 s = ~5.01 Mbps
I just did a quick check. Good news—nothing significant was found.
However, the AI is nitpicking about a few minor issues, including the fact that the default constructor only partially initializes the variables. `resetHistory` is called, but the...
The personal attacks are an ad hominem fallacy.
My username, alleged behavior years ago, and hobbies are irrelevant to the technical discussion.
That "user error" is in the core, too.
Edit:
It takes no more than about 5 seconds for anyone to...
Over the years, the Teensy platform set the gold standard for microcontrollers, driven by both exceptional hardware and a vibrant community that actively shaped the core.
Looking at recent releases, however, updates seem heavily dominated by...
I'll add: I'm working through some fixes and memory-related fixes and improvements, eg. to reduce churn. After that, I'll be diving into a WiFi driver, and then the multi-interface stuff.
MemoPan combines a handpan-inspired playing interface with a musical memory game inspired by Simon. It is designed to make music accessible, encourage musical exploration, and help children develop their listening and memory skills.
Features...
No Problem ;)
Just take it as a hint that the compiler is telling you where to look. Regardless of the microcontroller.
Well, the warnings do serve a purpose. That's the only reason they exist :)
I even use -Werror. That way, warnings are...
Friends,
Here is my latest tutorial, which brings on-the-fly DDS synthesis (16-bit) to port IO on the Teensy 4.1. It can go surprisingly fast, and through careful testing, I show that the arm_sin_q31() function is the fastest way to compute the...
Maybe there's some custom things going on in @jmarsh's system, but I read it as "perhaps there's aggressive caching." I'll get the obvious question out of the way: have you tried deleting all its caches or custom settings directories? I wonder...
bing search led me to this by some " Frank B " :)
https://forum.pjrc.com/index.php?threads/discussion-about-a-simple-way-to-change-the-sample-rate.38753/
No, no changes to the core were necessary. It was possible at runtime. No problem.
@defragster "the master of forum searches" will find it.
If not, I’ll take a look at my old code.
The audio board is not only working with a Teensy but also with an ESP32P4, and than we can use with a sample rate of 96KHz.
Yes yes I know it's not a Teensy but I like to say that the PRJC audio adapter is a awesome device.
Best,
Johan
I've been keeping the no-driver-constants QNEthernet branch up to date to accommodate an unknown initial MAC address. Would you mind testing it with your code? (I don't have the hardware to play with.)
I wrote a new driver (using the same "wrapper" method as before) for ASIX88772 USB to Ethernet adapters: repo here which uses the base USB driver here.
I ended up "faking" some of the methods, such as is_unknown() returning constant false and...
Thank you Frank…. I think :)
And to be fair… The ESP32 is also not needed for the Internet Radio to run at my desk since it can just run from a browser window. But of course, that is not why any of us are on this forum.
To state what I think...