ESP8266 Arduino Core v2.6.0

[OttoWinter]

Arduino Core v2.6.0 for ESP8266 was just released: https://github.com/esp8266/Arduino/releases/tag/2.6.0

There appear to be quite a few bugfixes that may help stability.

Some highlights:

• ets_intr_lock nested; from description appears to fix ESP8266 getting unreachable after a while - Ets intr lock nest esp8266/Arduino#6484, Esp8266 IP Address not reachable after a while esp8266/Arduino#2330

• Add option for slower flash chips, behavior seems to kinda be similar to the infamous ESP8266 Various Connection Issues issues#455; plus the fact that it depends on fw size/alignment would finally make sense; idk, don't want to get my hopes up too much yet - Add 20/26MHz Flash frequencies for slow/cheap flash chips on the Generic ESP Board esp8266/Arduino#6552

  • This appears to be an opt-in. How do we make this user-settable? Maybe we can do something like with flash mode: Manual upload is always with worst (dout) setting, but any subsequent OTA auto-overwrites the flash mode to the correct one.

• i2c transaction fix: Double I2C read in one transaction skips a clock pulse (#5528) esp8266/Arduino#6654

• NONOS-SDK updated to 2.2.x, and appears to be the new default; TODO evaluate stability, check tools/platformio-build.py - nonos-sdk: update to latest version 2.2.x branch, commitlog added in tree esp8266/Arduino#6257, Switch default FW to "2.2.2-dev(38a443e)" (menu:2.2.1+100) esp8266/Arduino#6272, FW: use NONOS-SDK branch 2.2.x from 2019-10-24 by default esp8266/Arduino#6724

Potential changes needed for:

• toolchain update - docker images - Upgrade to 2.5.0-4 toolchain w/improved pgm_read_x, C++ exceptions esp8266/Arduino#6273
• timezone handling, upstream now has its own copy of IANA database. But with ESPHome that part is handled by python generator, maybe that messes with the timezone code. time: import IANA timezone definitions, expose SNTP API esp8266/Arduino#6373
• attachInterruptArg is now public with a changed signature, probably esphal.cpp needs to be updated. - Bugfix: attach interrupt (#6049) esp8266/Arduino#6048
• LEAmDNS can handle multiple net interfaces now, we can remove the hack in util.cpp - mDNS: restriction to a single interface esp8266/Arduino#6224
• PSTR macro updated, maybe needs some changes to logger - Move all PSTRs to own section, allow string dedup esp8266/Arduino#6565

May want to copy some bugfixes downstream:

• WiFi STA static IP fix, not sure what exactly/why yet - lwip2: (re)fix setting static ip address esp8266/Arduino#6194
• Changes to Software Serial - https://github.com/esp8266/Arduino/pull/6722/files

Other useful stuff:

• precache() (like prefetch for instruction memory) - precache() - preload code into the flash cache. esp8266/Arduino#6628
• umm_malloc update - Update core with upstream umm_malloc esp8266/Arduino#6438
• NAT support; looks like wifi repeating is easy now; TODO evaluate stability and RAM usage - new network feature: NAPT (widely known as NAT) esp8266/Arduino#6360

Misc:

• getCycleCount forced to be inline (means that micros() might have had an issue in ISR context before) - Inline ESP::getCycleCount() to make it safe to call from ISRs esp8266/Arduino#6477
• memory fence for xt_rsil (disabling interrupts) - Added memory fence to xt_rsil(). Without this the compiler may use esp8266/Arduino#6301
  • Related: We should probably change the disable_interrupts()+enable_interrupts() function to be a scoped lock like esp8266::InterruptLock. Makes code cleaner
• Fix for random LEAmDNS crash on startup - MDNS: fix random crash on startup esp8266/Arduino#6261

[brandond]

esp8266/Arduino#6552 is actually super interesting if you read through the whole thread from the original issue in esp8266/Arduino#6366.

The issue is not that low-end flash chips are slow - the issue is that cheap boards are poorly laid out, and at higher TX powers the RF interference actually causes the flash to misread, resulting in all kinds of bad things including executing instructions that send the pointer to bad offsets in the stack. I wonder if it might also be affecting writes and corrupting flash - that wasn't tested in the thread.

The fix was to either reduce TX power, use an external antenna, or reduce the flash frequency: esp8266/Arduino#6366 (comment). ESPHome doesn't currently expose wifi power settings, but this might be a good reason to do so. See: esp8266/Arduino#764 (comment)

[brandond]

esp8266/Arduino#6674 (comment) is also interesting

[Marlor]

Just a heads-up, Core 2.6.2 just got released and is in platform-espressif8266 2.3.1