Merge upstream micropython master (for m_tracked_realloc and other 1.25-1.29 changes)

[andrewleech]

Why now

An open PR against lvgl/lvgl (lvgl/lvgl#10106) needs MicroPython's m_tracked_realloc, which landed upstream as commit 18735dde3becfee023f3ecf33ff36cfe1c7f5cc3 on 2026-04-08, after v1.29.0-preview was cut. The most recent tag therefore doesn't cover it, so this sync uses upstream/master (8ecd995041c2c65fd1d69f9f4b0bb763272d8f83) directly rather than a tag.

What's included

Merges micropython/micropython master at 8ecd995041c2c65fd1d69f9f4b0bb763272d8f83 into lv_micropython master. That's 2098 upstream commits since the last sync (v1.24.1, commit 06ebf74c1), spanning 1.25, 1.26, 1.27, 1.28 and the in-development 1.29.

The version bump is reflected in py/mpconfig.h: 1.24.1 → 1.29.0-pre.

Why master instead of a tag

m_tracked_realloc is not in any tag yet. Waiting for 1.29 to be cut is open-ended. The PR follows the historical pattern (merge upstream, no cherry-pick, no rebase) — just against master rather than a tag.

Conflict resolutions

15 files had conflicts. Trivial ones (additive blocks in different parts of the same file) were merged by taking both sides. The non-trivial ones:

• .gitmodules — kept upstream's added alif submodules (alif_ensemble-cmsis-dfp, alif-security-toolkit), kept the user_modules/lv_binding_micropython entry, and adopted upstream's micropython/nxp_driver URL. The previous hathach/nxp_driver URL was drift from before upstream's rename in v1.27.0 (commit bd7342d9e), not a deliberate local change.
• .gitignore, .github/workflows/ruff.yml, ports/esp32/mpconfigport.h, py/misc.h — additive merges. The local LVGL define block in mpconfigport.h (MICROPY_ENABLE_SCHEDULER, MICROPY_MODULE_BUILTIN_INIT, MICROPY_PY_SYS_SETTRACE=0) and the local XC16 branch in misc.h were preserved alongside the upstream changes.
• extmod/network_cyw43.c — pure upstream addition (STA disconnect on inactive). Other recent cyw43 fixes were cherry-picked into both sides and converged cleanly.
• py/mpconfig.h — took upstream's version (1.29.0-pre).
• ports/esp32/machine_pwm.c, machine_touchpad.c — took upstream entirely. Upstream dropped ESP-IDF <v5.2 / <v5.3 support and rewrote both files. The lv_micropython local fixes were cherry-picks of fixes that landed upstream first and are now superseded.
• ESP32 partition CSVs — upstream restructured to auto-detect flash size (commit 6201e77999), deleting most partitions-*.csv files and removing vfs entries from the survivors. Adopted upstream's layout. The lv_micropython local CSV adjustments (which enlarged app partitions for LVGL firmware) are obsolete under auto-detect. Follow-up needed: ESP32 LVGL board configs may need partition tweaks to fit the LVGL firmware size in the new layout.

Testing

Built ports/unix with VARIANT=lvgl (this variant exists specifically to exercise the binding under the unix port). Build succeeds:

LINK build-lvgl/micropython
   text	   data	    bss	    dec	    hex	filename
1994540	 236864	   3120	2234524	 22189c	build-lvgl/micropython

Smoke test: micropython -c 'import lvgl as lv; print(len(dir(lv)))' reports 560 attributes on the lvgl module — module loads, binding exposes the LVGL API surface, no missing-symbol errors.

Other ports were not built. ESP32 needs sdkconfig fix-ups for the new partition layout before its CI will be green; that's the main known regression to address as a follow-up.

---

Summary by cubic

Syncs this fork with upstream MicroPython master to bring in m_tracked_realloc and all changes up to 1.29.0-pre, unblocking the LVGL binding work and modernizing CI/docs. Conflicts were resolved to keep LVGL-specific config and user_modules/lv_binding_micropython; unix VARIANT=lvgl builds and loads lvgl successfully.

• Dependencies

  • Submodules: added Alif dependencies; switched lib/nxp_driver to micropython/nxp_driver; kept user_modules/lv_binding_micropython.
  • CI: new reusable ESP32 setup action; workflows bumped to actions/*@v6; Windows matrix simplified; code style tools updated (ruff, codespell).
  • Version: bumped to 1.29.0-pre; docs refreshed and expanded across many modules.

• Migration

  • ESP32 partitions now auto-detect flash size; upstream removed most partitions-*.csv and VFS entries. Update LVGL ESP32 board configs/partitions as needed to fit firmware.
  • ESP-IDF: oldest supported is v5.3 (CI tests v5.3 and v5.5.1). Ensure builds use a supported IDF; older IDF support was dropped upstream.

^{Written for commit 995cc14. Summary will update on new commits.}

[codecov]

Codecov Report

❌ Patch coverage is 99.57265% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 98.47%. Comparing base (da4b38e) to head (995cc14).
⚠️ Report is 4173 commits behind head on master.

Files with missing lines	Patch %	Lines
extmod/modtls_mbedtls.c	95.12%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff            @@
##           master     #106     +/-   ##
=========================================
  Coverage   98.47%   98.47%             
=========================================
  Files         153      176     +23     
  Lines       20139    22831   +2692     
=========================================
+ Hits        19832    22483   +2651     
- Misses        307      348     +41     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[cubic-dev-ai]

5 issues found across 2672 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="docs/develop/porting.rst">

<violation number="1" location="docs/develop/porting.rst:53">
P2: The porting example calls `mp_cstack_init_with_sp_here()` without including the header that declares it, so the documented `main.c` can fail to compile.</violation>
</file>

<file name="docs/library/network.PPP.rst">

<violation number="1" location="docs/library/network.PPP.rst:81">
P2: PPP.config example uses class-level call instead of instance-level call</violation>
</file>

<file name="docs/library/micropython.rst">

<violation number="1" location="docs/library/micropython.rst:70">
P3: Verbose mem_info docs hard-code a 16-byte GC block and 1KiB line size, but both depend on port/architecture and are incorrect on non-32-bit builds.</violation>
</file>

<file name=".github/actions/setup_esp32/action.yml">

<violation number="1" location=".github/actions/setup_esp32/action.yml:22">
P2: Third-party GitHub Actions are pinned to mutable tags instead of immutable SHAs.</violation>
</file>

<file name="docs/library/machine.Timer.rst">

<violation number="1" location="docs/library/machine.Timer.rst:52">
P2: Timer.init signature hard-codes `hard=True` but documentation text says default is port-specific, creating a contradiction.</violation>
</file>

_{Note: This PR contains a large number of files. cubic only reviews up to 75 files per PR, so some files may not have been reviewed. cubic prioritizes the most important files to review.
On a pro plan you can use ultrareview for larger PRs.}

[cubic-dev-ai]

P2: The porting example calls mp_cstack_init_with_sp_here() without including the header that declares it, so the documented main.c can fail to compile.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At docs/develop/porting.rst, line 53:

<comment>The porting example calls `mp_cstack_init_with_sp_here()` without including the header that declares it, so the documented `main.c` can fail to compile.</comment>

<file context>
@@ -51,7 +50,7 @@ The basic MicroPython firmware is implemented in the main port file, e.g ``main.
    int main(int argc, char **argv) {
        // Initialise the MicroPython runtime.
-       mp_stack_ctrl_init();
+       mp_cstack_init_with_sp_here(2048);
        gc_init(heap, heap + sizeof(heap));
        mp_init();
</file context>

[cubic-dev-ai]

P2: PPP.config example uses class-level call instead of instance-level call

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At docs/library/network.PPP.rst, line 81:

<comment>PPP.config example uses class-level call instead of instance-level call</comment>

<file context>
@@ -70,8 +75,11 @@ Methods
+   Sets or gets parameters of the PPP interface. The only parameter that can be
+   retrieved and set is the underlying stream, using::
+
+      stream = PPP.config("stream")
+      PPP.config(stream=stream)
 
</file context>

Suggested change

	stream = PPP.config("stream")
	PPP.config(stream=stream)
	stream = ppp.config("stream")
	ppp.config(stream=stream)

[cubic-dev-ai]

P2: Third-party GitHub Actions are pinned to mutable tags instead of immutable SHAs.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At .github/actions/setup_esp32/action.yml, line 22:

<comment>Third-party GitHub Actions are pinned to mutable tags instead of immutable SHAs.</comment>

<file context>
@@ -0,0 +1,47 @@
+
+  - name: Cached ESP-IDF install
+    id: cache_esp_idf
+    uses: actions/cache@v5
+    with:
+      path: |
</file context>

[cubic-dev-ai]

P2: Timer.init signature hard-codes hard=True but documentation text says default is port-specific, creating a contradiction.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At docs/library/machine.Timer.rst, line 52:

<comment>Timer.init signature hard-codes `hard=True` but documentation text says default is port-specific, creating a contradiction.</comment>

<file context>
@@ -4,41 +4,52 @@
 -------
 
-.. method:: Timer.init(*, mode=Timer.PERIODIC, freq=-1, period=-1, callback=None)
+.. method:: Timer.init(*, mode=Timer.PERIODIC, freq=-1, period=-1, callback=None, hard=True)
 
    Initialise the timer. Example::
</file context>

[cubic-dev-ai]

P3: Verbose mem_info docs hard-code a 16-byte GC block and 1KiB line size, but both depend on port/architecture and are incorrect on non-32-bit builds.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At docs/library/micropython.rst, line 70:

<comment>Verbose mem_info docs hard-code a 16-byte GC block and 1KiB line size, but both depend on port/architecture and are incorrect on non-32-bit builds.</comment>

<file context>
@@ -62,8 +62,33 @@ Functions
+   free.
+
+   The exact output of verbose mode varies between ports, but in general each
+   letter represents a single 16 byte block of memory. Each line of
+   output represents 0x400 bytes or 1KiB of RAM.
+
</file context>

Suggested change

	letter represents a single 16 byte block of memory. Each line of
	output represents 0x400 bytes or 1KiB of RAM.
	letter represents a single GC block of memory. Each line of
	output represents 64 GC blocks (the exact byte count depends on the port).