Fix Github CI for v1.4-branch (project-chip#40589)

* Fix TestInetEndPoint to pass even when computer is offline (project-chip#38015)

Description:

- TestInetEndPoint was not hermetic, it required at least one network
  interface connected to a link and up.
- There was a comment before to skip tests if the link was not up,
  but it was not actually skipping the test. This made it fail for
  me when working offline and trying to run unit tests.

Testing done:
- Test passes with and without link up.

* Add debug logs for test inet limits (project-chip#38790)

* Add debug log for endpoint full error

* Also add assertions that no in use items

* Clean up includes

* Fix typo in message

* Setup ipv6 environment in `REPL Tests - Linux` CI job

This change is picked from the PR [project-chip#38845](project-chip#38845)

* Setup ipv6 environment in other places where it's needed

* Add sudo to IPV6 setup commands for proper permissions

* Revert "Add sudo to IPV6 setup commands for proper permissions"

This reverts commit 025cc0c.

* Remove `add-apv6` action from `Build on Linux (python_lib)`

This doesn't work because of permission issues.

* Stop using old ubuntu version (project-chip#37870)

Co-authored-by: Andrei Litvin <[email protected]>

---------

Co-authored-by: Tennessee Carmel-Veilleux <[email protected]>
Co-authored-by: Andrei Litvin <[email protected]>
Co-authored-by: Andrei Litvin <[email protected]>

Author: 4 people

.github/actions/add-ipv6/action.yaml

@@ -0,0 +1,18 @@
+name: Set up an IPV6 environment
+description: Creates a socket pair with a "peer" network namespace that supports IPV6
+runs:
+  using: "composite"
+  steps:
+    - name: Set up IPV6
+      if: ${{ runner.os == 'Linux' }}
+      shell: bash
+      run: |
+          ip netns add peer
+
+          ip link add name veth0 type veth peer name veth0-peer
+          ip -6 addr add fd00:0:1:1::1/64 dev veth0
+          ip link set dev veth0 up
+
+          ip link set veth0-peer netns peer
+          ip netns exec peer ip -6 addr add fd00:0:1:1::2/64 dev veth0-peer
+          ip netns exec peer ip link set dev veth0-peer up

.github/workflows/build.yaml

@@ -66,6 +66,8 @@ jobs:
               run: |
                   mkdir /tmp/cores || true
                   sysctl -w kernel.core_pattern=/tmp/cores/core.%u.%p.%t || true
+            - name: Set up a IPV6 known environment
+              uses: ./.github/actions/add-ipv6
             - name: Checkout submodules & Bootstrap
               uses: ./.github/actions/checkout-submodules-and-bootstrap
               with:
@@ -162,6 +164,8 @@ jobs:
               run: |
                   mkdir /tmp/cores || true
                   sysctl -w kernel.core_pattern=/tmp/cores/core.%u.%p.%t || true
+            - name: Set up a IPV6 known environment
+              uses: ./.github/actions/add-ipv6
             - name: Checkout submodules & Bootstrap
               uses: ./.github/actions/checkout-submodules-and-bootstrap
               with:
@@ -468,6 +472,10 @@ jobs:
         steps:
             - name: Checkout
               uses: actions/checkout@v4
+
+            - name: Set up a IPV6 known environment
+              uses: ./.github/actions/add-ipv6
+
             - name: Checkout submodules & Bootstrap
               uses: ./.github/actions/checkout-submodules-and-bootstrap
               with:

.github/workflows/tests.yaml

@@ -78,6 +78,9 @@ jobs:
                   sysctl -w kernel.core_pattern=/tmp/cores/core.%u.%p.%t || true
                   mkdir objdir-clone || true
 
+            - name: Set up a IPV6 known environment
+              uses: ./.github/actions/add-ipv6
+
             - name: Validate that xml are parsable
               # The sub-items being run here are the same as the input XMLs listed
               # at src/app/zap-templates/zcl/zcl.json
@@ -473,6 +476,9 @@ jobs:
                   sysctl -w kernel.core_pattern=/tmp/cores/core.%u.%p.%t || true
                   mkdir objdir-clone || true
 
+            - name: Set up a IPV6 known environment
+              uses: ./.github/actions/add-ipv6
+
             - name: Build Python REPL and example apps
               # NOTE: the data-mode-check + check-failure-die is not 100% perfect as different
               #       encoding sizes for data that keeps changing may alter over time (e.g. anything relating to time

.github/workflows/zap_regeneration.yaml

@@ -28,7 +28,7 @@ jobs:
     zap_regeneration:
         name: ZAP Regeneration
 
-        runs-on: ubuntu-20.04
+        runs-on: ubuntu-latest
         container:
             image: ghcr.io/project-chip/chip-build:81
         defaults:

.github/workflows/zap_templates.yaml

@@ -33,7 +33,7 @@ jobs:
     zap_templates:
         name: ZAP templates generation
 
-        runs-on: ubuntu-20.04
+        runs-on: ubuntu-latest
         container:
             image: ghcr.io/project-chip/chip-build:81
         defaults:

src/inet/tests/TestInetEndPoint.cpp

@@ -33,7 +33,9 @@
 
 #include <CHIPVersion.h>
 #include <inet/IPPrefix.h>
+#include <inet/InetConfig.h>
 #include <inet/InetError.h>
+#include <inet/UDPEndPoint.h>
 #include <lib/core/StringBuilderAdapters.h>
 #include <lib/support/CHIPArgParser.hpp>
 #include <lib/support/CHIPMem.h>
@@ -277,8 +279,13 @@ TEST_F(TestInetEndPoint, TestInetEndPointInternal)
 
     err = InterfaceId::Null().GetLinkLocalAddr(&addr);
 
-    // We should skip the following checks if the interface does not have the Link local address
-    ASSERT_NE(err, INET_ERROR_ADDRESS_NOT_FOUND);
+    // We should skip the following checks if the interface does not have the Link local address.
+    // This can happen if you don't have network interfaces connected to any link (like happened
+    // to the author of this comment at YYZ before CSA 2025 Chicago Member Meeting).
+    if (err == INET_ERROR_ADDRESS_NOT_FOUND)
+    {
+        return;
+    }
 
     EXPECT_EQ(err, CHIP_NO_ERROR);
     intId = InterfaceId::FromIPAddress(addr);
@@ -381,12 +388,29 @@ TEST_F(TestInetEndPoint, TestInetEndPointLimit)
 
     CHIP_ERROR err = CHIP_NO_ERROR;
 
+    // we assume NO open endpoints
+    gUDP.ForEachEndPoint([](UDPEndPoint * ep) {
+        ChipLogError(Test, "NOTE: Unexpected UDP endpoint in use. Will free it");
+        ep->Free();
+        return Loop::Continue;
+    });
+
     int udpCount = 0;
     SYSTEM_STATS_RESET(System::Stats::kInetLayer_NumUDPEps);
     for (int i = INET_CONFIG_NUM_UDP_ENDPOINTS; i >= 0; --i)
     {
         err = gUDP.NewEndPoint(&testUDPEP[i]);
-        EXPECT_EQ(err, (i ? CHIP_NO_ERROR : CHIP_ERROR_ENDPOINT_POOL_FULL));
+
+        CHIP_ERROR expected_error = (i ? CHIP_NO_ERROR : CHIP_ERROR_ENDPOINT_POOL_FULL);
+        if (err != expected_error)
+        {
+            // have a log to debug things
+            ChipLogError(Test, "UDP: Failure on index %d: (out of %d)", i, INET_CONFIG_NUM_UDP_ENDPOINTS);
+
+            // this will fail after the above log
+            EXPECT_EQ(err, expected_error);
+        }
+
         if (err == CHIP_NO_ERROR)
         {
             ++udpCount;
@@ -397,12 +421,27 @@ TEST_F(TestInetEndPoint, TestInetEndPointLimit)
     EXPECT_TRUE(SYSTEM_STATS_TEST_HIGH_WATER_MARK(System::Stats::kInetLayer_NumUDPEps, udpHighWaterMark));
 
 #if INET_CONFIG_ENABLE_TCP_ENDPOINT
+    // we assume NO open endpoints
+    gTCP.ForEachEndPoint([](TCPEndPoint * ep) {
+        ChipLogError(Test, "NOTE: Unexpected TCP endpoint in use. Will free it");
+        ep->Free();
+        return Loop::Continue;
+    });
+
     int tcpCount = 0;
     SYSTEM_STATS_RESET(System::Stats::kInetLayer_NumTCPEps);
     for (int i = INET_CONFIG_NUM_TCP_ENDPOINTS; i >= 0; --i)
     {
-        err = gTCP.NewEndPoint(&testTCPEP[i]);
-        EXPECT_EQ(err, (i ? CHIP_NO_ERROR : CHIP_ERROR_ENDPOINT_POOL_FULL));
+        err                       = gTCP.NewEndPoint(&testTCPEP[i]);
+        CHIP_ERROR expected_error = (i ? CHIP_NO_ERROR : CHIP_ERROR_ENDPOINT_POOL_FULL);
+        if (err != expected_error)
+        {
+            // have a log to debug things
+            ChipLogError(Test, "TCP: Failure on index %d: (out of %d)", i, INET_CONFIG_NUM_TCP_ENDPOINTS);
+
+            // this will fail after the above log
+            EXPECT_EQ(err, expected_error);
+        }
         if (err == CHIP_NO_ERROR)
         {
             ++tcpCount;