temp: trace the dtls state machines

Author: aidangarske

.github/workflows/_resolve-wolfssl.yml

@@ -67,8 +67,13 @@ jobs:
             if [[ "$ref" =~ ^[0-9a-f]{40}$ ]]; then
               sha="$ref"
             else
+              # An annotated tag resolves to the tag object, not the commit, so
+              # ask for the peeled ref first and fall back for branches.
               sha=$(git ls-remote https://github.com/wolfSSL/wolfssl.git \
-                    "$ref" "refs/tags/$ref^{}" | head -n1 | cut -f1)
+                    "refs/tags/$ref^{}" | cut -f1)
+              [ -n "$sha" ] || sha=$(git ls-remote \
+                    https://github.com/wolfSSL/wolfssl.git "$ref" \
+                    | head -n1 | cut -f1)
             fi
             [ -n "$sha" ] || { echo "could not resolve wolfSSL ref '$ref'"; exit 1; }
             shas="${shas:+$shas,}$sha"

.github/workflows/diag.yml

@@ -16,11 +16,11 @@ jobs:
     steps:
       - uses: actions/checkout@v5
 
-      - name: Build wolfSSL (dtls profile)
+      - name: Build wolfSSL (dtls profile + debug)
         run: |
           set -euo pipefail
           sudo apt-get update -qq
-          sudo apt-get install -y --no-install-recommends gdb >/dev/null
+          sudo apt-get install -y --no-install-recommends gdb strace >/dev/null
           git clone -q --depth 1 https://github.com/wolfSSL/wolfssl /tmp/wolfssl
           cd /tmp/wolfssl
           ./autogen.sh >/dev/null 2>&1
@@ -31,56 +31,43 @@ jobs:
             --enable-static --enable-shared --enable-debug \
             CFLAGS="-DWOLFSSL_DTLS13_NO_HRR_ON_RESUME -g -O0" >/dev/null
           make -j"$(nproc)" >/dev/null && make install >/dev/null
-          echo "built"
+          echo built
 
-      - name: memory-bio-dtls -- where are the threads stuck
+      - name: memory-bio-dtls -- wolfSSL state machine trace
         run: |
           set -uo pipefail
           cd dtls
-          make memory-bio-dtls CFLAGS="-I/tmp/inst/include -g -O0" \
-            LDFLAGS="-L/tmp/inst/lib" 2>&1 | tail -3 || \
-            gcc -g -O0 -o memory-bio-dtls memory-bio-dtls.c \
-              -I/tmp/inst/include -L/tmp/inst/lib -lwolfssl -lpthread -lm
+          # flip the example's own `#if 0 wolfSSL_Debugging_ON()` on
+          sed -i '143s/#if 0/#if 1/' memory-bio-dtls.c
+          sed -n '142,145p' memory-bio-dtls.c
+          gcc -g -O0 -o mb memory-bio-dtls.c -I/tmp/inst/include \
+            -L/tmp/inst/lib -lwolfssl -lpthread -lm || exit 1
           export LD_LIBRARY_PATH=/tmp/inst/lib
-          stdbuf -oL -eL ./memory-bio-dtls > /tmp/mb.out 2>&1 &
-          PID=$!
-          sleep 15
-          if kill -0 $PID 2>/dev/null; then
-            echo "=== STILL RUNNING after 15s -- thread stacks:"
-            sudo gdb -batch -p $PID -ex 'thread apply all bt' 2>/dev/null \
-              | grep -vE '^\[|^Reading|^Downloading|warning:' | head -60
-            kill -9 $PID
-          else
-            echo "=== exited on its own, rc=$?"
-          fi
-          echo "=== stdout so far:"
-          cat /tmp/mb.out
+          timeout 12 stdbuf -oL -eL ./mb > /tmp/mb.out 2>&1
+          echo "rc=$?"
+          echo "=== total debug lines: $(wc -l < /tmp/mb.out)"
+          echo "=== FIRST 40:"; head -40 /tmp/mb.out
+          echo "=== LAST 60 (where it spins):"; tail -60 /tmp/mb.out
 
-      - name: dtls-nonblocking -- where are the two sides stuck
+      - name: dtls-nonblocking -- state machine + syscalls
         run: |
           set -uo pipefail
           cd dtls
           export LD_LIBRARY_PATH=/tmp/inst/lib
           for t in server-dtls-nonblocking client-dtls-nonblocking; do
-            gcc -g -O0 -o "$t" "$t.c" -I/tmp/inst/include -L/tmp/inst/lib \
-              -lwolfssl -lm 2>&1 | head -3
+            sed -i 's/wolfSSL_Init();/wolfSSL_Init(); wolfSSL_Debugging_ON();/' "$t.c"
+            gcc -g -O0 -o "${t%%-dtls*}X" "$t.c" -I/tmp/inst/include \
+              -L/tmp/inst/lib -lwolfssl -lm || exit 1
           done
-          stdbuf -oL -eL ./server-dtls-nonblocking > /tmp/s.out 2>&1 &
-          SPID=$!
+          strace -f -e trace=sendto,recvfrom,recvmsg,sendmsg,connect,bind \
+            -o /tmp/s.strace stdbuf -oL -eL ./serverX > /tmp/s.out 2>&1 &
           sleep 2
-          echo "hello" | stdbuf -oL -eL ./client-dtls-nonblocking 127.0.0.1 > /tmp/c.out 2>&1 &
-          CPID=$!
-          sleep 15
-          for n in server:$SPID client:$CPID; do
-            name=${n%%:*}; pid=${n##*:}
-            if kill -0 $pid 2>/dev/null; then
-              echo "=== $name STILL RUNNING -- stack:"
-              sudo gdb -batch -p $pid -ex 'bt' 2>/dev/null \
-                | grep -vE '^\[|^Reading|^Downloading|warning:' | head -25
-            else
-              echo "=== $name already exited"
-            fi
-          done
-          kill -9 $SPID $CPID 2>/dev/null || true
-          echo "=== server out:"; cat /tmp/s.out
-          echo "=== client out:"; cat /tmp/c.out
+          echo hello | timeout 12 strace -f -e trace=sendto,recvfrom,recvmsg,sendmsg,connect \
+            -o /tmp/c.strace stdbuf -oL -eL ./clientX 127.0.0.1 > /tmp/c.out 2>&1
+          echo "client rc=$?"
+          sleep 1
+          pkill -9 serverX 2>/dev/null || true
+          echo "=== SERVER wolfSSL trace (last 45):"; tail -45 /tmp/s.out
+          echo "=== SERVER syscalls (first 25):"; head -25 /tmp/s.strace
+          echo "=== CLIENT wolfSSL trace (last 45):"; tail -45 /tmp/c.out
+          echo "=== CLIENT syscalls (first 25):"; head -25 /tmp/c.strace