Files
threadx/scripts/check_ports.sh
T
Frédéric Desbiens 08b120d2bc Added port consistency checks and wired them into CI and release preparation (#591)
Three defects reached the repository through the port trees recently, and each
of them is mechanically detectable without a cross compiler. Add
scripts/check_ports.sh, which looks for exactly those three, and give CI and
the release process the same command a contributor can run locally.

The generated Cortex-M ports must be reproducible from ports_arch. Fixes were
applied to the generated copies instead of the source for eight months, and the
next run of the copy scripts would have reverted them.

Preprocessor directives must balance. A fix left the Cortex-M85 IAR tx_port.h
with one more #endif than #if, so that header could not compile.

No port header may carry a statement outside a function body. A fix left a
second, headerless copy of a function body in the Cortex-M4 AC6 tx_port.h,
which is issue 569. The check tracks brace depth while skipping preprocessor
lines, multi-line macro bodies and comments, and reports assignments,
dereferences and control statements that land at file scope. Headers under
example_build are excluded, since those trees vendor third party SDK code.

A fourth section reports, without failing the run, on port families that have
no copy script and so cannot be checked for reproducibility. It currently
observes that the Cortex-M0 ac5, ac6 and keil ports lack the barriers their gnu
and iar siblings have.

ports_arch_check now calls the script rather than inlining a copy and diff, so
CI and the command line check the same things by the same definition, and the
workflow now triggers on pull requests to dev as well as master. Triggering on
master alone is why the drift went unseen. prepare_release.sh runs the checks
before it branches or rewrites anything, and stops if they fail, with
SKIP_PORT_CHECKS=1 as the escape hatch.

Each check was verified by reintroducing the defect it exists to catch and
confirming that the script fails, then confirming it passes on a clean tree.

Assisted-by: Claude Code (Opus 5) <noreply@anthropic.com>
2026-08-09 10:48:29 -04:00

225 lines
8.6 KiB
Bash
Executable File

#!/bin/bash
##############################################################################
# Copyright (C) 2026 Eclipse ThreadX contributors
#
# This program and the accompanying materials are made available under the
# terms of the MIT License which is available at
# https://opensource.org/licenses/MIT.
#
# AI Disclosure: This file was largely AI-generated by Claude Code (Opus 5).
# The AI-generated portions may be considered public domain (CC0-1.0)
# and not subject to the project's licence. The human contributor has
# reviewed and verified that the code is correct.
#
# SPDX-License-Identifier: MIT and CC0-1.0
##############################################################################
# Consistency checks for the port trees. Run it before cutting a release, or
# any time the ports have been touched:
#
# scripts/check_ports.sh
#
# Options:
# --no-regen Skip the reproducibility check, which needs a clean tree.
# --quiet Print only failures and the summary.
#
# Exit status is 0 when every check passes and 1 otherwise, so the same
# command serves CI and the command line.
#
# Each check exists because a real defect reached the repository through it:
#
# 1. Reproducibility. The Cortex-M ports are generated by the copy scripts,
# but fixes were applied to the generated copies instead of the source for
# eight months. The next run of those scripts would have reverted them.
#
# 2. Preprocessor balance. A fix left ports/cortex_m85/iar/inc/tx_port.h with
# one more #endif than #if, so that header could not compile.
#
# 3. Code at file scope. A fix left a second, headerless copy of a function
# body in ports/cortex_m4/ac6/inc/tx_port.h, which placed statements
# outside any function. See issue 569.
#
# The last section reports, without failing, on port families that have no copy
# script and so cannot be checked for reproducibility.
#
# Headers under example_build are skipped: those trees vendor third party SDK
# code, which is not ours to hold to these rules.
set -u
cd "$(dirname "$(realpath "$0")")/.."
regen=1
quiet=0
for arg in "$@"; do
case "$arg" in
--no-regen) regen=0 ;;
--quiet) quiet=1 ;;
-h|--help) sed -n '17,30p' "$0"; exit 0 ;;
*) echo "Unknown option: $arg" >&2; exit 2 ;;
esac
done
failures=0
say() { [ "$quiet" -eq 1 ] || echo "$@"; }
fail() { echo " FAIL: $*"; failures=$((failures + 1)); }
# --------------------------------------------------------------------------
# 1. The generated ports must be reproducible from ports_arch.
# --------------------------------------------------------------------------
say ""
say "== Generated ports are reproducible from ports_arch =="
if [ "$regen" -eq 0 ]; then
say " skipped (--no-regen)"
elif ! command -v git >/dev/null 2>&1; then
say " skipped (git not available)"
elif [ -n "$(git status --porcelain -uno)" ]; then
fail "the working tree has uncommitted changes; commit or stash them, or pass --no-regen"
else
./scripts/copy_armv7_m.sh >/dev/null 2>&1
./scripts/copy_armv8_m.sh >/dev/null 2>&1
./scripts/copy_module_armv7_m.sh >/dev/null 2>&1
drift="$(git status --porcelain -uno)"
if [ -n "$drift" ]; then
fail "running the copy scripts changed $(echo "$drift" | wc -l) file(s)."
echo " The ports below were edited directly instead of through ports_arch."
echo " Re-apply the change to ports_arch and re-run the copy scripts."
echo "$drift" | sed 's/^/ /'
git checkout -- . >/dev/null 2>&1
else
say " ok: the copy scripts change nothing"
fi
fi
# --------------------------------------------------------------------------
# 2. Preprocessor directives must balance in every port header.
# --------------------------------------------------------------------------
say ""
say "== Preprocessor directives balance =="
unbalanced=0
while IFS= read -r f; do
result="$(awk '
/^[ \t]*#[ \t]*(if|ifdef|ifndef)/ { depth++ }
/^[ \t]*#[ \t]*endif/ {
depth--
if (depth < 0 && first == 0) { first = NR }
}
END { printf "%d %d", depth, first }
' "$f")"
depth="${result% *}"
first="${result#* }"
if [ "$depth" -ne 0 ]; then
if [ "$first" -ne 0 ]; then
fail "$f: #endif without a matching #if at line $first (final depth $depth)"
else
fail "$f: $depth unterminated #if (final depth $depth)"
fi
unbalanced=$((unbalanced + 1))
fi
done < <(find ports ports_arch ports_module ports_smp -name "*.h" -type f \
! -path "*/example_build/*" 2>/dev/null | sort)
[ "$unbalanced" -eq 0 ] && say " ok: every port header balances"
# --------------------------------------------------------------------------
# 3. No statements outside a function body in a port header.
# --------------------------------------------------------------------------
# Tracks brace depth, ignoring preprocessor lines, comments and strings, and
# reports any statement that lands at depth zero. Declarations, typedefs,
# externs and macro definitions are expected there; assignments, calls and
# dereferences are not, and an orphaned function body shows up as exactly that.
say ""
say "== No code at file scope in port headers =="
orphans=0
while IFS= read -r f; do
result="$(awk '
# Skip preprocessor lines, including multi-line macro bodies, whose
# continuations are statements by design.
/^[ \t]*#/ { if (/\\[ \t]*$/) { in_macro = 1 }; next }
in_macro { if (!/\\[ \t]*$/) { in_macro = 0 }; next }
# Skip comments, both kinds, including continuation lines.
in_comment { if (/\*\//) { in_comment = 0 }; next }
/^[ \t]*\/\// { next }
/\/\*/ { if (!/\*\//) { in_comment = 1 }; next }
{
line = $0
gsub(/"[^"]*"/, "", line)
gsub(/\/\/.*$/, "", line)
stripped = line
gsub(/^[ \t]+|[ \t]+$/, "", stripped)
# A statement at file scope ends in a semicolon and either assigns,
# dereferences or opens a control structure. Declarations and
# prototypes also end in a semicolon but match none of these.
if (depth == 0 && stripped ~ /;[ \t]*$/ &&
(stripped ~ /^\*\(/ ||
stripped ~ /^[A-Za-z_][A-Za-z0-9_]*([ \t]*(\[[^]]*\]|->[ \t]*[A-Za-z_][A-Za-z0-9_]*|\.[A-Za-z_][A-Za-z0-9_]*))*[ \t]*=[^=]/ ||
stripped ~ /^(return|if|while|for|switch|do)[ \t(]/)) {
print NR ": " stripped
found++
}
n = gsub(/{/, "{", line); m = gsub(/}/, "}", line)
depth += n - m
if (depth < 0) { depth = 0 }
}
END { exit (found > 0 ? 1 : 0) }
' "$f")"
if [ -n "$result" ]; then
fail "$f: statement(s) outside any function body"
echo "$result" | sed 's/^/ /'
orphans=$((orphans + 1))
fi
done < <(find ports ports_arch ports_module ports_smp -name "*.h" -type f \
! -path "*/example_build/*" 2>/dev/null | sort)
[ "$orphans" -eq 0 ] && say " ok: no port header carries code at file scope"
# --------------------------------------------------------------------------
# 4. Report only: families with no copy script.
# --------------------------------------------------------------------------
# These are maintained by hand, so a fix applied to one toolchain can silently
# miss the others. Nothing here fails the run; it is a prompt to look.
say ""
say "== Families with no copy script (report only) =="
if [ "$quiet" -eq 0 ]; then
for family in ports/cortex_m0 ports/cortex_m0+ ports/cortex_m23; do
[ -d "$family" ] || continue
for probe in "dsb 0xF" "isb 0xF"; do
have=""; missing=""
for header in "$family"/*/inc/tx_port.h; do
[ -f "$header" ] || continue
grep -q "0xE000ED04" "$header" || continue
tool="$(basename "$(dirname "$(dirname "$header")")")"
if grep -q "$probe" "$header"; then
have="$have $tool"
else
missing="$missing $tool"
fi
done
if [ -n "$have" ] && [ -n "$missing" ]; then
echo " $family: \"$probe\" present in$have but absent in$missing"
fi
done
done
say " (nothing above means the toolchains within each family agree)"
fi
# --------------------------------------------------------------------------
say ""
if [ "$failures" -eq 0 ]; then
say "All port consistency checks passed."
exit 0
fi
echo "$failures port consistency check(s) failed."
exit 1