coccinelle: rework how we run the Coccinelle transformations

Turns out that the original way we did things was quite broken, as it
skipped a _lot_ of code. This was because we just threw everything into
one pile and tried to spatch it, but this made Coccinelle sad, like when
man page examples redefined some of our macros, causing typedef
conflicts.

For example, with a minimal reproducer that defines a cleanup macro in
two source files, Coccinelle has no issues when spatch-ing each one
separately:

$ spatch --verbose-parsing --sp-file zz-drop-braces.cocci main.c
init_defs_builtins: /usr/lib64/coccinelle/standard.h
HANDLING: main.c
SPECIAL NAMES: adding _cleanup_ as a attribute with arguments
SPECIAL NAMES: adding _cleanup_free_ as a attribute

$ spatch --verbose-parsing --sp-file zz-drop-braces.cocci
logcontrol-example.c
init_defs_builtins: /usr/lib64/coccinelle/standard.h
HANDLING: logcontrol-example.c
SPECIAL NAMES: adding _cleanup_ as a attribute with arguments

But when you try to spatch both of them at once, Coccinelle starts
complaining and skipping the "bad" code:

$ spatch --verbose-parsing --sp-file zz-drop-braces.cocci main.c logcontrol-example.c
init_defs_builtins: /usr/lib64/coccinelle/standard.h
HANDLING: main.c logcontrol-example.c
SPECIAL NAMES: adding _cleanup_ as a attribute with arguments
SPECIAL NAMES: adding _cleanup_free_ as a attribute
remapping: _cleanup_ to an ident in macro name
ERROR-RECOV: found sync end of #define, line 44
parsing pass2: try again
ERROR-RECOV: found sync end of #define, line 44
parse error
 = File "logcontrol-example.c", line 44, column 21, charpos = 1719
  around = '__attribute__',
  whole content = #define _cleanup_(f) __attribute__((cleanup(f)))
badcount: 2
bad: #include <systemd/sd-journal.h>
bad:
BAD:!!!!! #define _cleanup_(f) __attribute__((cleanup(f)))

This was, unfortunately, hidden as it is visible only with
--verbose-parsing (or --parse-error-msg).

Another issue was how we handled includes. The original way of throwing
them into the pile of source files doesn't really work, leading up to
similar issues as above. The better way is to let Coccinelle properly
resolve all includes by telling it where to find our own include files
(basically the same thing we already do during compilation).

After fixing all this, Coccinelle now has a chance to process much more
of our code (there are still some issues in more complex macros, but
that requires further investigation). However, there's a huge downside
from all of this - doing a _proper_ code analysis is surprisingly time
and resource heavy; meaning that processing just one Coccinelle rule now
takes 15 - 30 minutes.

To make this slightly less painful, Coccinelle supports caching the
generated ASTs, which actually helps a lot - it gets the runtime of one
rule from 15 - 30 minutes down to ~1 minute. It, of course, has its own
downside - the cache is _really_ big (ATTOW the cache takes ~15 GiB).

However, even with the aggressive AST caching you're still looking at
~1 hour for one full Coccinelle run, which is a bit annoying, but I
guess that's the price of doing things _properly_ (but I'll definitely
look into ways of further optimizing this).

coccinelle/run-coccinelle.sh

@@ -2,6 +2,14 @@
 # SPDX-License-Identifier: LGPL-2.1-or-later
 set -e
 
+# FIXME:
+#   - Coccinelle doesn't like our TEST() macros, which then causes name conflicts; i.e. Cocci can't process
+#     that TEST(xsetxattr) yields test_xsetxattr() and uses just xsetxattr() in this case, which then conflicts
+#     with the tested xsetxattr() function, leading up to the whole test case getting skipped due to
+#     conflicting typedefs
+#   - something keeps pulling in src/boot/efi/*.h stuff, even though it's excluded
+#   - Coccinelle has issues with some of our more complex macros
+
 # Exclude following paths from the Coccinelle transformations
 EXCLUDED_PATHS=(
     "src/boot/efi/*"
@@ -10,13 +18,17 @@ EXCLUDED_PATHS=(
     # Symlinked to test-bus-vtable-cc.cc, which causes issues with the IN_SET macro
     "src/libsystemd/sd-bus/test-bus-vtable.c"
     "src/libsystemd/sd-journal/lookup3.c"
+    # Ignore man examples, as they redefine some macros we use internally, which makes Coccinelle complain
+    # and ignore code that tries to use the redefined stuff
+    "man/*"
 )
 
 TOP_DIR="$(git rev-parse --show-toplevel)"
+CACHE_DIR="$(dirname "$0")/.coccinelle-cache"
 ARGS=()
 
 # Create an array from files tracked by git...
-mapfile -t FILES < <(git ls-files ':/*.[ch]')
+mapfile -t FILES < <(git ls-files ':/*.c')
 # ...and filter everything that matches patterns from EXCLUDED_PATHS
 for excl in "${EXCLUDED_PATHS[@]}"; do
     # shellcheck disable=SC2206
@@ -37,12 +49,43 @@ fi
 
 [[ ${#@} -ne 0 ]] && SCRIPTS=("$@") || SCRIPTS=("$TOP_DIR"/coccinelle/*.cocci)
 
+mkdir -p "$CACHE_DIR"
+echo "--x-- Using Coccinelle cache directory: $CACHE_DIR"
+echo "--x--"
+echo "--x-- Note: running spatch for the first time without populated cache takes"
+echo "--x--       a _long_ time (15-30 minutes). Also, the cache is quite large"
+echo "--x--       (~15 GiB), so make sure you have enough free space."
+echo
+
 for script in "${SCRIPTS[@]}"; do
     echo "--x-- Processing $script --x--"
     TMPFILE="$(mktemp)"
     echo "+ spatch --sp-file $script ${ARGS[*]} ..."
-    parallel --halt now,fail=1 --keep-order --noswap --max-args=20 \
-             spatch --macro-file="$TOP_DIR/coccinelle/macros.h" --smpl-spacing --sp-file "$script" "${ARGS[@]}" ::: "${FILES[@]}" \
-             2>"$TMPFILE" || cat "$TMPFILE"
+    # A couple of notes:
+    #
+    # 1) Limit this to 10 files at once, as processing the ASTs is _very_ memory hungry - e.g. with 20 files
+    # at once one spatch process can take around 2.5 GiB of RAM, which can easily eat up all available RAM
+    # when paired together with parallel
+    #
+    # 2) Make sure spatch can find our includes via -I <dir>, similarly as we do when compiling stuff
+    #
+    # 3) Make sure to include includes from includes (--recursive-includes), but use them only to get type
+    # definitions (--include-headers-for-types) - otherwise we'd start formating them as well, which might be
+    # unwanted, especially for includes we fetch verbatim from third-parties
+    #
+    # 4) Use cache, since generating the full AST is _very_ expensive, i.e. the uncached run takes 15 - 30
+    # minutes (for one rule(!)), vs 30 - 90 seconds when the cache is populated. One major downside of the
+    # cache is that it's quite big - ATTOW the cache takes around 15 GiB, but the performance boost is
+    # definitely worth it
+    parallel --halt now,fail=1 --keep-order --noswap --max-args=10 \
+        spatch --cache-prefix "$CACHE_DIR" \
+               -I src \
+               --recursive-includes \
+               --include-headers-for-types \
+               --smpl-spacing \
+               --sp-file "$script" \
+               "${ARGS[@]}" ::: "${FILES[@]}" \
+               2>"$TMPFILE" || cat "$TMPFILE"
+    rm -f "$TMPFILE"
     echo -e "--x-- Processed $script --x--\n"
 done