0.9.1 --- the second register comes back from every call

This kernel returns a second value in x1 (rdx on x86_64) from every system call,
and the wrapper declared that register an input only. An optimizing compiler
therefore kept the first argument there across the call: the preopen table
stored the address of "/" from x1 after openat had cleared it, and every program
built with --release faulted in kal_fs_preopen before main. The register is now
an output as well. -fno-builtin keeps a counting loop from becoming strlen, and
CI runs this package's tests and the independence check in both profiles.

Author: Sunrisepeak

.github/workflows/ci.yml

@@ -227,19 +227,27 @@ jobs:
       #
       # The step is openkal-linux's, unchanged, including the assertion that
       # every suite ran: a suite that discovered nothing reports success.
+      # ⚠️ AND IN BOTH PROFILES. Until 0.9.1 every program built with --release
+      # faulted before `main': the system-call wrapper declared x1 an input
+      # only, this kernel returns a second value there, and an optimizing
+      # compiler kept an argument in it across the call. The dev profile keeps
+      # nothing in a register across a call and never showed it.
       - name: This package's own tests
         run: |
           set -euo pipefail
-          mcpp test 2>&1 | tee tests.log
+          for profile in dev release; do
+          echo "--- profile $profile"
+          mcpp test --profile "$profile" 2>&1 | tee tests.log
           # The list is derived from the files present rather than written out
           # here: a hand-written list names the suites that existed when it was
           # written, and one added afterwards escapes the assertion silently.
           missing=0
           for f in tests/*.cpp; do
             name="$(basename "$f" .cpp)"
-            grep -q "^$name \.\.\. ok" tests.log || { echo "did not run or did not pass: $name" >&2; missing=1; }
+            grep -q "^$name \.\.\. ok" tests.log || { echo "did not run or did not pass: $name ($profile)" >&2; missing=1; }
           done
           test "$missing" -eq 0
+          done
 
       # The other architecture, as far as this system allows it to be reached.
       #
@@ -279,7 +287,11 @@ jobs:
       # library through a macro.
       - name: The objects reference nothing of a C library but the two named
         run: |
-          rm -rf target && mcpp build --features standalone
+          # Both profiles: an optimizing compiler turns a counting loop into
+          # `strlen', which the dev profile never shows (hence -fno-builtin).
+          for profile in dev release; do
+          echo "--- profile $profile"
+          rm -rf target && mcpp build --features standalone --profile "$profile"
           objs="$(find target -path '*/obj/*' -name '*.o' ! -name '*.m.o' ! -name 'mcpp_*')"
           test -n "$objs" || { echo "no objects were found; the check would pass vacuously" >&2; exit 1; }
 
@@ -326,7 +338,8 @@ jobs:
             done
           done
           test "$bad" -eq 0
-          echo "the implementation reaches nothing of a C library but the two names that are named"
+          echo "the implementation reaches nothing of a C library but the two names that are named ($profile)"
+          done
 
       # A checker is only useful if it fails when it should.
       - name: The independence check detects a dependence

mcpp.toml

@@ -1,7 +1,7 @@
 [package]
 namespace   = "mcpplibs"
 name        = "openkal-macos"
-version     = "0.9.0"
+version     = "0.9.1"
 description = "An implementation of openkal for macOS, written on the kernel's own calls. Its purpose is as much to test the specification as to be used."
 license     = "Apache-2.0"
 
@@ -28,7 +28,7 @@ openkal = "0.12.0"
 # consumers rather than parts of the implementation.
 flags = [
   { glob = "src/**", cxxflags = ["-fno-exceptions", "-fno-rtti",
-                                 "-fno-stack-protector"] },
+                                 "-fno-stack-protector", "-fno-builtin"] },
 ]
 
 [features]

src/sys.h

@@ -54,9 +54,16 @@ inline okm_long sys(okm_long n, okm_long a = 0, okm_long b = 0, okm_long c = 0,
     register okm_long x4 __asm__("x4") = e;
     register okm_long x5 __asm__("x5") = f;
     okm_long failed;
-    __asm__ __volatile__("svc #0x80\n\tcset %1, cs"
-                         : "+r"(x0), "=r"(failed)
-                         : "r"(x16), "r"(x1), "r"(x2), "r"(x3), "r"(x4), "r"(x5)
+    // ⚠️⚠️ THE SECOND REGISTER COMES BACK TOO, AND UNTIL 0.9.1 IT WAS DECLARED AN
+    // INPUT ONLY. This kernel returns a second value in x1 from every call ---
+    // the two calls below read it --- and writes it whether or not the call has
+    // one. Declared as an input, x1 was assumed to survive, and an optimizing
+    // compiler kept the first argument there: the preopen table stored the
+    // address of "/" from x1 after `openat' had cleared it, and every program
+    // built with --release faulted in `kal_fs_preopen' before `main'.
+    __asm__ __volatile__("svc #0x80\n\tcset %2, cs"
+                         : "+r"(x0), "+r"(x1), "=r"(failed)
+                         : "r"(x16), "r"(x2), "r"(x3), "r"(x4), "r"(x5)
                          : "memory", "cc");
     return failed ? -x0 : x0;
 }
@@ -71,9 +78,10 @@ inline okm_long sys(okm_long n, okm_long a = 0, okm_long b = 0, okm_long c = 0,
     register okm_long r10 __asm__("r10") = d;
     register okm_long r8  __asm__("r8")  = e;
     register okm_long r9  __asm__("r9")  = f;
+    // The second value comes back in rdx, as x1 above: an output as well.
     __asm__ __volatile__("syscall"
-                         : "=a"(r), "=@ccc"(failed)
-                         : "a"(n | 0x2000000L), "D"(a), "S"(b), "d"(c),
+                         : "=a"(r), "+d"(c), "=@ccc"(failed)
+                         : "a"(n | 0x2000000L), "D"(a), "S"(b),
                            "r"(r10), "r"(r8), "r"(r9)
                          : "rcx", "r11", "memory", "cc");
     return failed ? -r : r;