Fix the ninja test for console ctrl-c

.github/workflows/runtest.yml

@@ -61,7 +61,6 @@ jobs:
             echo "test/Libs/LIBLITERALPREFIX.py" >> .excludes
             echo "test/Rpcgen/live.py" >> .excludes
             echo "test/builderrors.py" >> .excludes
-            echo "test/ninja/ninja_handle_control_c_rebuild.py" >> .excludes
             echo "test/ninja/shutdown_scons_daemon.py" >> .excludes
             echo "test/scons-time/run/option/verbose.py" >> .excludes
           elif [[ "${{matrix.os}}" == "windows"* ]]; then

CHANGES.txt

@@ -30,6 +30,7 @@ RELEASE  VERSION/DATE TO BE FILLED IN LATER
       outright.
 
   From Mats Wichmann:
+    - Fix the ninja ctrl-c test, which was failing on some platforms.
     - Reference manual improvements:
       * Additional clarification for SideEffect()
 

RELEASE.txt

@@ -49,6 +49,8 @@ FIXES
   the SCons documentation build and the docbook tool, which failed on any
   DocBook source pulling in an external entity module.
 
+- Fix the ninja ctrl-c test, which was failing on some platforms.
+
 IMPROVEMENTS
 ------------
 

test/ninja/ninja_test_sconscripts/sconstruct_control_c_ninja

@@ -13,4 +13,13 @@ env.Tool('ninja')
 env.Program(target='foo', source='foo.c')
 
 if ARGUMENTS.get('NINJA_DISABLE_AUTO_RUN', 0):
+    # Simulate a terminal Ctrl-C. We have to send two signals: from a terminal,
+    # SIGINT is delivered to the whole foreground process group, including
+    # both ninja and the subprocess scons busy regenerating build.ninja,
+    # however ninja intentionally does not deliver the signal to its children,
+    # so just sending a kill to ninja is not a sufficient to simulate.
+    # This test failed on some platforms because scons continued running,
+    # finished writing the new build.ninja, which thus had a new mtime,
+    # so ninja's shutdown processing would delete it because it's modified.
     os.kill(os.getppid(), signal.SIGINT)
+    os.kill(os.getpid(), signal.SIGINT)