all-tests: computing success or failure should hopefully be more robust now

diff --git a/all-tests b/all-tests
index 5c53854..6cae656 100755 (executable)
--- a/all-tests
+++ b/all-tests
@@ -43,17 +43,16 @@ function all_tests () {
     echo tests
     echo ========================================
     make PYTHON=$PYTHON all-tests
+    make_retcod=$!
     echo ========================================
     end=$(date +%s)
     echo all-tests : end at; date; echo total duration $(($end-$begin)) seconds
+    return $make_retcod
 }
 
-# WARNING: the current code for analyzing the output is not quite right
-# I've seen cases where the tests session goes very badly - with broken imports for example
-# but it's considered OK though
-# we mght wish to take the opposite angle : check we have as many OK as expected
-# but then the list of default targets should be declared here and not in Makefile
-# in any case this is good enough for now
+# WARNING
+# the current code for analyzing the output is very basic
+# this needs to be checked manually for some time
 
 all_tests >& $output
 retcod=$?