Let recollindex execute a script at startup to try and guess if it should retry failed files

src/filters/rclcheckneedretry.sh

@@ -0,0 +1,41 @@
+#!/bin/sh
+
+# Check /usr/bin and /usr/local/bin modification date against recorded
+# state, as recorded inside ~/.config/Recoll.org/needidxretrydate
+#
+# If any argument is given, we record the new state instead of
+# generating it (this should be used at the end of an indexing pass
+# with retry set).
+#
+# The script exits with 0 if retrying should be performed (something
+# changed), 1 else.
+
+# Bin dirs to be tested:
+bindirs="/usr/bin /usr/local/bin $HOME/bin /opt/*/bin"
+
+rfiledir=~/.config/Recoll.org
+rfile=$rfiledir/needidxretrydate
+nrfile=$rfiledir/tneedidxretrydate
+
+test -d $rfiledir || mkdir -p $rfiledir
+
+# If any argument is given, we are called just to record the new
+# state. We do not recompute it as it may have changed during
+# indexing, but just move the state in place
+if test $# != 0 ; then 
+    mv -f $nrfile $rfile
+    exit 0
+fi
+
+# Compute state of bin dirs and see if anything changed:
+> $nrfile
+for dir in $bindirs; do
+    ls -ld $dir >> $nrfile 2> /dev/null
+done
+
+if cmp -s $rfile $nrfile ; then 
+    exit 1
+else
+    exit 0
+fi
+