input.h: change meaning of interactive and forced_interactive global variables. * libinterp/corefcn/input.h, libinterp/corefcn/input.cc: the interactive var is set at the start of Octave session if the session would be interactive. The forced_interactive is set only by the "--interactive" option. This means that in a forced interactive session, interactive is false, and that in a not forced interactive session with the "--interactive" option, forced_interactive is true. This is a bit counter-intuitive. Also when it matters if a session is interactive or not, it should not matter if it was forced not, only whether is interactive. Change this logic. Interactive means we are running a interactive session, forced interactive means the interactive session needed to be forced. * libinterp/octave.cc: during Octave initialization, as soon as we identify if this would be an interactive session, check with forced_interactive to adjust the value of this variables. * libinterp/corefcn/error.cc, libinterp/corefcn/oct-stream.cc, libinterp/corefcn/pager.cc, libinterp/corefcn/dirfns.cc, libinterp/corefcn/sighandlers.cc, libinterp/corefcn/sysdep.cc, libinterp/corefcn/toplev.cc, libinterp/parse-tree/lex.ll, libinterp/parse-tree/oct-parse.in.yy: replace all checks for "(interactive || forced_interactive)" with simply interactive. On the few cases where forced_interactive was not checked, it was replaced with "(interactive && ! forced_interactive)", just like "! interactive" got replaced by "(! interactive || forced_interactive)", only to conserve the logic. I am uncertain if such checks are not actually bugs though.

--- a/libinterp/parse-tree/lex.ll	Tue Feb 24 10:41:57 2015 -0800
+++ b/libinterp/parse-tree/lex.ll	Tue Feb 24 17:09:42 2015 +0000
@@ -2236,7 +2236,7 @@
   // input.
 
   if (! quitting_gracefully
-      && (interactive || forced_interactive)
+      && interactive
       && ! (reading_fcn_file
             || reading_classdef_file
             || reading_script_file