Unbreak the build on 9.X (against old yacc(1) which apparently does not

support %token_table declaration).  Originally the port was relying on
bison(1) instead (via USES=bison) in yacc-compat mode, but was switched
to yacc(1) in r389653 because it failed to build against `devel/bison'
version update to 3.0.4 as part of an exp-run (see PR).

It turns out that actually generated code does not suffer from removing
%token_table: the only difference in generated `neural_parse.cc' is that
yyname[] table is not guarded by `#if YYDEBUG' and it #define's yytname
as yyname, but the code that references yytname in `neural_parse.cc' is
#if'ed 0 and YYDEBUG is defined to 1 earlier.  That said, simply comment
out %token_table line which allows the build to proceed (ir)regardless of
the yacc(1) version, on all FreeBSD releases (including 8.X).

While here, remove `ac_cv_lib_stdcpp_main=no' in CONFIGURE_ENV that was
supposed to fix the build against Clang added in r323095 as it seems no
longer required, augment COMMENT and port description, and fix WWW line
to point to working website, killing trailing whitespace along the way.

PR:	200833

audio/epos-devel/Makefile

@@ -8,16 +8,14 @@ CATEGORIES=	audio
 MASTER_SITES=	SF/${PORTNAME}/${PORTNAME}-dev/${PORTVERSION}
 
 MAINTAINER=	ports@FreeBSD.org
-COMMENT=	Text-to-Speech system
+COMMENT=	Language independent Text-to-Speech (TTS) system
 
 LICENSE=	GPLv2 LGPL20
 LICENSE_COMB=	multi
 
-BROKEN_FreeBSD_9=	does not build on FreeBSD < 10.x
-
 USES=		gmake
 GNU_CONFIGURE=	yes
-CONFIGURE_ENV=	ac_cv_lib_stdcpp_main=no YACC=${YACC}
+CONFIGURE_ENV=	YACC="${YACC}"
 
 post-patch:
 	@${REINPLACE_CMD} -e 's/^\(SUBDIRS =.*\)doc/\1/' ${WRKSRC}/Makefile.in

audio/epos-devel/files/patch-src__nnet__neural_parse.yy

@@ -1,6 +1,15 @@
 --- src/nnet/neural_parse.yy.orig
 +++ src/nnet/neural_parse.yy
-@@ -255,7 +255,7 @@
+@@ -104,7 +104,7 @@ make_tree (TTypedValue val = TTypedValue
+ /*                    SYMBOLS                        */
+ /* * * * * * * * * * * * * * * * * * * * * * * * * * */
+ 
+-%token_table
++//%token_table -- unsupported by the base yacc(1) on some FreeBSD versions
+ //%raw bison nefunguje s raw! (aspon yylex ne)
+ 
+ %union {
+@@ -255,7 +255,7 @@ unit_level: string_exp 
  
  int yyerror (char *s)
  {

audio/epos-devel/pkg-descr

@@ -1,8 +1,16 @@
 Epos is a language independent rule-driven Text-to-Speech (TTS) system
 primarily designed to serve as a research tool.  Epos is (or tries to be)
-independent of the language processed, linguistic description method, 
-and computing environment.
+independent of the language processed, linguistic description method, and
+computing environment.  It has the following features:
 
-This port is built from the development branch of Epos.
+  - Czech and Slovak text to speech synthesis configuration
+  - LPC and time domain speech synthesizers
+  - Prosody modelling driven by rules, prosody models, and artificial
+    neural networks
+  - All language dependencies are defined in the configuration files,
+    as opposed to the source code
+  - Documentation in English included in the distribution
 
-WWW: http://epos.ure.cas.cz/
+This is a port for the development version of Epos.
+
+WWW: http://epos.ufe.cz/