Add make_flex.sh and update lexer/parser generation documentation.

pars/make_bison.sh

@@ -1,7 +1,6 @@
 #!/bin/bash
 #
-# regenerate parser from bison input files as documented at the top of
+# generate parser files from bison input files.
-# pars0lex.l.
 
 set -eu
 

pars/make_flex.sh

@@ -0,0 +1,20 @@
+#!/bin/bash
+#
+# generate lexer files from flex input files.
+
+set -eu
+
+TMPFILE=_flex_tmp.c
+OUTFILE=lexyy.c
+
+flex -o $TMPFILE pars0lex.l
+
+# AIX needs its includes done in a certain order, so include "univ.i" first
+# to be sure we get it right.
+echo '#include "univ.i"' > $OUTFILE
+
+# flex assigns a pointer to an int in one place without a cast, resulting in
+# a warning on Win64. this adds the cast.
+sed -e 's/int offset = (yy_c_buf_p) - (yytext_ptr);/int offset = (int)((yy_c_buf_p) - (yytext_ptr));/;' < $TMPFILE >> $OUTFILE
+
+rm $TMPFILE

pars/pars0lex.l

@@ -12,27 +12,11 @@ not automatically generate them from pars0grm.y and pars0lex.l.
 
 How to make the InnoDB parser and lexer C files:
 
-1. First do
+1. Run ./make_flex.sh to generate lexer files.
-        bison -d pars0grm.y
-   That generates pars0grm.tab.c and pars0grm.tab.h.
 
-2. Rename pars0grm.tab.c to pars0grm.c and pars0grm.tab.h to pars0grm.h.
+2. Run ./make_bison.sh to generate parser files.
 
-3. Copy pars0grm.h also to /innobase/include
+These instructions seem to work at least with bison-1.875d and flex-2.5.31 on
-
-4. Do
-        flex pars0lex.l
-   That generates lex.yy.c.
-
-5. Rename lex.yy.c to lexyy.c.
-
-6. Add '#include "univ.i"' before #include <stdio.h> in lexyy.c
-   (Needed for AIX)
-
-7. Add a type cast to int to the assignment below the comment
-   'need more input.'  (Removes a warning on Win64)
-
-These instructions seem to work at least with bison-1.28 and flex-2.5.4 on
 Linux.
 *******************************************************/