Work around Java's ``code too large'' problem for parser tables.

[bison.git] / tests / torture.at

# Torturing Bison.                                    -*- Autotest -*-
# Copyright (C) 2001, 2002, 2004, 2005, 2006, 2007 Free Software Foundation,
# Inc.

# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.

AT_BANNER([[Torture Tests.]])


# AT_INCREASE_DATA_SIZE(SIZE)
# ---------------------------
# Try to increase the data size to SIZE KiB if possible.
m4_define([AT_INCREASE_DATA_SIZE],
[data_limit=`(ulimit -S -d) 2>/dev/null`
case $data_limit in
[[0-9]]*)
  if test "$data_limit" -lt $1; then
    AT_CHECK([ulimit -S -d $1 || exit 77])
    ulimit -S -d $1
  fi
esac])


## ------------------------------------- ##
## Creating a large artificial grammar.  ##
## ------------------------------------- ##

# AT_DATA_TRIANGULAR_GRAMMAR(FILE-NAME, SIZE)
# -------------------------------------------
# Create FILE-NAME, containing a self checking parser for a huge
# triangular grammar.
m4_define([AT_DATA_TRIANGULAR_GRAMMAR],
[AT_DATA([[gengram.pl]],
[[#! /usr/bin/perl -w

use strict;
my $max = $ARGV[0] || 10;

print <<EOF;
]AT_DATA_GRAMMAR_PROLOGUE[
%error-verbose
%debug
%{
#include <stdio.h>
#include <stdlib.h>

static int yylex (void);
static void yyerror (const char *msg);
%}
%union
{
  int val;
};

%token END "end"
%type <val> exp input
EOF

for my $size (1 .. $max)
  {
    print "%token t$size $size \"$size\"\n";
  };

print <<EOF;
%%
input:
  exp        { if (\@S|@1 != 0) abort (); \$\$ = \@S|@1; }
| input exp  { if (\@S|@2 != \@S|@1 + 1) abort (); \$\$ = \@S|@2; }
;

exp:
  END
    { \$\$ = 0; }
EOF

for my $size (1 .. $max)
  {
    use Text::Wrap;
    print wrap ("| ", "   ",
		(map { "\"$_\"" } (1 .. $size)),
		" END \n"),
		  "    { \$\$ = $size; }\n";
  };
print ";\n";

print <<EOF;
%%
static int
yylex (void)
{
  static int inner = 1;
  static int outer = 0;
  if (outer > $max)
    return 0;
  else if (inner > outer)
    {
      inner = 1;
      ++outer;
      return END;
    }
  return inner++;
}

static void
yyerror (const char *msg)
{
  fprintf (stderr, "%s\\n", msg);
}

int
main (void)
{
  yydebug = !!getenv ("YYDEBUG");
  return yyparse ();
}
EOF
]])

AT_CHECK([perl -w ./gengram.pl $2 || exit 77], 0, [stdout])
mv stdout $1
])


## -------------- ##
## Big triangle.  ##
## -------------- ##

AT_SETUP([Big triangle])

# I have been able to go up to 2000 on my machine.
# I tried 3000, a 29Mb grammar file, but then my system killed bison.
# With 500 and the new parser, which consume far too much memory,
# it gets killed too.  Of course the parser is to be cleaned.
AT_DATA_TRIANGULAR_GRAMMAR([input.y], [200])
AT_BISON_CHECK_NO_XML([-v -o input.c input.y])
AT_COMPILE([input])
AT_PARSER_CHECK([./input])

AT_CLEANUP



# AT_DATA_HORIZONTAL_GRAMMAR(FILE-NAME, SIZE)
# -------------------------------------------
# Create FILE-NAME, containing a self checking parser for a huge
# horizontal grammar.
m4_define([AT_DATA_HORIZONTAL_GRAMMAR],
[AT_DATA([[gengram.pl]],
[[#! /usr/bin/perl -w

use strict;
my $max = $ARGV[0] || 10;

print <<EOF;
]AT_DATA_GRAMMAR_PROLOGUE[
%error-verbose
%debug
%{
#include <stdio.h>
#include <stdlib.h>

static int yylex (void);
static void yyerror (const char *msg);
%}

%token
EOF
for my $size (1 .. $max)
  {
    print "    t$size $size \"$size\"\n";
  };

print <<EOF;

%%
EOF

use Text::Wrap;
print
  wrap ("exp: ", "  ",
	(map { "\"$_\"" } (1 .. $max)), ";"),
  "\n";

print <<EOF;
%%
static int
yylex (void)
{
  static int counter = 1;
  if (counter <= $max)
    return counter++;
  if (counter++ != $max + 1)
    abort ();
  return 0;
}

static void
yyerror (const char *msg)
{
  fprintf (stderr, "%s\\n", msg);
}

int
main (void)
{
  yydebug = !!getenv ("YYDEBUG");
  return yyparse ();
}
EOF
]])

AT_CHECK([perl -w ./gengram.pl $2 || exit 77], 0, [stdout])
mv stdout $1
])


## ---------------- ##
## Big horizontal.  ##
## ---------------- ##

AT_SETUP([Big horizontal])

# I have been able to go up to 10000 on my machine, but I had to
# increase the maximum stack size (* 100).  It gave:
#
# input.y      263k
# input.tab.c  1.3M
# input        453k
#
# gengram.pl 10000                 0.70s user 0.01s sys  99% cpu    0.711 total
# bison input.y                  730.56s user 0.53s sys  99% cpu 12:12.34 total
# gcc -Wall input.tab.c -o input   5.81s user 0.20s sys 100% cpu     6.01 total
# ./input                          0.00s user 0.01s sys 108% cpu     0.01 total
#
AT_DATA_HORIZONTAL_GRAMMAR([input.y], [1000])

# GNU m4 requires about 70 MiB for this test on a 32-bit host.
# Ask for 200 MiB, which should be plenty even on a 64-bit host.
AT_INCREASE_DATA_SIZE(204000)

AT_BISON_CHECK_NO_XML([-v -o input.c input.y])
AT_COMPILE([input])
AT_PARSER_CHECK([./input])

AT_CLEANUP



# AT_DATA_LOOKAHEAD_TOKENS_GRAMMAR(FILE-NAME, SIZE)
# --------------------------------------------------
# Create FILE-NAME, containing a self checking parser for a grammar
# requiring SIZE lookahead tokens.
m4_define([AT_DATA_LOOKAHEAD_TOKENS_GRAMMAR],
[AT_DATA([[gengram.pl]],
[[#! /usr/bin/perl -w

use strict;
use Text::Wrap;
my $max = $ARGV[0] || 10;

print <<EOF;
%error-verbose
%debug
%{
# include <stdio.h>
# include <stdlib.h>
# include <assert.h>

static int yylex (void);
static void yyerror (const char *msg);
%}
%union
{
  int val;
};

%type <val> input exp
%token token
EOF

print
  wrap ("%type <val> ",
	"            ",
	map { "n$_" } (1 .. $max)),
  "\n";

print "%token\n";
for my $count (1 .. $max)
  {
    print "    t$count $count \"$count\"\n";
  };

print <<EOF;
%%
input:
  exp        { assert (\@S|@1 == 1); \$\$ = \@S|@1; }
| input exp  { assert (\@S|@2 == \@S|@1 + 1); \$\$ = \@S|@2; }
;

exp:
  n1 "1" { assert (\@S|@1 == 1); \@S|@\@S|@ = \@S|@1; }
EOF

for my $count (2 .. $max)
  {
    print "| n$count \"$count\" { assert (\@S|@1 == $count); \@S|@\@S|@ = \@S|@1; }\n";
  };
print ";\n";

for my $count (1 .. $max)
  {
    print "n$count: token { \$\$ = $count; };\n";
  };

print <<EOF;
%%
static int
yylex (void)
{
  static int return_token = 1;
  static int counter = 1;
  if (counter > $max)
    {
      if (counter++ != $max + 1)
	abort ();
      return 0;
    }
  if (return_token)
    {
      return_token = 0;
      return token;
    }
  return_token = 1;
  return counter++;
}

static void
yyerror (const char *msg)
{
  fprintf (stderr, "%s\\n", msg);
}

int
main (void)
{
  yydebug = !!getenv ("YYDEBUG");
  return yyparse ();
}
EOF
]])

AT_CHECK([perl -w ./gengram.pl $2 || exit 77], 0, [stdout])
mv stdout $1
])


## ------------------------ ##
## Many lookahead tokens.   ##
## ------------------------ ##

AT_SETUP([Many lookahead tokens])

AT_DATA_LOOKAHEAD_TOKENS_GRAMMAR([input.y], [1000])

# GNU m4 requires about 70 MiB for this test on a 32-bit host.
# Ask for 200 MiB, which should be plenty even on a 64-bit host.
AT_INCREASE_DATA_SIZE(204000)

AT_BISON_CHECK([-v -o input.c input.y])
AT_COMPILE([input])
AT_PARSER_CHECK([./input])

AT_CLEANUP



# AT_DATA_STACK_TORTURE(C-PROLOGUE, [BISON-DECLS])
# ------------------------------------------------
# A parser specialized in torturing the stack size.
m4_define([AT_DATA_STACK_TORTURE],
[# A grammar of parens growing the stack thanks to right recursion.
# exp:
AT_DATA([input.y],
[[%{
#include <errno.h>
#include <limits.h>
#include <stdio.h>
#include <stdlib.h>
]$1[
  static int yylex (void);
  static void yyerror (const char *msg);
%}
]$2[
%error-verbose
%debug
%token WAIT_FOR_EOF
%%
exp: WAIT_FOR_EOF exp | ;
%%
static void
yyerror (const char *msg)
{
  fprintf (stderr, "%s\n", msg);
}

static int
yylex (void)
{
  if (yylval < 0)
    abort ();
  if (yylval--)
    return WAIT_FOR_EOF;
  else
    return EOF;
}

int
main (int argc, const char **argv)
{
  char *endp;
  YYSTYPE yylval_init;
  if (argc != 2)
    abort ();
  yylval_init = strtol (argv[1], &endp, 10);
  if (! (argv[1] != endp
	 && 0 <= yylval_init && yylval_init <= INT_MAX
	 && errno != ERANGE))
    abort ();
  yydebug = 1;
  {
    int count;
    int status;
]m4_bmatch([$2], [%push-],
[[    yypstate *ps = yypstate_new ();
]])[    for (count = 0; count < 2; ++count)
      {
        int new_status;
        yylval = yylval_init;
]m4_bmatch([$2], [%push-],
[[        new_status = yypull_parse (ps);
]],
[[        new_status = yyparse ();
]])[        if (count > 0 && new_status != status)
          abort ();
        status = new_status;
      }
]m4_bmatch([$2], [%push-],
[[    yypstate_delete (ps);
]])[    return status;
  }
}
]])
AT_BISON_CHECK([-o input.c input.y])
AT_COMPILE([input])
])


## -------------------------------------- ##
## Exploding the Stack Size with Alloca.  ##
## -------------------------------------- ##

AT_SETUP([Exploding the Stack Size with Alloca])

m4_pushdef([AT_USE_ALLOCA], [[
#if (defined __GNUC__ || defined __BUILTIN_VA_ARG_INCR \
     || defined _AIX || defined _MSC_VER || defined _ALLOCA_H)
# define YYSTACK_USE_ALLOCA 1
#endif
]])

AT_DATA_STACK_TORTURE([AT_USE_ALLOCA])

# Below the limit of 200.
AT_PARSER_CHECK([./input 20], 0, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])
# Two enlargements: 2 * 2 * 200.
AT_PARSER_CHECK([./input 900], 0, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])
# Fails: beyond the limit of 10,000 (which we don't reach anyway since we
# multiply by two starting at 200 => 5120 is the last possible).
AT_PARSER_CHECK([./input 10000], 2, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])

# The push parser can't use alloca since the stacks can't be locals.  This test
# just helps guarantee we don't let the YYSTACK_USE_ALLOCA feature affect
# push parsers.
AT_DATA_STACK_TORTURE([AT_USE_ALLOCA],
[[%define api.push_pull "both"
]])
AT_PARSER_CHECK([./input 20], 0, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])
AT_PARSER_CHECK([./input 900], 0, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])
AT_PARSER_CHECK([./input 10000], 2, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])

m4_popdef([AT_USE_ALLOCA])

AT_CLEANUP




## -------------------------------------- ##
## Exploding the Stack Size with Malloc.  ##
## -------------------------------------- ##

AT_SETUP([Exploding the Stack Size with Malloc])

m4_pushdef([AT_USE_ALLOCA], [[#define YYSTACK_USE_ALLOCA 0]])

AT_DATA_STACK_TORTURE([AT_USE_ALLOCA])

# Below the limit of 200.
AT_PARSER_CHECK([./input 20], 0, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])
# Two enlargements: 2 * 2 * 200.
AT_PARSER_CHECK([./input 900], 0, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])
# Fails: beyond the limit of 10,000 (which we don't reach anyway since we
# multiply by two starting at 200 => 5120 is the possible).
AT_PARSER_CHECK([./input 10000], 2, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])

AT_DATA_STACK_TORTURE([AT_USE_ALLOCA],
[[%define api.push_pull "both"
]])
AT_PARSER_CHECK([./input 20], 0, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])
AT_PARSER_CHECK([./input 900], 0, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])
AT_PARSER_CHECK([./input 10000], 2, [], [ignore],
                [[VALGRIND_OPTS="$VALGRIND_OPTS --log-fd=1"]])

m4_popdef([AT_USE_ALLOCA])

AT_CLEANUP