Version 2.1.42 (20050930)
=========================
This version fixes some bugs if the installation process and
introduces some minor enhancements.

ejudge-setup
------------
The command of creation of the symlink to the serve-control CGI
program is added to the installation script generation code.

Makefile
--------
Handling of situation, when the localization utilities xgettext,
etc were not found during the ejudge configuration is improved.

userlist-server
---------------
All the user cookies are dropped when the password is changed.

super-serve,serve-control
-------------------------
Editing of the `checker_env' problem configuration variable is
supported.

The new standard checkers (see below) are supported.

When the content of the header and the footer files and the
start script is edited, it is automatically converted from the
DOS text format to the UNIX text format, because browsers send
the data in the DOS text format (i. e. with \r\n as line
terminators).

The `master' and `judge' CGI-programs are opened in a new
browser window when clicking on the corresponding links.

Editing of the `variant_num' problem configuration variable is
supported. Editing of the variant assingment table (variant map)
is supported. Checking of contest settings for variant problems
is supported.

libchecker
----------
New functions for comparing floating-point numbers are added.

int checker_eq_double(double v1, double v2, double eps);
int checker_eq_long_double(long double v1, long double v2, long double eps);
int checker_eq_float(float v1, float v2, float eps);

These functions compare two floating-point numbers v1 and v2 of
the corresponding types with the given precision eps. Special
values NaN, +Inf, -Inf are handled correctly. If the absolute
values of the numbers do not exceed 1, the absolute value of
their difference must be less than 2*eps, i.e.
|v1 - v2| < 2*eps. Else both numbers must have the same sign,
must have the exponents with difference <= 1, and then their
exponent are changed, so the both numbers fit into the [0,2)
interval, then the condition |v1 - v2| < 2*eps. Thus the eps
value determines how many matching decimal digits do the numbers
have.

Several new standard checkers are added:
cmp_float, cmp_double, cmp_long_double - compare the output of
the program being checked, which must be the only floating-point
number of the corresponding type, and the correct answers. The
floating-point comparison functions described above are used.
The eps parameter must be defined by the EPS environment
variable, which must be set using the `checker_env' problem
configuration variable.
cmp_bytes - compares the output of the program being checked and
the correct answer byte by byte.
cmp_int_seq, cmp_unsigned_int_seq, cmp_long_long_seq,
cmp_unsigned_long_long_seq - compare the sequence of the integer
numbers of the corresponding type in the output of the program
being tested and the correct answer. The checked output must
consist only of integer numbers separated with whitespace
characters. The total number of numbers and the numbers themself
in the output must match those in the correct answer.
cmp_double_seq, cmp_long_double_seq - compare the sequence of
the floating-point numbers of the corresponding type in the
output of the program being checked and in the correct answer.
The checked output must consist only of floating-point numbers
separated with whitespace characters. The total number of
numbers and the numbers themself in the output must match those
in the correct answer.