* src/expr.c (die): New "noreturn" function to wrap one-arg use of
error
(string_too_long): Use die rather than error.
(toint): Remove definition of now-unused function.
(eval6): Remove a little duplication.
Use die rather than error.
(dodivide): Remove declaration of now-unused variable.
* src/Makefile.am (expr_LDADD): Link expr against GNU MP.
* doc/coreutils.texi (expr invocation): Describe --bignum,
--no-bignum. Explain the new arbitrary-precision functionality.
* NEWS: Indicate that arbitrary-precision arithmetic is now
supported in expr.
* src/expr.c (enum valtype): Added mp_integer, signifying a GNU MP
number.
(usage): Document the new options --bignum and --no-bignum which
force and prohibit the use of arbitrary-precision arithmetic,
respectively.
(long_options): data structure for getopt_long, which we need to
use to parse the options mentioned above.
(main): parse these options with getopt_long instead of
parse_long_options.
(valinfo): Downgrade the numeric member of the union from
intmax_t to signed long, since MP lacks functions for promoting an
intmax_t to an arbitrary-precision quantity.
(enum arithmetic_mode): Represents the current choice between
--bignum, --no-bignum and the default (automatically switch from
one to the other if needed).
(integer_overflow): issue a more explicit error message indicating
that MP is not available.
(string_too_long): new function, emits a fatal error message for
the case where an argument to the 'index' expression is too long
for a string offset to be represented.
(int_value): With --bignum, create the value as mp_integer rather
than plain integer.
(substr_value): factored out of eval6; implements "substr".
(freev): also destroy mp_integer values. Check that no mp_integer
values exist if --no-bignum was specified.
(printv, null, tostring): support mp_integer.
(toint): new funtion for converting from string or mp_integer to
integer.
(getsize): extracts a size_t value from a VALUE object; used to
implement substr.
(promote): promotes a value from integer to mp_integer.
(domult, dodivide): functions for multiplication and division,
factored out of eval4.
(doadd): addition/subraction function, factpred out of eval3.
(eval3): support mp_integer types; call doadd.
(eval4): support mp_integer types; call domult, dodivide.
(eval6): support mp_integer offsets and lengths for "substr" and
"index".
* TODO: Mention that expr supports arbitrary-precision arithmetic,
and suggest that this might also be a good idea for seq.
* AUTHORS (expr): Add James Youngman.
* src/expr.c (docolon): Free the regexp buffer using regfree, rather
than doing it manually, being careful to set fastmap to NULL first.
Free any re_regs.start and .end members, if necessary.
what POSIX requires. Also, don't let anchors match
newline; this fixes an incompatibility with tradition and with POSIX.
Don't warn about leading ^. POSIX says it is unspecified whether
^ is a special character, which means that implementations can
either treat it as special or not, but either way a warning is not
allowed (unless the regexp is otherwise invalid). Instead, anchor
the expression but treat ^ as an anchor; this is the traditional
behavior (e.g., Solaris 10).
(eval4, eval3, eval2): Treat non-numeric args, division by zero,
and the like as invalid expressions (exit status 2), not as
failure of 'expr' (exit status 3). This is more consistent with
how Solaris behaves.
(looks_like_integer): New function.
(toarith): Use it. Also, use xstrtoimax rather than rolling our
own diagnostics.
(eval2): Don't look for trouble if !evaluate; this simplifies things.
Compare numbers using string comparison, so that overflow is
not possible.
rather than suffering silently.
Before, expr would silently overflow and wrap around:
$ expr 9223372036854775808 = 0 # $(echo 2^63|bc)
1
Now it detects the problem and exits nonzero:
$ ./expr $(echo 2^63|bc) = 0
./expr: 9223372036854775808: integer is too large
too few operands ("missing operand after `xxx'") or
too many operands ("extra operand `xxx'").
Include "quote.h" and/or "error.h" if it wasn't already being included.
Accept a bool argument specifying whether to evaluate the
expression. This is to allow short-circuit evaluation. All
callers changed.
(null): Report that a string is zero even if it has
a form like "-0" or "00".
(eval1, eval): Use short-circuit evaluation for | and &.
(eval): Return 0 if both arguments are null or zero, instead
of returning the first argument.
(main): Use initialize_exit_failure rather than
setting exit_failure directly; this optimizes away redundant assignments.
(EXPR_FAILURE): Renamed from EXPR_ERROR, for
consistency with the other programs' naming conventions.
All uses changed.
of 2003-09-19. Now, AUTHORS is a comma-separated list of strings.
Update the call to parse_long_options so that `AUTHORS, NULL' are the
last parameters.
* src/true.c (main): Append NULL to version_etc argument list.
* src/sys2.h (case_GETOPT_VERSION_CHAR): Likewise.
(EXPR_INVALID, EXPR_ERROR): New constants.
(nomoreargs, null, toarith, nextarg): Return bool, not int.
(syntax_error): New function, exiting with status 2. Use it
insteading of printing "syntax error" ourselves.
(main): Initialize exit_failure to EXPR_ERROR.
Exit with EXPR_INVALID on syntax error (too few arguments).
(nextarg): Use strcmp, not strcoll; strcoll might return
an undesirable 0, or might fail.
(docolon, eval4, eval3): Exit with status 3 on invalid argument type
or other such error.
(eval2): Report an error if strcoll fails in a string comparison.