Added "STATIC bool blum_initialized = false" to zrandom.c to improve
how the code detects if the Blum-Blum-Shub pseudo-random number
generator is seeded or not, and how to free the state correctly.
NOTE: There is a very minor memory leak in zrandom.c that will be
fixed in a later release.
So far, the Makefile checked for sys/param.h instead and that suspiciously
looks like a copy & paste error.
Found by the Debian hurd-i386 build daemon because calc failed to compile
there and it seems to lack sys/mount.h. Not that this would be an extremely
relevant architecture these days, but nevertheless... ;-)
On ancient 32 bit big endian machines (e.g. MIPS) file.c fails to build
when FILEPOS is a struct and not a scalar. This happens because the
SWAP_HALF_IN_FILEPOS macro essentially expands to a simple assignment in
this case but the types are incompatile (HALF and FILEPOS).
Detailed compile error (this is from an older version so line numbers might
be wrong but the error is the same):
gcc -DCALC_SRC -DCUSTOM -Wall -g -O2 -fstack-protector-strong -Wformat -Werror=format-security -O3 -g3 -Wno-error=long-long -Wno-long-long -c file.c
In file included from qmath.h:32,
from cmath.h:32,
from value.h:33,
from calc.h:33,
from file.c:39:
file.c: In function 'filepos2z':
zmath.h:85:46: error: incompatible types when assigning to type 'HALF' {aka 'long unsigned int'} from type 'FILEPOS' {aka 'struct _G_fpos_t'}
#define SWAP_HALF_IN_B32(dest, src) (*(dest) = *(src))
^
fposval.h:15:42: note: in expansion of macro 'SWAP_HALF_IN_B32'
#define SWAP_HALF_IN_FILEPOS(dest, src) SWAP_HALF_IN_B32(dest, src)
^~~~~~~~~~~~~~~~
file.c:1370:2: note: in expansion of macro 'SWAP_HALF_IN_FILEPOS'
SWAP_HALF_IN_FILEPOS(ret.v, &pos);
^~~~~~~~~~~~~~~~~~~~
Fix this by adding suitable casts to the definition of SWAP_HALF_IN_FILEPOS
on big endian machines. Strictly speaking, the casts only seem to be
necessary when using SWAP_HALF_IN_B32, but they can't hurt in the other
cases either.
The chk_tree will check for files in distlist that
would otherwise be ignored by trailblank or that
are used, by convention, as tempory / test files.
Fixed a shellcheck nit in trailblank.
Updated BUGS about MSYS2 on Windows compiling of calc.
Added more git related checks and sanity checks to chk_tree.
Added ${FSANITIZE} make variable to Makefile.config to hold
common Address Sanitizer (ASAN) optins to modern Linux and macOS.
The Address Sanitizer is NOT enabled not compiled in by default.
Improved comments in Makefile.local for RHEL9.2 (Linux) and for
macOS 14.0 that, when uncommented and calc is recompiled (i.e.,
make clobber all) will enable the Address Sanitizer (ASAN) for calc.
Before, make prep would show but not object to chk_tree failures.
Now, if all except chk_tree is OK, make prep will report:
almost satifactory except for chk_tree
The funclist.sed sed script, when transforming func.c
into help/funclist.c, was creating nested comments.
While those were hardless, a single change to func.c
avoids the naive funclist.sed processing and avoids
(harmless) warnings when compiling help/funclist.c
in the course of building the function list.
The following are the changes in this release:
The tarball for calc version 2.15.0.0 was missing version.h.
The version.h is now listed as part of the calc distribution.
Added the following new trigonometric functions:
versin(x [,eps]) versed trigonometric sine
coversin(x [,eps]) coversed trigonometric sine
vercos(x [,eps]) versed trigonometric cosine
covercos(x [,eps]) coversed trigonometric cosine
aversin(x [,eps]) inverse versed trigonometric sine
acoversin(x [,eps]) inverse coversed trigonometric sine
avercos(x [,eps]) inverse versed trigonometric cosine
acovercos(x [,eps]) inverse coversed trigonometric cosine
haversin(x [,eps]) half versed trigonometric sine
hacoversin(x [,eps]) half coversed trigonometric sine
havercos(x [,eps]) half versed trigonometric cosine
hacovercos(x [,eps]) half coversed trigonometric cosine
ahaversin(x [,eps]) inverse half versed trigonometric sine
ahacoversin(x [,eps]) inverse half coversed trigonometric sine
ahavercos(x [,eps]) inverse half versed trigonometric cosine
ahacovercos(x [,eps]) inverse half coversed trigonometric cosine
exsec(x [,eps]) exterior trigonometric secant
aexsec(x [,eps]) inverse exterior trigonometric secant
excsc(x [,eps]) exterior trigonometric cosecant
aexcsc(x [,eps]) inverse exterior trigonometric cosecant
crd(x [,eps]) trigonometric chord of a unit circle
acrd(x [,eps]) inverse trigonometric chord of a unit circle
cas(x [,eps]) trigonometric cosine plus sine
cis(x [,eps]) Euler's formula
As Msys2 is a fork of Cygwin, if the OSNAME is Msys, the Cygwin
target will be used. Thanks to GitHub user @iahung2 for the
pull request.
Support for win32 and DJGPP has been dropped. Calc version
2.14.3.5 was the last to make references to win32 and make
references to DJGPP. Future versions of calc may work under
those systems, we just elected to remove the somewhat out of
date and awkward `win32.mkdef` and related win32 references.
If you are a win32 user, please feel free to create a win32
target in Makefile.target and submit as a pull request.
If you are a DJGPP user, please feel free to create a DJGPP
target in Makefile.target and submit as a pull request.
Until someone can test such systems, we prefer to wait
until someone is able to test and supply a pull request.
Added PTR_LEN (length of a pointer) and PTR_BITS (bit length
of a pointer) to longbits.h.
Moved calc version definition from version.c to version.h.
Sorted the order of symbols printed by "make env".
Test if <stdbool.h> exists and set HAVE_STDBOOL_H accordingly
in have_stdbool.h. Added HAVE_STDBOOL_H to allow one to force
this value.
Added "bool.h" include file to support use of boolean symbols,
true and false for pre-c99 C compilers. The "bool.h" include
file defines TRUE as true, FALSE as false, and BOOL as bool:
for backward compatibility.
Replaced in C source, TRUE with true, FALSE with false, and
BOOL with bool.
Fixed have_statfs optional executable file extension ${EXT{ in
the ${UTIL_PROGS} make variable.
Test if <stdint.h> exists and set HAVE_STDINT_H accordingly
in have_stdint.h. Added HAVE_STDINT_H to allow one to force
this value.
Test if <inttypes.h> exists and set HAVE_INTTYPES_H accordingly
in have_inttypes.h. Added HAVE_INTTYPES_H to allow one to force
this value.
Added c_chk.c to check the compiler and C include for calc
requirements. If you are unable to compile this program, or
if this program when compiles does not exit 0, then your C
compiler and/or C include fails to meet calc requirements.
Compilers that are at least c99 MUST be able to compile this
program such that when run will exit 0.
The "make hsrc" file will attempt to compile and run c_chk and
will warn if the C compiler and/or C include fails to meet
calc requirements. The "make debug" system will run c_chk -c
to print information about the C compiler and C include.
Currently failure to compile cc_chk.c or c_chk exiting non-0
will just print "WARNING!!" strings to stderr.
The make chk_c file also forms status.chk_c.h which either
defines CHK_C when the C compiler and select include files
appear to meet calc requirements, or undefines CHK_C
when it does not.
Added int.h as a central place for calc integer types and
integer macros.
Removed `-R release_file` and `-r release_file` command
line options from `ver_calc`. Add `-h` option. Updated
comments in "README.RELEASE", which serves as the contents
of the calc command "help release".
Added log2(x [,eps]) builtin function. When x is an integer
power of 2, log2(x) will return an integer, otherwise it will
return the equivalent of ln(x)/ln(2).
Removed CALC2_COMPAT in favor of ckecking if MAJOR_VER < 3.
The sign element in a ZVALUE is now of type SIGN, which is either
SB32 when MAJOR_VER < 3, or a bool otherwise.
The len element in a ZVALUE is of type LEN. LEN type is SB32 when
MAJOR_VER < 3, or a uintptr_t otherwise.
Setting an invalid epsilon via the epsilon(value) or confiv("epsilon",
value) triggers an error. The epsilon value must be: 0 < epsilon < 1.
Added new logn(x, n [,eps]) builtin to compute logarithms to base n.
Verify that eps arguments (error tolerance arguments that override
the default epsilon value) to builtin functions have proper values.
Previously the eps argument had little to no value checks for
many builtin functions.
Document in help files for builtin functions that take eps arguments,
the LIMIT range for such eps values.
Removed old Makefile testing rules for make dbx and make gdb.
Improved "make run" to execute calccalc using shared libraries
from the local directory, and with reading of the startup scripts
disabled.
Changed "make prep" to perform various tests that are used to
help verify that calc is ready for a release. Added the
update_ver tool, (formerly verupdate) and the trailblank tool
that existed outside of the calc source base but neverthless
used in the calc release process. Both of these tools are used
by "make prep".
Added Makefile testing rule "make testfuncsort" to check for
the sort of the builtin function list. Changed the order that
builtin functions are listed by "show builtin" and the help/builtin
to match the sorting of "LANG=C LC_ALL=C sort -d -u".
Added c_to_q(COMPLEX *c, bool cfree) to make is easier to convert
a COMPLEX value that is real (imag part is 0) into a NUMBER and
optionally free the COMPLEX value. The func.c code now uses c_to_q().
Added q_to_c(NUMBER *q) to make it easier to convert a NUMBER
into an allocated COMPLEX value.
Added new vercos(x, [,eps]) for versed cosine and covercos(x, [,eps])
for inverse versed cosine.
Added new avercos(x, [,eps]) for inverse versed cosine and acovercos(x, [,eps])
for inverse coversed cosine.
Improved comments about use of the ${CALC_ENV} Makefile variable.
Noted in Makefile.cal where and how the ${CALC_ENV} is used.
Use ${CALC_ENV} Makefile variable were needed.
Modified regression test cal/regress.cal: in some cases test numbers
were adjusted. Add comments indicate which test numbers apply to
which code. Indicated where there is room for new tests.
Expanded the end of test numbers from 9999 to 99999.
To make the meaning a bit more clear in cal/regress.cal, we have
renamed the following test calc resource files that are related to
the calc regression test suite:
cal/test1700.cal -> cal/test8000.read.cal
cal/test2300.cal -> cal/test2300.obj_incdec.cal
cal/test2600.cal -> cal/test2600.numfunc.cal
cal/test2700.cal -> cal/test2700.isqrt.cal
cal/test3100.cal -> cal/test3100.matobj.cal
cal/test3400.cal -> cal/test3400.trig.cal
cal/test4000.cal -> cal/test4000.ptest.cal
cal/test4100.cal -> cal/test4100.redc.cal
cal/test4600.cal -> cal/test4600.fileop.cal
cal/test5100.cal -> cal/test5100.newdecl.cal
cal/test5200.cal -> cal/test5200.globstat.cal
cal/test8400.cal -> cal/test8400.quit.cal
cal/test8500.cal -> cal/test8500.divmod.cal
cal/test8600.cal -> cal/test8600.maxargs.cal
cal/set8700.cal -> cal/test8700.dotest.cal
cal/test8900.cal -> cal/test8900.special.cal
cal/test9300.cal -> cal/test9300.frem.cal
Added to test 94dd, read of a number of new calc resource files
that are not already read as a result of the calc regression test suite.
Fixed more documentation and code comments that referred to the
old additive 55 (a55) shuffle pseudo-random number generator.
We have been using the subtractive 100 shuffle pseudo-random
number generator in place of the additive 55 generator for a
while now.
Improved help files trigonometric functions. They were corrected
to indicate that complex arguments are allowed: an oversight
from long ago when those trigonometric functions were expanded
to include complex arguments. The EXAMPLE sections were expanded
and made consistent, where applicable, across the trigonometric
help files. Documented libcalc functions in the SEE ALSO sections.
Improved "SEE ALSO" for the hyperbolic function help files.
Expanded the calc regression test suite test 34dd to test various
real and complex values for trigonometric functions.
Added complex multiple approximation function to commath.c so
that users of libcalc may directly round complex number to
nearest multiple of a given real number:
E_FUNC COMPLEX *cmappr(COMPLEX *c, NUMBER *e, long rnd, bool cfree);
For example:
COMPLEX *c; /* complex number to round to nearest epsilon */
NUMBER *eps; /* epsilon rounding precision */
COMPLEX *res; /* c rounded to nearest epsilon */
long rnd = 24L; /* a common rounding mode */
bool ok_to_free; /* true ==> free c, false ==> do not free c */
...
res = cmappr(c, eps, ok_to_free);
The complex trigonometric functions tan, cot, sec, csc were
implemented in func.c as calls to complex sin and complex cos.
We added the following direct calls to comfunc.c so that users
of libcalc may call them directly:
E_FUNC COMPLEX *c_tan(COMPLEX *c, NUMBER *eps);
E_FUNC COMPLEX *c_cot(COMPLEX *c, NUMBER *eps);
E_FUNC COMPLEX *c_sec(COMPLEX *c, NUMBER *eps);
E_FUNC COMPLEX *c_cot(COMPLEX *c, NUMBER *eps);
Added help/errorcodes rule to the top level Makefile.
Added E_USERMAX symbol (== 32767) to indicate the maximum value
allowed for user error codes.
Improved help/error. Added text about error code ranges and
range symbols.
Changed calc_errno a global int variable so that is may be directly
accessed by libcalc users.
Further improve help files for help/errno, help/error, help/newerror,
help/stoponerror and help/strerror by adding to documentation
of the calc error code system as well as libcalc interface
where applicable.
Changed #define E_USERDEF to #define E__USERDEF.
Removed use of E_USERDEF, E__BASE, E__COUNT, and E__HIGHEST
from custom/c_sysinfo because the c_sysinfo is just a demo
and this will simplify the custom/Makefile.
The include file calcerr.h is now the errsym.h include file.
The calcerr.tbl has been replaced by errtbl.c and errtbl.h.
The calcerr_c.awk, calcerr_c.sed, calcerr_h.awk, and
calcerr_h.sed files are now obsolete and have been removed.
The calcerr.c and calcerr.h now obsolete and are no longer built.
The calc computation error codes, symbols and messages are now in
a error_table[] array of struct errtbl.
An E_STRING is a string corresponds to an error code #define.
For example, the E_STRING for the calc error E_STRCAT,
is the string "E_STRING". An E_STRING must now match
the regular expression: "^E_[A-Z0-9_]+$".
The old array error_table[] of error message strings has been
replaced by a new error_table[] array of struct errtbl. The struct
errtbl array holds calc errnum error codes, the related E_STRING
symbol as a string, and the original related error message.
To add new computation error codes, add them near the bottom of the
error_table[] array, just before the NULL entry.
The ./errcode utility, when run, will verify the consistency of
the error_table[] array.
The Makefile uses ./errcode -e to generate the contents of
help/errorcodes file. The help errorcodes now prints
information from the new cstruct errtbl error_table[] array.
The help/errorcodes.hdr and help/errorcodes.sed files are
now obsolete and have been removed.
The Makefile uses ./errcode -d to generate the contents of the
errsym.h include file.
Code that used the old array error_table[] of error message strings
such as:
#include "calcerr.h"
char *msg; /* calc computation error message */
msg = error_table[errnum - E__BASE];
where errnum is the calc computation error code
E__BASE <= errnum <= E__HIGHEST, may now use:
#include "errtbl.h"
#include "errsym.h"
char *msg; /* calc computation error message */
msg = error_table[errnum - E__BASE].errmsg;
Rename the #define E__COUNT to ECOUNT to avoid confusion
with "E_STRING" error symbols.
Renamed "E_1OVER0" to "E_DIVBYZERO".
Renamed "E_0OVER0" to "E_ZERODIVZERO".
The verify_error_table() function that does a verification
the error_table[] array and setup private_error_alias[] array
is now called by libcalc_call_me_first().
Fix comment about wrong include file in have_sys_mount.h.
Removed unused booltostr() and strtobool() macros from bool.h.
Moved define of math_error(char *, ...) from zmath.h to errtbl.h.
The errtbl.h include file, unless ERRCODE_SRC is defined
also includes attribute.h and errsym.h.
Added E_STRING to error([errnum | "E_STRING"]) builtin function.
Added E_STRING to errno([errnum | "E_STRING"]) builtin function.
Added E_STRING to strerror([errnum | "E_STRING"]) builtin function.
Calling these functions with an E_STRING errsym is the same as calling
them with the matching errnum code.
Standardized on calc computation error related E_STRING strings
where there are a set of related codes. Changed "E_...digits" into
"E_..._digits". For example, E_FPUTC1 became E_FPUTC_1, E_FPUTC2
became E_FPUTC_2, and E_FPUTC3 became E_FPUTC_3. In a few cases
such as E_APPR became E_APPR_1, because there was a E_APPR2 (which
became E_APPR_2) and E_APPR3 (which became E_APPR_3). To other
special cases, E_ILOG10 became E_IBASE10_LOG and E_ILOG2 became
E_IBASE2_LOG because E_ILOG10 and E_ILOG2 are both independent calc
computation error related E_STRING strings. Now related sets of
E_STRING strings end in _ (underscore) followed by digits.
Added errsym builtin function. The errsym(errnum | "E_STRING")
builtin, , when given a valid integer errnum that corresponds to a
calc error condition, will return an E_STRING string, AND when given
a valid E_STRING string that is associated with a calc error
condition, will return errnum integer that corresponds to a calc
error condition.
Supplying a non-integer numeric errnum code to error(), errno(),
strerror(), or errsym() will result in an error.
Added tests to the calc regression test suite (cal/regress.cal) to
verify that the errnum calc computation error codes and their
E_STRING values have not changed.
Improved the clarity of calc regression suite (regress.cal) to mostly
use E_STRING errsym instead of numeric errnum values for error()
and errno() related tests.
Fixed SEE ALSO typo in help randperm.
Fixed calc regression test 42dd to set the display value back to 20.
Added to test 95dd and test9500.trigeq.cal to the calc regression test
suite to perform extensive test of trigonometric functions.
Added to test 34dd, some if the missing inverse trigonometric tests.
Improved builtin function strings, as printed by help builtin,
that use an optional accuracy (epsilon) arg by adding a comma.
The following are the changes from calc version 2.14.3.4 to 2.14.3.5:
Under macOS, to reduce dependency chains, we remove functions
and data that are unreachable by the entry point or exported
symbols. In particular, the macOS linker is used with both
"-dead_strip" and "-dead_strip_dylibs".
The libcalc shared library is now linked with libcustcalc.
The config("triground") controls rounding for the following
trigonometric and hyperbolic functions:
sin, cos, tan, cot, sec, csc
asin, acos, atan, acot, asec, acsc
versin, coversin, vercos, covercos
aversin, acoversin, avercos, acovercos
haversin, hacoversin, havercos, hacovercos
ahaversin, hacoversin, havercos, ahacovercos
exsec, aexsec, excsc, aexcsc
crd, acrd
cas, cis
sinh, cosh, tanh, coth, sech, csch
asinh, acosh, atanh, acoth, asech, acsch
In addition to taking a complex root (such as via the power
function on a complex value), "triground" is used for:
exp, polar
For the above mentioned functions, the rounding mode used to
round the result to the nearest epsilon value is controlled by,
and defaults to:
config("triground", 24)
As with other config options, the call returns the previous mode,
without a 2nd argument, returns the current mode without changing it:
config("triground")
When printing an error, calc used to print the errnum (error number):
; 1/0
Error 10001
Calc now prints the errsym (errsym):
; 1/0
Error E_DIVBYZERO
Added errsym E_LN_3 for ln(0).
Added errsym E_LOG_5 for log(0).
Added errsym E_LOG2_4 for log2(0).
Added errsym E_LOGN_6 for logn(0,base).
Added a chk_tree tool to help look for problems such as files that are
result of building calc that are also part of the calc distribution,
and files that are part of the calc source that are missing from the
calc distribution, and files that are of unknown status that are either
result of building calc nor missing from the calc distribution.
Updated file lists in Makefile, sorting as needed.
Updated Makefile PHONY rule to include Makefile rules that are NOT files.
Reduced make chatter for rules that build lists.
Added make verifydist to verify the existence of files that are part of
the calc distribution.
Added make verifydist to make prep.
Added a chk_tree double check, one after make clobber, one before
the final make chk, to make prep. Added double pass of chk_tree to
make full_debug (and thus to the make debug output).
Improved notes for install locations in Makefile.config.
Added printing of ${BUILD_ALL} to make env output.
Improved notes for install locations in Makefile.config.
Added printing of ${BUILD_ALL} to make env output.
Added double pass of chk_tree to make full_debug (and thus to the make
debug output).
The tarball for calc version 2.15.0.0 was missing version.h.
The version.h is now listed as part of the calc distribution.
Added a chk_tree tool to help look for problems such as files that are
result of building calc that are also part of the calc distribution,
and files that are part of the calc source that are missing from the
calc distribution, and files that are of unknown status that are either
result of building calc nor missing from the calc distribution.
Updated file lists in Makefile, sorting as needed.
Updated Makefile PHONY rule to include Makefile rules that are NOT files.
Reduced make chatter for rules that build lists.
Added make verifydist to verify the existence of files that are part of
the calc distribution.
Added make verifydist to make prep.
Added a chk_tree double check, one after make clobber, one before the
final make chk, to make prep.
The following are the changes in this release:
Added the following new trigonometric functions:
versin(x [,eps]) versed trigonometric sine
coversin(x [,eps]) coversed trigonometric sine
vercos(x [,eps]) versed trigonometric cosine
covercos(x [,eps]) coversed trigonometric cosine
aversin(x [,eps]) inverse versed trigonometric sine
acoversin(x [,eps]) inverse coversed trigonometric sine
avercos(x [,eps]) inverse versed trigonometric cosine
acovercos(x [,eps]) inverse coversed trigonometric cosine
haversin(x [,eps]) half versed trigonometric sine
hacoversin(x [,eps]) half coversed trigonometric sine
havercos(x [,eps]) half versed trigonometric cosine
hacovercos(x [,eps]) half coversed trigonometric cosine
ahaversin(x [,eps]) inverse half versed trigonometric sine
ahacoversin(x [,eps]) inverse half coversed trigonometric sine
ahavercos(x [,eps]) inverse half versed trigonometric cosine
ahacovercos(x [,eps]) inverse half coversed trigonometric cosine
exsec(x [,eps]) exterior trigonometric secant
aexsec(x [,eps]) inverse exterior trigonometric secant
excsc(x [,eps]) exterior trigonometric cosecant
aexcsc(x [,eps]) inverse exterior trigonometric cosecant
crd(x [,eps]) trigonometric chord of a unit circle
acrd(x [,eps]) inverse trigonometric chord of a unit circle
cas(x [,eps]) trigonometric cosine plus sine
cis(x [,eps]) Euler's formula
As Msys2 is a fork of Cygwin, if the OSNAME is Msys, the Cygwin
target will be used. Thanks to GitHub user @iahung2 for the
pull request.
Support for win32 and DJGPP has been dropped. Calc version
2.14.3.5 was the last to make references to win32 and make
references to DJGPP. Future versions of calc may work under
those systems, we just elected to remove the somewhat out of
date and awkward `win32.mkdef` and related win32 references.
If you are a win32 user, please feel free to create a win32
target in Makefile.target and submit as a pull request.
If you are a DJGPP user, please feel free to create a DJGPP
target in Makefile.target and submit as a pull request.
Until someone can test such systems, we prefer to wait
until someone is able to test and supply a pull request.
Added PTR_LEN (length of a pointer) and PTR_BITS (bit length
of a pointer) to longbits.h.
Moved calc version definition from version.c to version.h.
Sorted the order of symbols printed by "make env".
Test if <stdbool.h> exists and set HAVE_STDBOOL_H accordingly
in have_stdbool.h. Added HAVE_STDBOOL_H to allow one to force
this value.
Added "bool.h" include file to support use of boolean symbols,
true and false for pre-c99 C compilers. The "bool.h" include
file defines TRUE as true, FALSE as false, and BOOL as bool:
for backward compatibility.
Replaced in C source, TRUE with true, FALSE with false, and
BOOL with bool.
Fixed have_statfs optional executable file extension ${EXT{ in
the ${UTIL_PROGS} make variable.
Test if <stdint.h> exists and set HAVE_STDINT_H accordingly
in have_stdint.h. Added HAVE_STDINT_H to allow one to force
this value.
Test if <inttypes.h> exists and set HAVE_INTTYPES_H accordingly
in have_inttypes.h. Added HAVE_INTTYPES_H to allow one to force
this value.
Added c_chk.c to check the compiler and C include for calc
requirements. If you are unable to compile this program, or
if this program when compiles does not exit 0, then your C
compiler and/or C include fails to meet calc requirements.
Compilers that are at least c99 MUST be able to compile this
program such that when run will exit 0.
The "make hsrc" file will attempt to compile and run c_chk and
will warn if the C compiler and/or C include fails to meet
calc requirements. The "make debug" system will run c_chk -c
to print information about the C compiler and C include.
Currently failure to compile cc_chk.c or c_chk exiting non-0
will just print "WARNING!!" strings to stderr.
The make chk_c file also forms status.chk_c.h which either
defines CHK_C when the C compiler and select include files
appear to meet calc requirements, or undefines CHK_C
when it does not.
Added int.h as a central place for calc integer types and
integer macros.
Removed `-R release_file` and `-r release_file` command
line options from `ver_calc`. Add `-h` option. Updated
comments in "README.RELEASE", which serves as the contents
of the calc command "help release".
Added log2(x [,eps]) builtin function. When x is an integer
power of 2, log2(x) will return an integer, otherwise it will
return the equivalent of ln(x)/ln(2).
Removed CALC2_COMPAT in favor of ckecking if MAJOR_VER < 3.
The sign element in a ZVALUE is now of type SIGN, which is either
SB32 when MAJOR_VER < 3, or a bool otherwise.
The len element in a ZVALUE is of type LEN. LEN type is SB32 when
MAJOR_VER < 3, or a uintptr_t otherwise.
Setting an invalid epsilon via the epsilon(value) or confiv("epsilon",
value) triggers an error. The epsilon value must be: 0 < epsilon < 1.
Added new logn(x, n [,eps]) builtin to compute logarithms to base n.
Verify that eps arguments (error tolerance arguments that override
the default epsilon value) to builtin functions have proper values.
Previously the eps argument had little to no value checks for
many builtin functions.
Document in help files for builtin functions that take eps arguments,
the LIMIT range for such eps values.
Removed old Makefile testing rules for make dbx and make gdb.
Improved "make run" to execute calccalc using shared libraries
from the local directory, and with reading of the startup scripts
disabled.
Changed "make prep" to perform various tests that are used to
help verify that calc is ready for a release. Added the
update_ver tool, (formerly verupdate) and the trailblank tool
that existed outside of the calc source base but neverthless
used in the calc release process. Both of these tools are used
by "make prep".
Added Makefile testing rule "make testfuncsort" to check for
the sort of the builtin function list. Changed the order that
builtin functions are listed by "show builtin" and the help/builtin
to match the sorting of "LANG=C LC_ALL=C sort -d -u".
Added c_to_q(COMPLEX *c, bool cfree) to make is easier to convert
a COMPLEX value that is real (imag part is 0) into a NUMBER and
optionally free the COMPLEX value. The func.c code now uses c_to_q().
Added q_to_c(NUMBER *q) to make it easier to convert a NUMBER
into an allocated COMPLEX value.
Added new vercos(x, [,eps]) for versed cosine and covercos(x, [,eps])
for inverse versed cosine.
Added new avercos(x, [,eps]) for inverse versed cosine and acovercos(x, [,eps])
for inverse coversed cosine.
Improved comments about use of the ${CALC_ENV} Makefile variable.
Noted in Makefile.cal where and how the ${CALC_ENV} is used.
Use ${CALC_ENV} Makefile variable were needed.
Modified regression test cal/regress.cal: in some cases test numbers
were adjusted. Add comments indicate which test numbers apply to
which code. Indicated where there is room for new tests.
Expanded the end of test numbers from 9999 to 99999.
To make the meaning a bit more clear in cal/regress.cal, we have
renamed the following test calc resource files that are related to
the calc regression test suite:
cal/test1700.cal -> cal/test8000.read.cal
cal/test2300.cal -> cal/test2300.obj_incdec.cal
cal/test2600.cal -> cal/test2600.numfunc.cal
cal/test2700.cal -> cal/test2700.isqrt.cal
cal/test3100.cal -> cal/test3100.matobj.cal
cal/test3400.cal -> cal/test3400.trig.cal
cal/test4000.cal -> cal/test4000.ptest.cal
cal/test4100.cal -> cal/test4100.redc.cal
cal/test4600.cal -> cal/test4600.fileop.cal
cal/test5100.cal -> cal/test5100.newdecl.cal
cal/test5200.cal -> cal/test5200.globstat.cal
cal/test8400.cal -> cal/test8400.quit.cal
cal/test8500.cal -> cal/test8500.divmod.cal
cal/test8600.cal -> cal/test8600.maxargs.cal
cal/set8700.cal -> cal/test8700.dotest.cal
cal/test8900.cal -> cal/test8900.special.cal
cal/test9300.cal -> cal/test9300.frem.cal
Added to test 94dd, read of a number of new calc resource files
that are not already read as a result of the calc regression test suite.
Fixed more documentation and code comments that referred to the
old additive 55 (a55) shuffle pseudo-random number generator.
We have been using the subtractive 100 shuffle pseudo-random
number generator in place of the additive 55 generator for a
while now.
Improved help files trigonometric functions. They were corrected
to indicate that complex arguments are allowed: an oversight
from long ago when those trigonometric functions were expanded
to include complex arguments. The EXAMPLE sections were expanded
and made consistent, where applicable, across the trigonometric
help files. Documented libcalc functions in the SEE ALSO sections.
Improved "SEE ALSO" for the hyperbolic function help files.
Expanded the calc regression test suite test 34dd to test various
real and complex values for trigonometric functions.
Added complex multiple approximation function to commath.c so
that users of libcalc may directly round complex number to
nearest multiple of a given real number:
E_FUNC COMPLEX *cmappr(COMPLEX *c, NUMBER *e, long rnd, bool cfree);
For example:
COMPLEX *c; /* complex number to round to nearest epsilon */
NUMBER *eps; /* epsilon rounding precision */
COMPLEX *res; /* c rounded to nearest epsilon */
long rnd = 24L; /* a common rounding mode */
bool ok_to_free; /* true ==> free c, false ==> do not free c */
...
res = cmappr(c, eps, ok_to_free);
The complex trigonometric functions tan, cot, sec, csc were
implemented in func.c as calls to complex sin and complex cos.
We added the following direct calls to comfunc.c so that users
of libcalc may call them directly:
E_FUNC COMPLEX *c_tan(COMPLEX *c, NUMBER *eps);
E_FUNC COMPLEX *c_cot(COMPLEX *c, NUMBER *eps);
E_FUNC COMPLEX *c_sec(COMPLEX *c, NUMBER *eps);
E_FUNC COMPLEX *c_cot(COMPLEX *c, NUMBER *eps);
Added help/errorcodes rule to the top level Makefile.
Added E_USERMAX symbol (== 32767) to indicate the maximum value
allowed for user error codes.
Improved help/error. Added text about error code ranges and
range symbols.
Changed calc_errno a global int variable so that is may be directly
accessed by libcalc users.
Further improve help files for help/errno, help/error, help/newerror,
help/stoponerror and help/strerror by adding to documentation
of the calc error code system as well as libcalc interface
where applicable.
Changed #define E_USERDEF to #define E__USERDEF.
Removed use of E_USERDEF, E__BASE, E__COUNT, and E__HIGHEST
from custom/c_sysinfo because the c_sysinfo is just a demo
and this will simplify the custom/Makefile.
The include file calcerr.h is now the errsym.h include file.
The calcerr.tbl has been replaced by errtbl.c and errtbl.h.
The calcerr_c.awk, calcerr_c.sed, calcerr_h.awk, and
calcerr_h.sed files are now obsolete and have been removed.
The calcerr.c and calcerr.h now obsolete and are no longer built.
The calc computation error codes, symbols and messages are now in
a error_table[] array of struct errtbl.
An E_STRING is a string corresponds to an error code #define.
For example, the E_STRING for the calc error E_STRCAT,
is the string "E_STRING". An E_STRING must now match
the regular expression: "^E_[A-Z0-9_]+$".
The old array error_table[] of error message strings has been
replaced by a new error_table[] array of struct errtbl. The struct
errtbl array holds calc errnum error codes, the related E_STRING
symbol as a string, and the original related error message.
To add new computation error codes, add them near the bottom of the
error_table[] array, just before the NULL entry.
The ./errcode utility, when run, will verify the consistency of
the error_table[] array.
The Makefile uses ./errcode -e to generate the contents of
help/errorcodes file. The help errorcodes now prints
information from the new cstruct errtbl error_table[] array.
The help/errorcodes.hdr and help/errorcodes.sed files are
now obsolete and have been removed.
The Makefile uses ./errcode -d to generate the contents of the
errsym.h include file.
Code that used the old array error_table[] of error message strings
such as:
#include "calcerr.h"
char *msg; /* calc computation error message */
msg = error_table[errnum - E__BASE];
where errnum is the calc computation error code
E__BASE <= errnum <= E__HIGHEST, may now use:
#include "errtbl.h"
#include "errsym.h"
char *msg; /* calc computation error message */
msg = error_table[errnum - E__BASE].errmsg;
Rename the #define E__COUNT to ECOUNT to avoid confusion
with "E_STRING" error symbols.
Renamed "E_1OVER0" to "E_DIVBYZERO".
Renamed "E_0OVER0" to "E_ZERODIVZERO".
The verify_error_table() function that does a verification
the error_table[] array and setup private_error_alias[] array
is now called by libcalc_call_me_first().
Fix comment about wrong include file in have_sys_mount.h.
Removed unused booltostr() and strtobool() macros from bool.h.
Moved define of math_error(char *, ...) from zmath.h to errtbl.h.
The errtbl.h include file, unless ERRCODE_SRC is defined
also includes attribute.h and errsym.h.
Added E_STRING to error([errnum | "E_STRING"]) builtin function.
Added E_STRING to errno([errnum | "E_STRING"]) builtin function.
Added E_STRING to strerror([errnum | "E_STRING"]) builtin function.
Calling these functions with an E_STRING errsym is the same as calling
them with the matching errnum code.
Standardized on calc computation error related E_STRING strings
where there are a set of related codes. Changed "E_...digits" into
"E_..._digits". For example, E_FPUTC1 became E_FPUTC_1, E_FPUTC2
became E_FPUTC_2, and E_FPUTC3 became E_FPUTC_3. In a few cases
such as E_APPR became E_APPR_1, because there was a E_APPR2 (which
became E_APPR_2) and E_APPR3 (which became E_APPR_3). To other
special cases, E_ILOG10 became E_IBASE10_LOG and E_ILOG2 became
E_IBASE2_LOG because E_ILOG10 and E_ILOG2 are both independent calc
computation error related E_STRING strings. Now related sets of
E_STRING strings end in _ (underscore) followed by digits.
Added errsym builtin function. The errsym(errnum | "E_STRING")
builtin, , when given a valid integer errnum that corresponds to a
calc error condition, will return an E_STRING string, AND when given
a valid E_STRING string that is associated with a calc error
condition, will return errnum integer that corresponds to a calc
error condition.
Supplying a non-integer numeric errnum code to error(), errno(),
strerror(), or errsym() will result in an error.
Added tests to the calc regression test suite (cal/regress.cal) to
verify that the errnum calc computation error codes and their
E_STRING values have not changed.
Improved the clarity of calc regression suite (regress.cal) to mostly
use E_STRING errsym instead of numeric errnum values for error()
and errno() related tests.
Fixed SEE ALSO typo in help randperm.
Fixed calc regression test 42dd to set the display value back to 20.
Added to test 95dd and test9500.trigeq.cal to the calc regression test
suite to perform extensive test of trigonometric functions.
Added to test 34dd, some if the missing inverse trigonometric tests.
Improved builtin function strings, as printed by help builtin,
that use an optional accuracy (epsilon) arg by adding a comma.
When printing an error, calc used to print the errnum (error number):
; 1/0
Error 10001
Calc now prints the errsym (errsym):
; 1/0
Error E_DIVBYZERO
Added errsym E_LN_3 for ln(0).
Added errsym E_LOG_5 for log(0).
Added errsym E_LOG2_4 for log2(0).
Added errsym E_LOGN_6 for logn(0,base).
