north-numerical-computing/cpfloat

Custom-Precision Floating-point numbers.

cpfloat's Issues

emax/emin potential typo

Probably line 1202 was meant to say params.emin = DEFEMIN; because params.emax is already set on line 1201.

cpfloat/src/cpfloat_template.h

Lines 1201 to 1202 in 7dabbe1

	params.emax = DEFEMAX; \
	params.emax = DEFEMIN; \

Documentation about scripts needed to run tests locally

This is not an issue in the code but more on the documentation side. I am putting it here so that we don't forget to address it at some point.

I think we need to provide some information in README on the testing framework and what dependencies need to be installed.

I tried to set it up on macOS and I noticed that I am missing:

check
checkmk
subunit
rt

Finding these packages is quite difficult since their names are quite generic. We could provide links (and their full names, short descriptions) to the sources of these packages in README which may help. I was able to easily install check and checkmk through Homebrew but I could not find subunit there and had to build it from source. As for rt, I still do not know what it is (perhaps the Request Tracker https://rt-wiki.bestpractical.com/wiki/ManualInstallation).

Potential retval inconsistencies

retval settings on

cpfloat/src/cpfloat_template.h

Lines 297 to 304 in 7dabbe1

	if (precision > DEFPREC) {
	precision = DEFPREC;
	*retval = 1;
	}
	if (emax > DEFEMAX) {
	emax = DEFEMAX;
	*retval = 2;
	}

are not consistent with what is done in VALIDATE_INPUT. Do we need these checks in COMPUTE_GLOBAL_PARAMS when it is already done in VALIDATE_INPUT?