I've just published Will Clinger's latest improvements to
the reference implementation of SRFI 144.

Here is the diff:

  <https://github.com/scheme-requests-for-implementation/srfi-144/commit/8eecfc6eef1142e7b728aff75d98ce28a3adc844>

Here are Will's comments on these changes:

  The sample implementation has now been tested in Larceny,
  Sagittarius, Chibi Scheme, Kawa, and Gauche. So far as I
  can tell, all failed tests are caused by roundoff or other
  minor errors in the current versions of those systems,
  often involving infinities or NaNs. Larceny's current
  development system passes all tests, with or without using
  an FFI.

  The hyperbolic functions are now more accurate, thanks to
  Bradley Lucier's comment on the previous version.

  Commit Summary

  * Merge pull request #1 from scheme-requests-for-implementation/master
  * Fixed several portability problems.
  * Changed flexponent to correspond with C99 logb.
  * Merge https://github.com/scheme-requests-for-implementation/srfi-144
  * Improved accuracy of hyperbolic functions.
  * Minor cleanup.
  * One-line cleanup.

  File Changes

  * M srfi-144-constants.scm (6)
  * M srfi/144.body.scm (149)
  * A srfi/144.constants.scm (61)
  * M srfi/144.r6rs.scm (55)
  * M srfi/144.sld (1)
  * M srfi/144.special.scm (93)
  * M srfi/README (6)
  * M tests/README (16)
  * M tests/scheme/flonum.sld (14)
  * M tests/scheme/test.sld (1)
  * M tests/srfi-144-test.scm (30)

Regards,

SRFI Editor