PoissonBinomial 1.2.7

Updated Makevars file, because it contained hardcoded paths which could cause problems on some exotic configurations, e.g. MacPorts
Minor performance improvement for DFT-CF procedure
Documentation updates

PoissonBinomial 1.2.6

Addresses [-Wformat-security] compiler warning on newer clang configuration. See https://github.com/RcppCore/Rcpp/issues/1287
change of maintainer eMail to personal address
GitHub repo moved to personal account

Further optimizations of determining the number of splits for “DivideFFT” procedures.

Performance improvements for “Convolve” (and subsequently “DivideFFT”) procedures.
GCD optimizations for generalized Poisson distributions have been moved to the respective C++ functions, so that packages that import them may benefit from them as well.
Removed dependence on BH package, as it was only needed for one constant in the C++ code, which has now been defined manually.
Adjustment to vignettes. Benchmarks (for performance comparisons) are now done with 51 runs (instead of 100) to accelerate building of the package.
Fixed a minor bug with qgpbinom.

Performance improvements for exact methods of dgpbinom and pgpbinom.
Input variable of the Rcpp implementations of all methods are made const to prevent inadvertent changes to them. Any package that imports headers must be updated. The ‘Imports’ field of the DESCRIPTION file should include a version requirement, i.e. PoissonBinomial (>= 1.2.0).
Added new random generation methods for rpbinom and rgpbinom (see function documentation). They are much faster than the old quantile-based inversion method, which has been removed.

Improved numerical accuracy of normal approximations of dpbinom and dgpbinom.

Bug fixes and performance improvements for qpbinom and qgpbinom that also affect rpbinom and rgpbinom. Quantiles were off be one; all code that uses the quantile functions should be reviewed!
When requesting cumulative probabilities, the respective C++ implementations are now capable of computing these values for lower.tail = FALSE on their own, which improves accuracy.

Bug fixes in ppbinom and pgpbinom that caused incorrect calculation of logarithms and cumulative upper-tail probabilities.

Added exact and approximate algorithms for the generalized Poisson binomial distribution described in Zhang, Hong & Balakrishnan (2018). The non-generalized distribution is now referred to as the ‘ordinary’ Poisson binomial distribution.
Restructured vignettes. Added tables of content and fixed smaller issues.
Minor bug fixes for dbinom, ppbinom and qpbinom functions.

Improvements of C++ helper function norm_dpb to achieve better normalization.
Bug fix of DFT-CF method (“Characteristic”) so that negative probabilities are no longer possible.
Reworked vignette structure.
Added author acknowledgments to the Makevars.win file (original author was Geoff99 (https://github.com/Geoff99)).

Fixed a bug in the C++ helper function “norm_dpb” that could cause infinite loops (the function is invisible to the user, since it is only used in the C++ domain).