Changes in version 1.2.0 (2026-01-28)                  

  - nlsic(): added reuse of jacobian in non linear iterations
      - con$reuse_jac=TRUE (do it by default)
      - con$max_reuse=5L (default maximal number of jacobian reuses)
  - nlsic(): con$btfrac default value is brought to 0.8 instead of 0.5.
  - lsi() and lsi_ln() return x with aqr attribute containing QR
    decomposition of matrix a. This is used to spare some calculations
    in non linear iterations

                 Changes in version 1.1.1 (2025-06-27)                  

  - ls_ln() and ls_0(): fixed case when rhs, b, is a multi-column matrix

                 Changes in version 1.1.0 (2025-05-16)                  

  - lsie_ln(): added mnorm and x0 parameters
  - lsi_ln(): rewritten from scratch to use lsie_ln() recursively
  - added pnull() and ls_0() functions
  - ls_ln() now gets mnorm and x0 parameters
  - complited unit tests
  - warning() is called in problematic situations. So user can intercept
    and debug them

                 Changes in version 1.0.4 (2023-06-26)                  

  - nlsic(): fixed error reporting when step has NA
  - nlsic(): fixed error reporting when inequalities are not satisfied

                        Changes in version 1.0.3                        

  - nlsic(): fixed error when par is a matrix and u=NULL
  - nlsic(): added paro field in returned result with original structure
    of par

                 Changes in version 1.0.2 (2022-04-12)                  

  - now, lsi() returns NA vector if rhs is all NA.

                 Changes in version 1.0.1 (2022-01-11)                  

  - Added a NEWS.md file to track changes to the package.
  - minor fixes in documentation