NEWS

coinclp 0.1.2

The licence changes from EPL to MIT. License: EPL in an R package means EPL 1.0, which the Free Software Foundation and the Eclipse Foundation both record as incompatible with the GPL, and that blocked GPL-licensed packages from building on this one. MIT is compatible with the GPL and with the Eclipse Public License of Clp alike, so both sides work. The code is unchanged, and the permissions granted to users are wider than before, not narrower.
inst/COPYRIGHTS now records that binary builds contain Clp object code under EPL 2.0, because the Clp libraries are static. The previous wording described only the source package, where Clp is genuinely absent.

CRAN's valgrind run reported uninitialised bytes written to a file from clp_save_model(). The bytes are the trailing padding of a struct that Clp's own saveModel() writes whole (ClpSimplex.cpp, Clp_scalars), so the report comes from inside the Clp library and is harmless, but it cannot be silenced from R. The snapshot round trip is therefore no longer exercised in CRAN's checks: the example on the clp_save_model help page is marked not to run, and the corresponding test runs only where NOT_CRAN is set, as it is on the package's continuous integration. The functions themselves are unchanged, and the help page explains the finding for anyone who runs valgrind themselves.
src/Makevars.win now honours the CLP_CFLAGS and CLP_LIBS environment variables, so a Clp installed outside the Rtools tree can be built against on Windows too. They take precedence over pkg-config, which in turn precedes the Rtools library tree, matching the order configure uses on Unix.
The claim that Rtools ships Clp is now stated as what has been verified (Rtools 4.5, with 4.3 and 4.4 using the same library tree) rather than as a blanket one. Rtools 4.2 and earlier do not carry Clp, so source builds on R 4.2 or older need CLP_CFLAGS and CLP_LIBS.
README says what a Windows user actually needs: nothing at all for a CRAN binary, since Clp is static and ends up inside coinclp.dll; Rtools only for a source build.

First release.
clp_solve() solves a linear program in one call, from a dense matrix, a Matrix sparse matrix, a slam::simple_triplet_matrix or i/j/v triplets, with one-sided or ranged rows.
Complete bindings to the Clp callable library: problem construction, bounds, objective, matrix access, all six solve entry points, presolve options, basis access and warm starts, tolerances and limits, status queries, names, MPS input and output, and model snapshots.
initProbCLP() and the other *CLP functions reproduce the interface of the archived clpAPI package, including the clpPtr S4 class.
configure finds Clp through --with-clp-*, CLP_CFLAGS/CLP_LIBS, pkg-config or the default compiler paths, and detects optional Clp entry points by link test. On Windows the Clp that Rtools ships is used.