hadron missing headers in configure.ac

[kostrzewa]

probably, we should fix this...

[email addresses and recipients redacted]

A recent update to the toolchain on macOS is causing configure failures if
exit() is used undeclared in conftest.c fragments (#include <stdlib.h> is
needed).

This appears to affect packages

DRIP diversitree hadron island lwgeom rgdal rgeos ridge uFTIR vapour

NB: for some of these some but not all configure tests include stdlib.h, so
this list comes from examining the code, not just empirical testing. I failed
to install some of the spatial packages on macOS using SU's binaries, but

DRIP diversitree hadron island

all failed to install and ridge failed to find GSL.

It seems likely that this toolchain update was premature (by Apple -- has
happened before), as it installed a beta toolchain. But it would be good to
have this corrected sooner rather than later as it is likely this will become
the toolchain for macOS 11.0 within a few weeks.

[urbach]

see branch CRAN-3.1.0-bugfix.

I don't know yet how to upload a bugfix to CRAN. We should probably also keep this branch for further bugfixes.

[urbach]

or shall we just take the latest version?

[martin-ueding]

I would just publish whatever latest version that we have. I mean we don't have to support some particular feature versions.

[martin-ueding]

Also that fix in the configure.ac is trivial, and should not break anything. I think that we can just merge that into the master and then try to update a new version for CRAN.

[urbach]

Also that fix in the `configure.ac` is trivial, and should not break anything. I think that we can just merge that into the `master` and then try to update a new version for CRAN.

I think I already merged it. Do we actually import anything from package R6?

[martin-ueding]

Yes, I use the R6 in the new_matrixfit models.