Precision flag not passed from run_card properly (overwritten by cudacpp_config)

[zeniheisser]

The parameter 'floating_type' in the run_card is not passed along to the cudacpp compilation.

In launch_plugin.py, lines 35-37, the parameter is explicitly read from the run_card and is passed along using the common_run_interface.CommonRunCmd.update_make_opts_full function to Source/make_opts

However, in cudacpp.mk, the file cudacpp_config.mk is included before make_opts. In cudacpp_config.mk, lines 18-23,

 ifeq ($(FPTYPE),)
    # OLD DEFAULT UP TO v1.00.00 INCLUDED (inconsistent with default floating_type='m' in run_card.dat)
    ###override FPTYPE = d
    # NEW DEFAULT (#995) AS OF v1.00.01 (now consistent with default floating_type='m' in run_card.dat)
    override FPTYPE = m
 endif

FPTYPE is set to m with the override directive, which has higher priority than setting the variable anywhere else. In make_opts, FPTYPE is set without the override flag, so the default is used, regardless of any definition in make_opts.

Could be solved by either changing include order or adding an override directive specifically to the FPTYPE flag in make_opts.

[zeniheisser]

Partially fixed with #1030 which changes the declaration in make_opts from "FPTYPE" to "override FPTYPE".

Not a general solution: At runtime, additional arguments are appended at the TOP of make_opts, so changing the run_card will not affect future runs in the same directory