The hamfisted heuristic `*.c*` matching trick will fail pretty soon...

[xparq]

Matching false positives like local.cfg or broken.cpp.off etc.

[xparq]

Added obj_source_exts=.cpp .cxx .c, but that's mostly just a reminder... Only used by the lib source collector rule yet, and even that's for a slightly different purpose! :)

[xparq]

Hehh, OK, didn't have to wait long... If there's just a dir with no files, but only subdirs, in the tree (like a collection of modules or a dispatching point etc., as in the sfw repo, where I've been testing this), it blows right up already...

• [x] As a (kinda also hamfisted...) workaround for that, the driver loop now checks for existing sources in every dir and calls NMAKE only if there're things to do:

  if exist %%i\*.cpp !_make_! /c compiling DIR=!_dir_!
  if exist %%i\*.cxx !_make_! /c compiling DIR=!_dir_!
  if exist %%i\*.c   !_make_! /c compiling DIR=!_dir_!

• [x] Well, as we have $(obj_source_exts) (as .cpp .cxx .c) already, this should be an inner for loop iterating that!

• [x] Also, it could probably send the type info then too, which could allow removing the *.c* hack finally.

• [ ] But another similar hack still remains at the lib creation rule:

  for /r "$(src_dir)\$(lib_src_subdir)" %%o in ($(units_pattern).c*) do (