Re: Unicode-based FreeBSD

Alexander,

OK, now it's almost clear. Modified and added kernel source files just stay
where they are, integration test application goes to "src/tools/regression".

The only thing remaining to discuss is unit-tests. Currently I have a single
user-mode application that performs unit-tests for UTF-8 decoder. The
problem with just moving it to "src/tools/regression" is that unlike
integration tests, unit-tests link statically to code they test. For example
if I've added file utf8decoder.c to sys/dev/syscons/unicode, then related
unit-test application would consist of utf8decoder-test.c and utf8decoder.c,
producing "utf8deocder-unit-test" binary.

Linking statically to several kernel source files from
"src/tools/regression" and compiling them separately for that purpose seems
very obscure for me. Isn't it?

For my working copy I've chosen the following way: put unit-test sources
under sys/dev/syscons/unicode with their own Makefiles and let them compile
kernel source files in their own way. As I understand, this does not
interfere with currently used kernel build system, since those unit-test
sources are not listed in kernel build description file.

The purpose of this discussion is to verify that this approach is allowed
for the FreeBSD kernel. Is this permitted?

Alexander Churanov