Add >>IMP INCLUDE directive

[engboris]

Closes #114 and completes FR176 on SourceForge.

A new directive >> IMP INCLUDE FILE_1 ... FILE_n is introduced, allowing to include multiple C headers. Only files with name finishing by .h are allowed.

I'm not sure about that but I marked the feature with /* GnuCOBOL 3.3 extension*/.

[GitMensch]

Thanks for this first take - please copy the tests for --include and adjust those for the directive. In gnbucobol.texi and NEWS add a remark in the --include description, that the directive can be used as an alternative.

[GitMensch]

We don't need two lists, but we need to reset the list for each source file specified on the command line to the state from cobc command line. Therefore either backup the whole list and restore it (similar to cb_source_format) or just use two lists and reset the second before each new source file. I'm ok with both ways but think the two-list approach is easier to handle.

[GitMensch]

still some requested changes - but we're on the finish line with this PR

[GitMensch]

"approved with minor changes"

[GitMensch]

changes to the testcases needed, overal fine

[engboris]

@GitMensch Could you check this? It was almost finished and it was a rather minor change but I wasn't sure about the tests. Thank you!

[GitMensch]

code changes requested for @engboris - I'll adjust the used_binaries.at, so please don't force-push

[GitMensch]

Currently the memory was allocated with cobc_malloc and would need to be always freed, and that must also be done when there is no codegen (preprocess/syntaxy-check only) so must be done outside of codegen.

I think the parse_ memory functions would be more reasonable and drop the need to execute the free, also removing the need to keep the "last" element here; the only thing missing then is to initialize cb_include_file_list_directive at the right place

[engboris]

If I understand correctly, I should free those variables outside/independently of codegen but still after the running of codegen? Maybe after ylex_clear_all () in cobc.c and add a parse_clear_all () function or something like that?

[GitMensch]

I've rebased this PR and adjusted the "real" test for IMP INCLUDE. There are still open changes (see review notes), but I'm sure @engboris will have a look at those over the next days (or hours)?

[engboris]

I'm not sure if this was what you expected. The only error I get for missing arguments is this one. There is no other errors displayed. You talked about "resetting the error".