Typedefs and function pointers

[aldanor]

Typedefs and function pointers

[aldanor]

... so that should probably be converted to:

alias tibrvMsgData_Encoder = tibrv_status function(
    char** wire_buffer, tibrv_u32 mem_available, tibrvMsgField* field);

[jacob-carlborg]

Clang cannot compile that code, that's also what the error says 😉. It seems it cannot find tibrv_status. I think Clang uses int here as a form of default type.

[aldanor]

Oh, so that could have happened because I was converting each header one by one? The status thing is just a typedef defined in another header that gets included. This hasn't caused problems with anything but this issue though.

[jacob-carlborg]

You basically need to convert the headers one by one. That's only what's currently supported by DStep. C is a very annoying language in this area. If a given header is always included together with other headers it could very well be that this header is using symbols from other headers but is not explicitly including them itself. Personally I would say that is a broken design.

If this is the case I would recommend you contribute a fix to the project, if it's open source. Otherwise you could just add the missing includes when running DStep on the files and then remove them.

Hmm, I'm wondering if it's not able to find a header file, but in that case it would say so. In that case you need to use the -I flag to indicate where it can find the header files.

I just noticed that Clang has the a flag that could be useful

-include <file>         Include file before parsing

[aldanor]

Yea. I wonder if one could figure out a correct merge order based on local #include "..." directives from a set of header files? And then just dump everything into a single huge D module. There's many issues if you do it one by one and then try to merge it manually -- for instance, I had to fix a good number of Anonymous_0, Anonymous_1 name clashes (see #8). Or issues like the one above.

[jacob-carlborg]

I updated my comment above.

[jacob-carlborg]

The idea is that you should not merge the D modules. You need to manually add import-statements instead. That's something DStep doesn't handle automatically.

[aldanor]

In regards to the original question, indeed, the tibrv_status was defined in the file that wasn't directly included in the target header file (status.h), so one would need to pass the -include flag, as @jacob-carlborg has pointed out above:

dstep handler.h -o handler.d -include status.h

which would then correctly generate

extern(C):

alias  function (char**, uint, tibrvMsgField*) tibrvMsgData_Encoder;

Perhaps this could be put in the docs as it happens very often in C libraries.

[jacob-carlborg]

Personally I think this is a workaround and not a solution. I think it's a broken design of the header files. But I put it in the documentation anyway.

[jacob-carlborg]

@aldanor I added a section for limitations/known issue to the readme: https://github.com/jacob-carlborg/dstep#limitationsknown-issues.