Parse C/C++ tags wrapped in an extern "C" {...} block

[ntrel]

Geany fails to parse tags wrapped in an extern "C" {...} block, even in C++. The construct should be recognised for the C parser also, so this pattern is supported:

#ifdef __cplusplus
extern "C"
{
#endif

extern int i; // should be parsed as a global

#ifdef __cplusplus
}
#endif

[elextr]

Agree, although inside {} these are really global declarations.

[ntrel]

@elextr Yep, added extern.

[ntrel]

BTW fishman ctags does parse extern C blocks, and our code is probably meant to (e.g. isExternCDecl). I tried to investigate but couldn't find the problem, the parsers are a bit too divergent.

[b4n]

Err, what are you using? For me it works just fine, and produces an Extern Variables i tag for line 6, just as expected.

[codebrainz]

I'm having this issue too, on Windows 10, types are not highlighted correctly inside extern "C" blocks (whether guarded or not). For example,

#ifdef __cplusplus
extern "C" {
#endif
typedef struct SomeStruct SomeTypedef;
struct SomeStruct
{
    int x;
};
#ifdef __cplusplus
}
#endif

Doesn't highlight the type name "SomeTypedef" in C, but it does for the "SomeStruct". If I switch the filetype to C++, neither type is highlighted.

It must be Windows-specific, I know this works fine Linux for sure.

[elextr]

Confirm that like @b4n the test code used by @codebrainz works fine on linux.

[codebrainz]

As a workaround, putting the extern "C" { in a separate header a la G_BEGIN_DECLS fixes tag parsing inside the block.

[eht16]

Just for the record, I tested the code snippet on Windows with the ctags binary we ship with Geany-Plugins (originating from the MSYS2 ctags package) and this binary also does not parse the extern variable.
Still no clue what could cause the different behaviour in parsing on Windows :(.

[masatake]

As far as trying the input directly with u-ctags, it captures all expected tags.

[techee]

@eht16 Did the new cxx parser fix this issue on Windows?

[eht16]

It does!

I guess we can close this now.

[techee]

Good, one more mysterious bug away.

[kugel-]

Maybe have a unit test for this?

[techee]

Maybe have a unit test for this?

Yeah, but parser problems should mostly be business of uctags now. And I think the new cxx parser is tested quite well there.