C identifiers with 2 leading underscores fail to parse correctly

[puredrivel]

Identifiers with 2 leading underscores fail to parse correctly.

---

Library versions:

pyparsing 2.4.6
pyclibrary 0.1.7

---

Examples:

int fail(unsigned __flags);
int success(unsigned _flags);

struct success_s {
struct {
int aaa;
} _success;
};

struct fail_s {
struct {
int bbb;
} __fail;
} fail_t;

struct success2_s {
int _ccc;
};

struct fail2_s {
int __ddd;
};

---

Parsed results from above example (removed macros and empty sections):

============== types ==================
{ 'struct anon_struct0': Type('struct', 'anon_struct0'),
'struct anon_struct1': Type('struct', 'anon_struct1'),
'struct anon_struct2': Type('struct', 'anon_struct2'),
'struct anon_struct3': Type('struct', 'anon_struct3'),
'struct anon_struct4': Type('struct', 'anon_struct4'),
'struct anon_struct5': Type('struct', 'anon_struct5'),
'struct fail2_s': Type('struct', 'fail2_s'),
'struct fail_s': Type('struct', 'fail_s'),
'struct success2_s': Type('struct', 'success2_s'),
'struct success_s': Type('struct', 'success_s')}
============== variables ==================
{'bbb': (None, Type('int'))}
============== structs ==================
{ 'anon_struct0': Struct(('aaa', Type('int'), None)),
'anon_struct1': Struct(('bbb', Type('int'), None)),
'anon_struct2': Struct(('bbb', Type('int'), None)),
'anon_struct3': Struct(('aaa', Type('int'), None)),
'anon_struct4': Struct(('bbb', Type('int'), None)),
'anon_struct5': Struct(('bbb', Type('int'), None)),
'fail2_s': Struct(),
'fail_s': Struct(),
'success2_s': Struct(('_ccc', Type('int'), None)),
'success_s': Struct(('_success', Type('struct anon_struct3'), None))}
============== functions ==================
{ 'fail': Type(Type('int'), ((None, Type('unsigned', type_quals=(('__',),)), None),)),
'success': Type(Type('int'), (('_flags', Type('unsigned'), None),))}
============== values ==================
{'bbb': None}

---

Failed parsing from above examples:

struct fail_s {
struct {
} __fail;
}
struct fail2_s {
int __ddd;
};

[MatthieuDartiailh]

Looking quickly it appears the parser believe it is dealing with a function qualifier rather than an identifier. Honestly I won't have the bandwidth to look in more details soon. However if you can put together a PR I will do my best to review. Otherwise feel free to ping me again in a couple of weeks it may be better then.

[puredrivel]

I think you foresaw this issue...

# Removes '__name' from all type specs. may cause trouble.
underscore_2_ident = (WordStart(wordchars) + ~keyword + '__' +
                      Word(alphanums, alphanums+"_$") +
                      WordEnd(wordchars)).setParseAction(lambda t: t[0])
type_qualifier = ZeroOrMore((underscore_2_ident + Optional(nestedExpr())) |
                            kwl(qualifiers))

Removing this fixes the issue. I'm assuming it's here for a reason though. If you have a second to comment on the logic behind this, I'll see if I can work on a fix based on it's intended purpose.

[MatthieuDartiailh]

Actually the parser largely predates me taking over the maintenance of the project. However I believe this is here to handle the kind of things discussed in https://stackoverflow.com/questions/1449181/what-does-double-underscore-const-mean-in-c for example. Having read that in what context did you find double underscore name in a header you need to use ?