一、isa->類和元類

上篇文章分析了對象的isa底層實(shí)現(xiàn)以及是如何與cls關(guān)聯(lián)的，這邊文章繼續(xù)分析類的結(jié)構(gòu)。

HPObject *obj = [HPObject alloc];

對obj打斷點(diǎn)查看：

image.png

• x/4gx獲取obj的isa指針。
• isa & mask獲取類對象HPObject。
• po驗(yàn)證獲取到的是HPObject。

這個(gè)時(shí)候如果對類對象繼續(xù)x/4gx呢？

image.png

1.1 類對象內(nèi)存?zhèn)€數(shù)

既然上面驗(yàn)證出來類對象也是會開辟空間的，那么類對象在內(nèi)存中有多少份呢？
驗(yàn)證代碼：

void verifyClassNumber() {
    Class class1 = [HPObject class];
    Class class2 = [HPObject alloc].class;
    Class class3 = object_getClass([HPObject alloc]);
    Class class4 = objc_getClass("HPObject");
    //再次創(chuàng)建對象
    Class class5 = [HPObject alloc].class;
    NSLog(@"\n%p\n%p\n%p\n%p\n%p",class1,class2,class3,class4,class5);
}

結(jié)果：

0x1000082f0
0x1000082f0
0x1000082f0
0x1000082f0
0x1000082f0

可以看到都指向一個(gè)內(nèi)存地址，所以只存在一份。這幾種獲取類對象的方式區(qū)別如下：

獲取類對象對比

上面通過HPObject的isa &mask后也是HPObject，但是兩者的地址不一樣。說明類對象的isa指針獲取的不是類對象。那么它是什么？（元類）。
這個(gè)時(shí)候可以用MachOView查看下符號表：

image.png

二、isa走位圖和繼承鏈

2.1 isa走位圖

上面得到了實(shí)例對象(isa)->類對象(isa)->元類，這個(gè)時(shí)候又有一個(gè)疑問，元類的isa指向哪里呢？

image.png

• 實(shí)例對象(isa)->類對象(sia)->元類(isa)->根元類(NSObject isa)->自身
• 根類實(shí)例對象(isa)->根元類（isa）->自身

代碼驗(yàn)證下:

void verifyIsaLinked() {
    //NSObject 實(shí)例對象
    NSObject *obj = [NSObject alloc];
    //NSObject類
    Class class = object_getClass(obj);
    //NSObject元類
    Class metaClass = object_getClass(class);
    //NSObject根元類
    Class rootMetaClass = object_getClass(metaClass);
    //NSObject根根元類
    Class rootRootMetaClass = object_getClass(rootMetaClass);
    NSLog(@"\n實(shí)例對象:%p \n類:%p \n元類:%p \n根元類:%p \n根根元類:%p",obj,class,metaClass,rootMetaClass,rootRootMetaClass);
}

結(jié)果：

實(shí)例對象:0x10136bf30 
類:0x100358140 
元類:0x1003580f0 
根元類:0x1003580f0 
根根元類:0x1003580f0

這樣就得到了isa走位圖：

isa走位圖

2.2 類和元類的繼承鏈

上面分析到了元類，那么元類有父類么？

HPObject元類的父類

Class hpMetaClass = object_getClass(HPObject.class);
//HPObject元類的父類
Class hpSuperMetaClass = class_getSuperclass(hpMetaClass);
NSLog(@"\nHPObject元類的父類：%@ - %p",hpSuperMetaClass,hpSuperMetaClass);
//NSObject元類
Class objMetaClass = object_getClass(NSObject.class);
NSLog(@"\nNSObject元類：%@ - %p",objMetaClass,objMetaClass);

結(jié)果:

HPObject元類的父類：NSObject - 0x1003580f0
NSObject元類：NSObject - 0x1003580f0

可以得出結(jié)論：HPObject元類的父類是NSObject的元類。

HPObject子類（HPSubobject）元類的父類
新建一個(gè)HPObject的子類HPSubobject同樣獲取它的元類的父類：

///HPObject元類
Class hpMetaClass = object_getClass(HPObject.class);
NSLog(@"\nHPObject元類：%@ - %p",hpMetaClass,hpMetaClass);
//HPSubobject元類
Class hpsMetaClass = object_getClass(HPSubobject.class);
//HPSubobject元類的父類
Class hpsSuperMetaClass = class_getSuperclass(hpsMetaClass);
NSLog(@"\nHPSubobject元類的父類：%@ - %p",hpsSuperMetaClass,hpsSuperMetaClass);

結(jié)果:

HPObject元類：HPObject - 0x1000083b0
HPSubobject元類的父類：HPObject - 0x1000083b0

所以 元類也有繼承鏈。

NSObject（根元類）的父類
那么NSObject的元類也就是根元類的父類呢？

//NSObject 實(shí)例對象
NSObject *obj = [NSObject alloc];
//NSObject類
Class class = object_getClass(obj);
//NSObject元類
Class metaClass = object_getClass(class);
//NSObject元類的父類
Class superMetaClass = class_getSuperclass(metaClass);
NSLog(@"\n類：%@ - %p \n元類的父類：%@ - %p",class,class,superMetaClass,superMetaClass);

結(jié)果：

類：NSObject - 0x100358140 
元類的父類：NSObject - 0x100358140

可以看到是根元類的父類是NSObject，萬物基于NSObject。至此元類的繼承鏈就清晰了。

類的繼承關(guān)系
已知HPSubobject->HPObject->NSObject，需要驗(yàn)證NSObject的父類：

Class objSuperClass = class_getSuperclass(NSObject.class);
NSLog(@"\n%@ - %p",objSuperClass,objSuperClass);

結(jié)果：

(null) - 0x0

所以NSObject不存在父類。

這樣就得到了完整的類和元類的繼承鏈：

類的繼承鏈

通過isa的走位鏈和類的繼承關(guān)系就得到了那張?zhí)O果官網(wǎng)著名的圖：

isa流程圖.jpg

三、源碼分析類結(jié)構(gòu)

3.1類的內(nèi)存結(jié)構(gòu)

既然類也有isa，那么類的結(jié)構(gòu)是怎樣的呢？類在底層是objc_class類型，在runtime.h(OBJC2_UNAVAILABLE)與objc-runtime-new.h中都有聲明?，F(xiàn)在使用的都是objc-runtime-new.h中的objc_class。它是一個(gè)結(jié)構(gòu)體要研究它的結(jié)構(gòu)需要看它的成員變量。
類的結(jié)構(gòu)如下：

struct objc_object {
    Class _Nonnull isa  OBJC_ISA_AVAILABILITY;
};
struct objc_class : objc_object {
    // Class ISA;//繼承
    Class superclass;
    cache_t cache;             // formerly cache pointer and vtable
    class_data_bits_t bits;    // class_rw_t * plus custom rr/alloc flags
};

• ISA：繼承自objc_object，指向元類。
• superclass：指向父類。
• cache：方法緩存，當(dāng)調(diào)用一次方法后就會緩存進(jìn)vtable中，加速下次調(diào)用。
• bits：具體類信息（成員變量、屬性、方法）。

在源碼中能找到class_rw_t中封裝了獲取methods、properties、protocols的方法。而class_rw_t是存在bits中的。那么怎么通過objc_class獲取bits數(shù)據(jù)呢？

3.2 類的結(jié)構(gòu)內(nèi)存計(jì)算

既然類的底層數(shù)據(jù)是結(jié)構(gòu)體，那么只要找到首地址通過偏移就能得到bits數(shù)據(jù)的地址。ISA和superclass都是結(jié)構(gòu)體指針分別占用8字節(jié)，那么cache占多大空間呢？

cache_t中內(nèi)容很多包括很多函數(shù)和static的常量（不占結(jié)構(gòu)體空間），其實(shí)只需要關(guān)注成員變量即可。
cache_t結(jié)構(gòu)如下：

struct cache_t {
private:
    explicit_atomic<uintptr_t> _bucketsAndMaybeMask; //unsigned long 8字節(jié)
    union {
        struct {
            explicit_atomic<mask_t>    _maybeMask;//uint32_t 4字節(jié)
#if __LP64__
            uint16_t                   _flags;//2字節(jié)
#endif
            uint16_t                   _occupied;//2字節(jié)
        };
        explicit_atomic<preopt_cache_t *> _originalPreoptCache;//指針 8字節(jié)
    };
};

cache_t包含兩部分_bucketsAndMaybeMask（unsigned long 8字節(jié)）與聯(lián)合體，聯(lián)合體中有一個(gè)_originalPreoptCache(指針 8字節(jié))與結(jié)構(gòu)體（_originalPreoptCache與結(jié)構(gòu)體只需要計(jì)算一個(gè)，共用一塊內(nèi)存）。所以cache_t大小為16字節(jié)。

那么只需要類的首地址偏移32字節(jié)（0x20 = ISA(8) + superclass (8) + cache (16)）就能得到bits的地址。

指針的步長與指針類型有關(guān)。

3.3 lldb分析類的結(jié)構(gòu)

3.3.1 ISA

image.png

• 類的isa就是一個(gè)純指針，指向元類。

3.3.2 superclass

image.png

• 類的superclass就是類的父類。

3.3.3 cache

cache方法緩存相關(guān)的內(nèi)容cache。

3.3.4 bits

bits的類型是class_data_bits_t結(jié)構(gòu)如下：

struct class_data_bits_t {
    uintptr_t bits;
};

在源碼中有class_rw_t * plus custom rr/alloc flags注釋，也就是說class_data_bits_t的核心是class_rw_t。查找源碼發(fā)現(xiàn)data()返回的是class_rw_t*類型。

data()

data()實(shí)現(xiàn)如下：

#if __LP64__
#define FAST_DATA_MASK        0x00007ffffffffff8UL
#else
#define FAST_DATA_MASK        0xfffffffcUL
#endif


class_rw_t* data() {
   return (class_rw_t *)(bits & FAST_DATA_MASK);
}

• 64位下data()占用44位[3~46]，32位下data()占用30位[2~31]。
• 所以也就是bits(class_data_bits_t)的[3~46]/[2~31]位是data()(class_rw_t)。

同理源碼中有以下代碼：

#if __LP64__

// class is a Swift class from the pre-stable Swift ABI
#define FAST_IS_SWIFT_LEGACY    (1UL<<0)
// class is a Swift class from the stable Swift ABI
#define FAST_IS_SWIFT_STABLE    (1UL<<1)
// class or superclass has default retain/release/autorelease/retainCount/
//   _tryRetain/_isDeallocating/retainWeakReference/allowsWeakReference
#define FAST_HAS_DEFAULT_RR     (1UL<<2)
// data pointer
#define FAST_DATA_MASK          0x00007ffffffffff8UL

#else

// class is a Swift class from the pre-stable Swift ABI
#define FAST_IS_SWIFT_LEGACY  (1UL<<0)
// class is a Swift class from the stable Swift ABI
#define FAST_IS_SWIFT_STABLE  (1UL<<1)
#define FAST_DATA_MASK        0xfffffffcUL

#endif // __LP64__


bool isSwiftStable() {
    return getBit(FAST_IS_SWIFT_STABLE);
}

bool isSwiftLegacy() {
    return getBit(FAST_IS_SWIFT_LEGACY);
}

bool hasCustomRR() const {
    return !bits.getBit(FAST_HAS_DEFAULT_RR);
}

• FAST_IS_SWIFT_LEGACY：第0位類是否來自穩(wěn)定的Swift ABI的 Swift 類。(遺留的類)
• FAST_IS_SWIFT_STABLE：第1位類是否來自穩(wěn)定的Swift ABI的 Swift 類。
• FAST_HAS_DEFAULT_RR：第2位判斷當(dāng)前類或者父類是否含有默認(rèn)的retain/release/autorelease/retainCount/_tryRetain/_isDeallocating/retainWeakReference/allowsWeakReference方法。（僅64位）

前面已經(jīng)得出結(jié)論類的首地址偏移32字節(jié)就能得到bits的地址：

image.png

• 首地址 + 偏移 得到 bits的地址$8。
• 強(qiáng)轉(zhuǎn)bits地址（$8）為class_data_bits_t指針（$9）。
• bits指針($9)調(diào)用data()函數(shù)獲取class_rw_t指針($10)。
• 打印class_rw_t指針?biāo)赶虻闹担?code>*$10）。

class_rw_t

既然data()是class_rw_t結(jié)構(gòu)，它的內(nèi)存結(jié)構(gòu)如下：

struct class_rw_t {
    uint32_t flags;
    uint16_t witness;
#if SUPPORT_INDEXED_ISA
    uint16_t index;
#endif

    explicit_atomic<uintptr_t> ro_or_rw_ext;

    Class firstSubclass;
    Class nextSiblingClass;
};

• flags：
• witness：
• index：32位下有效，記錄類在數(shù)組中的索引。
• ro_or_rw_ext：存儲屬性，方法，協(xié)議，成員變量。
• firstSubclass：第一個(gè)子類。
  在上面的調(diào)試中firstSubclass為nil是因?yàn)樗鼪]有被使用。是懶加載類，如果使用了或者實(shí)現(xiàn)了+ load方法則會指向子類。
• nextSiblingClass：相鄰類。

分析到這里顯然核心就是ro_or_rw_ext了，查看class_rw_t源碼發(fā)現(xiàn)提供了methods、properties、protocols方法。

修改HPObject以及添加方法，文件如下：

image.png

HPObject:

@interface HPObject : NSObject {
    int height;
    NSString *sex;
}
@property (nonatomic, copy) NSString *name;
@property (nonatomic, assign) int age;

- (void)instanceMethod;
+ (void)classMethod;

@end

HPObject+Additions1:

@interface HPObject (Additions1)

- (void)additions1InstanceMethod;

+ (void)additions1ClassMethod;

@end

HPObject+additions2:

@interface HPObject (additions2)

- (void)additions2InstanceMethod;

+ (void)additions2ClassMethod;

@end

??：方法要有對應(yīng)的實(shí)現(xiàn)。

properties()

(lldb) x/6gx HPObject.class
0x1000083d8: 0x00000001000083b0 0x0000000100358140
0x1000083e8: 0x000000010034f360 0x0000803400000000
0x1000083f8: 0x00000001032babd4 0x00000001000ac920
(lldb) p (class_data_bits_t *)0x1000083f8
(class_data_bits_t *) $1 = 0x00000001000083f8
(lldb) p $1->data()
(class_rw_t *) $2 = 0x00000001032babd0
(lldb) p $2->properties()
(const property_array_t) $3 = {
  list_array_tt<property_t, property_list_t, RawPtr> = {
     = {
      list = {
        ptr = 0x0000000100008308
      }
      arrayAndFlag = 4295000840
    }
  }
}

通過properties獲取到的屬性是一個(gè)property_array_t，結(jié)構(gòu)如下：

class property_array_t : 
    public list_array_tt<property_t, property_list_t, RawPtr>
{
    typedef list_array_tt<property_t, property_list_t, RawPtr> Super;

 public:
    property_array_t() : Super() { }
    property_array_t(property_list_t *l) : Super(l) { }
};

property_array_t繼承自list_array_tt是一個(gè)兩層結(jié)構(gòu)property_list_t < property_t>。list_array_tt中有iterator意味著它有遍歷能力。
通過list能獲取到RawPtr<property_list_t>：

(lldb) p $3.list
(const RawPtr<property_list_t>) $4 = {
  ptr = 0x0000000100008308
}

通過訪問ptr能夠獲取到property_list_t數(shù)組:

(lldb) p $4.ptr
(property_list_t *const) $5 = 0x0000000100008308
(lldb) p *$5
(property_list_t) $6 = {
  entsize_list_tt<property_t, property_list_t, 0, PointerModifierNop> = (entsizeAndFlags = 16, count = 2)
}

$5就相當(dāng)于迭代器了,property_list_t源碼結(jié)構(gòu)如下：

struct property_list_t : entsize_list_tt<property_t, property_list_t, 0> {
};

property_list_t是一個(gè)空實(shí)現(xiàn)繼承自entsize_list_tt，entsize_list_tt中有一個(gè)get方法：

struct entsize_list_tt {
    uint32_t entsizeAndFlags;
    uint32_t count;
    Element& get(uint32_t i) const { 
        ASSERT(i < count);
        return getOrEnd(i);
    }
};

所以可以根據(jù)get方法獲取元素：

(lldb) p $6.get(0)
(property_t) $7 = (name = "name", attributes = "T@\"NSString\",C,N,V_name")
(lldb) p $6.get(1)
(property_t) $8 = (name = "age", attributes = "Ti,N,V_age")
(lldb) p $6.get(2)
Assertion failed: (i < count), function get, file /Volumes/HOTPOTCAT/sourcecode/objc4/objc4-818.2/runtime/objc-runtime-new.h, line 625.
error: Execution was interrupted, reason: signal SIGABRT.
The process has been returned to the state before expression evaluation.

這個(gè)時(shí)候并沒有成員變量height和sex。只有name和age兩個(gè)屬性。

methods()

與properties相同，通過class_rw_t的methods方法獲取方法：

(lldb) p $2->methods()
(const method_array_t) $3 = {
  list_array_tt<method_t, method_list_t, method_list_t_authed_ptr> = {
     = {
      list = {
        ptr = 0x0000000100008098
      }
      arrayAndFlag = 4295000216
    }
  }
}
(lldb) p $3.list.ptr
(method_list_t *const) $4 = 0x0000000100008098
(lldb) p *$4
(method_list_t) $5 = {
  entsize_list_tt<method_t, method_list_t, 4294901763, method_t::pointer_modifier> = (entsizeAndFlags = 27, count = 8)
}
(lldb) p $5.get(0)
(method_t) $6 = {}

• 看到有8個(gè)方法。
• 由于method_list_t也是繼承自entsize_list_tt，所以直接通過get()獲取，結(jié)果返回空。

那么分別看下property_t和method_t的實(shí)現(xiàn)。
property_t:

struct property_t {
    const char *name;
    const char *attributes;
};

method_t:

struct method_t {
//......
    struct big {
        SEL name;
        const char *types;
        MethodListIMP imp;
    };
    big &big() const {
        ASSERT(!isSmall());
        return *(struct big *)this;
    }
//......
};

兩者的區(qū)別是property_t有成員變量，method_t沒有成員變量。索引method_t打印為空。但是method_t中有一個(gè)結(jié)構(gòu)體big中有name和imp，并且提供了一個(gè)big方法，所以可以通過big方法獲取：

(lldb) p $5.get(0).big()
(method_t::big) $7 = {
  name = "additions1InstanceMethod"
  types = 0x0000000100003f3f "v16@0:8"
  imp = 0x0000000100003c30 (HPObjcTest`-[HPObject(Additions1) additions1InstanceMethod])
}
(lldb) p $5.get(1).big()
(method_t::big) $8 = {
  name = "instanceMethod"
  types = 0x0000000100003f3f "v16@0:8"
  imp = 0x0000000100003c50 (HPObjcTest`-[HPObject instanceMethod])
}
(lldb) p $5.get(2).big()
(method_t::big) $9 = {
  name = "additions2InstanceMethod"
  types = 0x0000000100003f3f "v16@0:8"
  imp = 0x0000000100003d50 (HPObjcTest`-[HPObject(additions2) additions2InstanceMethod])
}
(lldb) p $5.get(3).big()
(method_t::big) $10 = {
  name = ".cxx_destruct"
  types = 0x0000000100003f3f "v16@0:8"
  imp = 0x0000000100003d00 (HPObjcTest`-[HPObject .cxx_destruct])
}
(lldb) p $5.get(4).big()
(method_t::big) $11 = {
  name = "name"
  types = 0x0000000100003f55 "@16@0:8"
  imp = 0x0000000100003c60 (HPObjcTest`-[HPObject name])
}
(lldb) p $5.get(5).big()
(method_t::big) $12 = {
  name = "setName:"
  types = 0x0000000100003f5d "v24@0:8@16"
  imp = 0x0000000100003c90 (HPObjcTest`-[HPObject setName:])
}
(lldb) p $5.get(6).big()
(method_t::big) $13 = {
  name = "age"
  types = 0x0000000100003f68 "i16@0:8"
  imp = 0x0000000100003cc0 (HPObjcTest`-[HPObject age])
}
(lldb) p $5.get(7).big()
(method_t::big) $14 = {
  name = "setAge:"
  types = 0x0000000100003f70 "v20@0:8i16"
  imp = 0x0000000100003ce0 (HPObjcTest`-[HPObject setAge:])
}

• 除了兩個(gè)屬性的4個(gè)setter + getter方法外還有類和分類的實(shí)例方法以及. cxx_destruct。

protocols()

(lldb)  x/6gx HPObject.class
0x1000088a0: 0x0000000100008878 0x000000010036b140
0x1000088b0: 0x0000000100362360 0x0000803c00000000
0x1000088c0: 0x0000000100637d64 0x00000001000b9980
(lldb) p (class_data_bits_t *)0x1000088c0
(class_data_bits_t *) $16 = 0x00000001000088c0
(lldb) p $16->data()
(class_rw_t *) $17 = 0x0000000100637d60
(lldb) p $17->protocols()
(const protocol_array_t) $18 = {
  list_array_tt<unsigned long, protocol_list_t, RawPtr> = {
     = {
      list = {
        ptr = 0x0000000100008678
      }
      arrayAndFlag = 4295001720
    }
  }
}

protocols()獲取的是protocol_array_t:

class protocol_array_t : 
    public list_array_tt<protocol_ref_t, protocol_list_t, RawPtr>
{
    typedef list_array_tt<protocol_ref_t, protocol_list_t, RawPtr> Super;

 public:
    protocol_array_t() : Super() { }
    protocol_array_t(protocol_list_t *l) : Super(l) { }
};

繼承自list_array_tt與其屬性和方法一致。通過list.ptr能獲取到protocol_list_t：

(lldb) p $18.list.ptr
(protocol_list_t *const) $19 = 0x0000000100008678
(lldb) p *$19
(protocol_list_t) $20 = (count = 1, list = protocol_ref_t [] @ 0x00007fb14e4f68b8)

結(jié)構(gòu)如下：

struct protocol_list_t {
    // count is pointer-sized by accident.
    uintptr_t count;
    protocol_ref_t list[0]; // variable-size
};

它沒有繼承自entsize_list_tt。protocol_ref_t是個(gè)無符號長整形：

typedef uintptr_t protocol_ref_t;  // protocol_t *, but unremapped

嘗試用get函數(shù)獲取元素：

(lldb) p $20.get(0).big()
error: <user expression 22>:1:5: no member named 'get' in 'protocol_list_t'
$20.get(0).big()
~~~ ^
(lldb) p $20.get(0)
error: <user expression 23>:1:5: no member named 'get' in 'protocol_list_t'
$20.get(0)
~~~ ^

果然都失敗了，可以看到protocol_ref_t既然不是個(gè)結(jié)構(gòu)體所以沒有名字相關(guān)的數(shù)據(jù)。沒有繼承自entsize_list_tt所以沒有get函數(shù)。protocol_ref_t的注視看著與protocol_t有關(guān)，
查看下protocol_t的結(jié)構(gòu)：

struct protocol_t : objc_object {
    const char *mangledName;
    struct protocol_list_t *protocols;
    method_list_t *instanceMethods;
    method_list_t *classMethods;
    method_list_t *optionalInstanceMethods;
    method_list_t *optionalClassMethods;
    property_list_t *instanceProperties;
    uint32_t size;   // sizeof(protocol_t)
    uint32_t flags;
    // Fields below this point are not always present on disk.
    const char **_extendedMethodTypes;
    const char *_demangledName;
    property_list_t *_classProperties;
};

現(xiàn)在的問題就是protocol_ref_t怎么轉(zhuǎn)變成protocol_t。
搜索后發(fā)現(xiàn)在remapProtocol中protocol_ref_t直接強(qiáng)轉(zhuǎn)成了protocol_t:

image.png

(lldb) p $20.list[0]
(protocol_ref_t) $21 = 4295002416
(lldb) p (protocol_t *)$21
(protocol_t *) $22 = 0x0000000100008930
(lldb) p *$22
(protocol_t) $23 = {
  objc_object = {
    isa = {
      bits = 4298551496
      cls = Protocol
       = {
        nonpointer = 0
        has_assoc = 0
        has_cxx_dtor = 0
        shiftcls = 537318937
        magic = 0
        weakly_referenced = 0
        unused = 0
        has_sidetable_rc = 0
        extra_rc = 0
      }
    }
  }
  mangledName = 0x0000000100003c13 "HPObjectProtocol"
  protocols = 0x00000001000085c8
  instanceMethods = 0x00000001000085e0
  classMethods = 0x0000000100008600
  optionalInstanceMethods = 0x0000000000000000
  optionalClassMethods = 0x0000000000000000
  instanceProperties = 0x0000000000000000
  size = 96
  flags = 0
  _extendedMethodTypes = 0x0000000100008620
  _demangledName = 0x0000000000000000
  _classProperties = 0x0000000000000000
}

這樣就獲取到了protocol_t中的數(shù)據(jù)。

protocol_t數(shù)據(jù)結(jié)構(gòu)獲取：bits->data()->protocols().list.ptr.list[0]->(protocol_t *)強(qiáng)轉(zhuǎn)。

instanceMethods

(lldb) p $23.instanceMethods
(method_list_t *) $25 = 0x00000001000085e0
(lldb) p *$25
(method_list_t) $26 = {
  entsize_list_tt<method_t, method_list_t, 4294901763, method_t::pointer_modifier> = (entsizeAndFlags = 24, count = 1)
}
(lldb) p $26.get(0).big()
(method_t::big) $27 = {
  name = "protocolInstanceMethod"
  types = 0x0000000100003eb5 "v16@0:8"
  imp = 0x0000000000000000
}

• 實(shí)例方法的獲取與methods相同。

protocols
可以看到protocol_t中有protocols那么它是什么數(shù)據(jù)呢？

(lldb) p $23.protocols
(protocol_list_t *) $24 = 0x00000001000085c8
(lldb) p *$24
(protocol_list_t) $28 = (count = 1, list = protocol_ref_t [] @ 0x00007fb151e12bf8)
(lldb) p  $28.list[0]
(protocol_ref_t) $31 = 4295002320
(lldb) p (protocol_t *)$31
(protocol_t *) $32 = 0x00000001000088d0
(lldb) p *$32
(protocol_t) $33 = {
  objc_object = {
    isa = {
      bits = 0
      cls = nil
       = {
        nonpointer = 0
        has_assoc = 0
        has_cxx_dtor = 0
        shiftcls = 0
        magic = 0
        weakly_referenced = 0
        unused = 0
        has_sidetable_rc = 0
        extra_rc = 0
      }
    }
  }
  mangledName = 0x0000000100003c0a "NSObject"
  protocols = 0x0000000000000000
  instanceMethods = 0x00000001000082f0
  classMethods = 0x0000000000000000
  optionalInstanceMethods = 0x00000001000084c0
  optionalClassMethods = 0x0000000000000000
  instanceProperties = 0x00000001000084e0
  size = 96
  flags = 0
  _extendedMethodTypes = 0x0000000100008528
  _demangledName = 0x0000000000000000
  _classProperties = 0x0000000000000000
}

可以看到是屬于NSObject的。這個(gè)時(shí)候protocols就沒有指向了。instanceMethods還仍然有數(shù)據(jù)：

(lldb) p $33.instanceMethods
(method_list_t *) $34 = 0x00000001000082f0
(lldb) p *$34
(method_list_t) $35 = {
  entsize_list_tt<method_t, method_list_t, 4294901763, method_t::pointer_modifier> = (entsizeAndFlags = 24, count = 19)
}
(lldb) p $35.get(0).big()
(method_t::big) $36 = {
  name = "isEqual:"
  types = 0x0000000100003ebd "c24@0:8@16"
  imp = 0x0000000000000000
}
(lldb) p $35.get(1).big()
(method_t::big) $37 = {
  name = "class"
  types = 0x0000000100003ec8 "#16@0:8"
  imp = 0x0000000000000000
}
(lldb) p $35.get(2).big()
(method_t::big) $38 = {
  name = "self"
  types = 0x0000000100003ed0 "@16@0:8"
  imp = 0x0000000000000000
}

可以看到有19個(gè)方法，都有哪些呢？如下圖：

image.png

classMethods

(lldb) p $57.get(0).big()
(method_t::big) $58 = {
  name = "protocolClassMethod"
  types = 0x0000000100003eb5 "v16@0:8"
  imp = 0x0000000000000000
}

??只要遵循協(xié)議就能關(guān)聯(lián)到，不需要實(shí)現(xiàn)方法。本質(zhì)上與對象通過isa關(guān)聯(lián)cls相同。協(xié)議在底層也繼承自objc_object。也有isa的數(shù)據(jù)結(jié)構(gòu)。也可以添加屬性。

ro()

在properties中并沒有找到實(shí)例變量，但是在class_rw_t的methods()附近發(fā)現(xiàn)了ro()，它返回class_ro_t類型：

struct class_ro_t {
    uint32_t flags;
    uint32_t instanceStart;
    uint32_t instanceSize;
#ifdef __LP64__
    uint32_t reserved;
#endif

    union {
        const uint8_t * ivarLayout;
        Class nonMetaclass;
    };

    explicit_atomic<const char *> name;
    // With ptrauth, this is signed if it points to a small list, but
    // may be unsigned if it points to a big list.
    void *baseMethodList;
    protocol_list_t * baseProtocols;
    const ivar_list_t * ivars;

    const uint8_t * weakIvarLayout;
    property_list_t *baseProperties;
};

發(fā)現(xiàn)class_rw_t中有ivar。

(lldb) p $2->ro()
(const class_ro_t *) $3 = 0x0000000100008238
(lldb) p *$3
(const class_ro_t) $4 = {
  flags = 388
  instanceStart = 8
  instanceSize = 40
  reserved = 0
   = {
    ivarLayout = 0x0000000100003e7a "\x11\x11"
    nonMetaclass = 0x0000000100003e7a
  }
  name = {
    std::__1::atomic<const char *> = "HPObject" {
      Value = 0x0000000100003e71 "HPObject"
    }
  }
  baseMethodList = 0x0000000100008098
  baseProtocols = 0x0000000000000000
  ivars = 0x0000000100008280
  weakIvarLayout = 0x0000000000000000
  baseProperties = 0x0000000100008308
  _swiftMetadataInitializer_NEVER_USE = {}
}
(lldb) p $4.ivars
(const ivar_list_t *const) $5 = 0x0000000100008280
(lldb) p *$5
(const ivar_list_t) $6 = {
  entsize_list_tt<ivar_t, ivar_list_t, 0, PointerModifierNop> = (entsizeAndFlags = 32, count = 4)
}

• ivars結(jié)構(gòu)為ivar_list_t類型，也是繼承自entsize_list_tt，所以同樣應(yīng)該能夠根據(jù)get()去獲取ivar_t。
• 有4個(gè)成員變量。

ivar_t結(jié)構(gòu)如下：

struct ivar_t {
    int32_t *offset;
    const char *name;
    const char *type;
    // alignment is sometimes -1; use alignment() instead
    uint32_t alignment_raw;
    uint32_t size;

    uint32_t alignment() const {
        if (alignment_raw == ~(uint32_t)0) return 1U << WORD_SHIFT;
        return 1 << alignment_raw;
    }
};

獲取實(shí)例變量：

(lldb) p $6.get(0)
(ivar_t) $7 = {
  offset = 0x0000000100008340
  name = 0x0000000100003ec3 "height"
  type = 0x0000000100003f47 "i"
  alignment_raw = 2
  size = 4
}
(lldb) p $6.get(1)
(ivar_t) $8 = {
  offset = 0x0000000100008348
  name = 0x0000000100003eca "sex"
  type = 0x0000000100003f49 "@\"NSString\""
  alignment_raw = 3
  size = 8
}
(lldb) p $6.get(2)
(ivar_t) $9 = {
  offset = 0x0000000100008350
  name = 0x0000000100003ece "_age"
  type = 0x0000000100003f47 "i"
  alignment_raw = 2
  size = 4
}
(lldb) p $6.get(3)
(ivar_t) $10 = {
  offset = 0x0000000100008358
  name = 0x0000000100003ed3 "_name"
  type = 0x0000000100003f49 "@\"NSString\""
  alignment_raw = 3
  size = 8
}

類方法

在methods()中并沒有找到類方法，那么類方法存儲在哪里呢？根據(jù)isa的走位直接進(jìn)去元類查看。元類的結(jié)構(gòu)也是objc_class所以可以嘗試同樣的方式獲取。

(lldb) x/4gx HPObject.class
0x1000083d8: 0x00000001000083b0 0x0000000100358140
0x1000083e8: 0x000000010034f360 0x0000803400000000
(lldb) x/6gx 0x00000001000083b0
0x1000083b0: 0x00000001003580f0 0x00000001003580f0
0x1000083c0: 0x0000000102930780 0x0002e03500000003
0x1000083d0: 0x00000001029303e4 0x00000001000083b0
(lldb) p (class_data_bits_t *)0x1000083d0
(class_data_bits_t *) $1 = 0x00000001000083d0
(lldb) p $1->data()
(class_rw_t *) $2 = 0x00000001029303e0
(lldb) p $2->methods()
(const method_array_t) $3 = {
  list_array_tt<method_t, method_list_t, method_list_t_authed_ptr> = {
     = {
      list = {
        ptr = 0x0000000100008048
      }
      arrayAndFlag = 4295000136
    }
  }
}
(lldb) p $3.list.ptr
(method_list_t *const) $4 = 0x0000000100008048
(lldb) p *$4
(method_list_t) $5 = {
  entsize_list_tt<method_t, method_list_t, 4294901763, method_t::pointer_modifier> = (entsizeAndFlags = 27, count = 3)
}
(lldb) p $5.get(0).big()
(method_t::big) $6 = {
  name = "additions1ClassMethod"
  types = 0x0000000100003f3f "v16@0:8"
  imp = 0x0000000100003c20 (HPObjcTest`+[HPObject(Additions1) additions1ClassMethod])
}
(lldb) p $5.get(1).big()
(method_t::big) $7 = {
  name = "classMethod"
  types = 0x0000000100003f3f "v16@0:8"
  imp = 0x0000000100003c40 (HPObjcTest`+[HPObject classMethod])
}
(lldb) p $5.get(2).big()
(method_t::big) $8 = {
  name = "additions2ClassMethod"
  types = 0x0000000100003f3f "v16@0:8"
  imp = 0x0000000100003d40 (HPObjcTest`+[HPObject(additions2) additions2ClassMethod])
}

• 通過類的isa找到元類。
• 元類的結(jié)構(gòu)也是objc_class，在元類中以同樣的方式獲取bits的data()中methods()就能獲取到類方法了。
• 元類的作用就是存儲類方法。（在底層沒有所謂的類方法，都是對象方法。）

結(jié)論：類和分類的類方法存儲在元類中。

至此，屬性、方法和實(shí)例變量的結(jié)構(gòu)就清晰了，結(jié)構(gòu)如下圖：

objc_class結(jié)構(gòu)

總結(jié)

• 屬性獲取：類->bits(offset 0x20)->data()->properties().list.ptr.get(index)
• 實(shí)例方法獲?。?code>類->bits(offset 0x20)->data()->methods().list.ptr.get(index).big()
• 成員變量獲?。?code>類->bits(offset 0x20)->ro().ivars.get(index)
• 類方法獲?。?類->isa->bits(offset 0x20)->data()->methods().list.ptr.get(index).big()

四、 __has_feature(ptrauth_calls)

在源碼分析中經(jīng)?？吹?code>__has_feature(ptrauth_calls)，那么它究竟有什么作用呢？

• __has_feature：判斷編譯器是否支持某個(gè)功能
• ptrauth_calls：指針身份驗(yàn)證，針對arm64e架構(gòu)；使用Apple A12或更高版本A系列處理器的設(shè)備（iPhoneX以后，不包括X）支持arm64e架構(gòu)。
  也就是iPhone X以上的設(shè)備支持指針驗(yàn)證。

Devices using the Apple A12 or later A-series processor — like the iPhone XS, iPhone XS Max, and iPhone XR — support the arm64e architecture. To test your adoption, you have to run your app on one of these devices. You can’t test using the Simulator.
參考鏈接
可以在Build Settings -> Architectures中配置：

image.png

PAC

這里就引出了一個(gè)問題，什么是PAC？
PAC即Pointer Authentication，它的目的即檢測和保護(hù)地址不被意外或惡意修改，使得應(yīng)用執(zhí)行更加安全。PAC特性是由硬件提供的，保護(hù)了函數(shù)調(diào)用期間，?？臻g和地址的安全。
具體可以參考：arm64e與PAC