前文我们了解到，在android系统中打log，其实就是向几个设备文件中的一个中写入字符串。但我们又要如何来查看这些log呢？是logcat工具，它会读取log，对log做一个格式化，然后呈现给我们。下面我们来看一下logcat的实现。
logcat的用法
先来看一下logcat的用法。下面是 logcat打印出来的help信息：
Usage: logcat [options] [filterspecs]
options include:
-sSet default filter to silent.
Like specifying filterspec '*:s'
-f <filename>Log to file. Default to stdout
-r [<kbytes>]Rotate log every kbytes. (16 if unspecified). Requires -f
-n <count>Sets max number of rotated logs to <count>, default 4
-v <format>Sets the log print format, where <format> is one of:
brief process tag thread raw time threadtime long
-cclear (flush) the entire log and exit
-ddump the log and then exit (don't block)
-t <count>print only the most recent <count> lines (implies -d)
-gget the size of the log's ring buffer and exit
-b <buffer>Request alternate ring buffer, 'main', 'system', 'radio'
or 'events'. Multiple -b parameters are allowed and the
results are interleaved. The default is -b main -b system.
-Boutput the log in binary
filterspecs are a series of
<tag>[:priority]
where <tag> is a log component tag (or * for all) and priority is:
VVerbose
DDebug
IInfo
WWarn
EError
FFatal
SSilent (supress all output)
'*' means '*:d' and <tag> by itself means <tag>:v
If not specified on the commandline, filterspec is set from ANDROID_LOG_TAGS.
If no filterspec is found, filter defaults to '*:I'
If not specified with -v, format is set from ANDROID_PRINTF_LOG
or defaults to "brief"
由上面的help信息，我们可以看到，logcat主要有3个功能，一是清除log设备中的log(-c参数)；二是获取log的ring buffer的大小(-g参数，所谓ring buffer指的也就是某个log设备)；三是打印格式化的log(其他)，这也是logcat最为主要的功能。用于控制格式化的log输出的参数比较多，大致包括，是否为非阻塞输出log(-d)，输出的log的内容(-t)，输出log的buffer(-b，此参数对三种功能都有效)，输出log到文件(-f)，log文件最大大小(-r)，输出log文件的最大个数(-n)，是否为 binary输出(-B)，输出log的格式(-v)，log的 filter(-s)等。控制log输出的这么多参数看得人眼花缭乱，总结一下，打印log，也就是将所选择的log设备中，内容符合特定规则的特定数量的log，以一定的格式写入到特定的最大数量及最大大小有限的文件中去。接下来我们会看一下logcat是如何组织它的code，以实现这些功能的。
logcat工具的相关数据结构
不难想到，logcat要实现这些功能，势必会抽象出来一些数据结构，以方便代码的组织。没错，logcat主要是定义了struct log_device_t来描述一个log设备，也就是log的来源，及AndroidLogFormat，用于对所要输出的log进行过滤。先来看一下 struct log_device_t的定义：
struct queued_entry_t {
union {
unsigned char buf[LOGGER_ENTRY_MAX_LEN + 1] __attribute__((aligned(4)));
struct logger_entry entry __attribute__((aligned(4)));
};
queued_entry_t* next;
queued_entry_t() {
next = NULL;
}
};
struct log_device_t {
char* device;
bool binary;
int fd;
bool printed;
char label;
queued_entry_t* queue;
log_device_t* next;
log_device_t(char* d, bool b, char l) {
device = d;
binary = b;
label = l;
queue = NULL;
next = NULL;
printed = false;
}
void enqueue(queued_entry_t* entry) {
if (this->queue == NULL) {
this->queue = entry;
} else {
queued_entry_t** e = &this->queue;
while (*e && cmp(entry, *e) >= 0) {
e = &((*e)->next);
}
entry->next = *e;
*e = entry;
}
}
};
由此，我们可以看到，log设备的属性主要是其文件描述符，设备名称等。有那个 next指针，我们不难想到系统在组织log设备中的log内容时，是会将各个log设备用一个链表连接起来。而每个log设备都有自己的一个 queued_entry_t类型的链表，用于存放该设备中的log。queued_entry_t的 数据部分为一个char型的buf或struct logger_entry，表示log设备的一个项。上面的LOGGER_ENTRY_MAX_LEN和struct logger_entry定义在system/core/include/cutils/logger.h文件中：
/*
* The userspace structure for version 1 of the logger_entry ABI.
* This structure is returned to userspace by the kernel logger
* driver unless an upgrade to a newer ABI version is requested.
*/
struct logger_entry {
uint16_tlen;/* length of the payload */
uint16_t__pad;/* no matter what, we get 2 bytes of padding */
int32_tpid;/* generating process's pid */
int32_ttid;/* generating process's tid */
int32_tsec;/* seconds since Epoch */
int32_tnsec;/* nanoseconds */
charmsg[0]; /* the entry's payload */
};
/*
* The maximum size of a log entry which can be read from the
* kernel logger driver. An attempt to read less than this amount
* may result in read() returning EINVAL.
*/
#define LOGGER_ENTRY_MAX_LEN (5*1024)
组织log设备及log设备中的内容的结构大致就如上面所述的那样。然后我们再来看一下AndroidLogFormat，它表示用户想要查看的log的格式，其定义为(在system/core/include/cutils/logprint.h中)：
41typedef struct AndroidLogFormat_t AndroidLogFormat;
再来看struct AndroidLogFormat_t的定义(在system/core/liblog/logprint.c中)：
typedef struct FilterInfo_t {
char *mTag;
android_LogPriority mPri;
struct FilterInfo_t *p_next;
} FilterInfo;
struct AndroidLogFormat_t {
android_LogPriority global_pri;
FilterInfo *filters;
AndroidLogPrintFormat format;
};
这个结构描述了用户想要的log的内容及输出log的格式。用户可以指定自己所想要的log，其TAG及其优先级为什么，各个TAG及其优先级用一个FilterInfo来描述，并最终用一个链表连接起来。其中的优先级用android_LogPriority 描述，在system/core /include/android/log.h中定义：
/*
* Android log priority values, in ascending priority order.
*/
typedef enum android_LogPriority {
ANDROID_LOG_UNKNOWN = 0,
ANDROID_LOG_DEFAULT,/* only for SetMinPriority() */
ANDROID_LOG_VERBOSE,
ANDROID_LOG_DEBUG,
ANDROID_LOG_INFO,
ANDROID_LOG_WARN,
ANDROID_LOG_ERROR,
ANDROID_LOG_FATAL,
ANDROID_LOG_SILENT,/* only for SetMinPriority(); must be last */
} android_LogPriority;
而输出log的格式则用 AndroidLogPrintFormat描述，其定义在system/core/include/cutils/logprint.h中：
typedef enum {
FORMAT_OFF = 0,
FORMAT_BRIEF,
FORMAT_PROCESS,
FORMAT_TAG,
FORMAT_THREAD,
FORMAT_RAW,
FORMAT_TIME,
FORMAT_THREADTIME,
FORMAT_LONG,
} AndroidLogPrintFormat;
总结一下， AndroidLogFormat描述了用户想要查看的log及输出log的格式，所想要查看的 log指的是用户可以指定log的TAG及其优先级。
logcat 功能实现
接着我们来看一下logcat的这些功能是如何被一个个实现的。
参数的解析及处理
我们会先从main() 函数的参数解析部分开始，下面是这个部分的code：
int main(int argc, char **argv)
{
int err;
int hasSetLogFormat = 0;
int clearLog = 0;
int getLogSize = 0;
int mode = O_RDONLY;
const char *forceFilters = NULL;
log_device_t* devices = NULL;
log_device_t* dev;
bool needBinary = false;
g_logformat = android_log_format_new();
if (argc == 2 && 0 == strcmp(argv[1], "--test")) {
logprint_run_tests();
exit(0);
}
if (argc == 2 && 0 == strcmp(argv[1], "--help")) {
android::show_help(argv[0]);
exit(0);
}
for (;;) {
int ret;
ret = getopt(argc, argv, "cdt:gsQf:r::n:v:b:B");
if (ret < 0) {
break;
}
switch(ret) {
case 's':
// default to all silent
android_log_addFilterRule(g_logformat, "*:s");
break;
case 'c':
clearLog = 1;
mode = O_WRONLY;
break;
case 'd':
g_nonblock = true;
break;
case 't':
g_nonblock = true;
g_tail_lines = atoi(optarg);
break;
case 'g':
getLogSize = 1;
break;
case 'b': {
char* buf = (char*) malloc(strlen(LOG_FILE_DIR) + strlen(optarg) + 1);
strcpy(buf, LOG_FILE_DIR);
strcat(buf, optarg);
bool binary = strcmp(optarg, "events") == 0;
if (binary) {
needBinary = true;
}
if (devices) {
dev = devices;
while (dev->next) {
dev = dev->next;
}
dev->next = new log_device_t(buf, binary, optarg[0]);
} else {
devices = new log_device_t(buf, binary, optarg[0]);
}
android::g_devCount++;
}
break;
case 'B':
android::g_printBinary = 1;
break;
case 'f':
// redirect output to a file
android::g_outputFileName = optarg;
break;
case 'r':
if (optarg == NULL) {
android::g_logRotateSizeKBytes
= DEFAULT_LOG_ROTATE_SIZE_KBYTES;
} else {
long logRotateSize;
char *lastDigit;
if (!isdigit(optarg[0])) {
fprintf(stderr,"Invalid parameter to -r\n");
android::show_help(argv[0]);
exit(-1);
}
android::g_logRotateSizeKBytes = atoi(optarg);
}
break;
case 'n':
if (!isdigit(optarg[0])) {
fprintf(stderr,"Invalid parameter to -r\n");
android::show_help(argv[0]);
exit(-1);
}
android::g_maxRotatedLogs = atoi(optarg);
break;
case 'v':
err = setLogFormat (optarg);
if (err < 0) {
fprintf(stderr,"Invalid parameter to -v\n");
android::show_help(argv[0]);
exit(-1);
}
hasSetLogFormat = 1;
break;
case 'Q':
/* this is a *hidden* option used to start a version of logcat*/
/* in an emulated device only. it basically looks for androidboot.logcat=*/
/* on the kernel command line. If something is found, it extracts a log filter */
/* and uses it to run the program. If nothing is found, the program should*/
/* quit immediately*/
#defineKERNEL_OPTION"androidboot.logcat="
#defineCONSOLE_OPTION "androidboot.console="
{
intfd;
char*logcat;
char*console;
intforce_exit = 1;
static charcmdline[1024];
fd = open("/proc/cmdline", O_RDONLY);
if (fd >= 0) {
intn = read(fd, cmdline, sizeof(cmdline)-1 );
if (n < 0) n = 0;
cmdline[n] = 0;
close(fd);
} else {
cmdline[0] = 0;
}
logcat= strstr( cmdline, KERNEL_OPTION );
console = strstr( cmdline, CONSOLE_OPTION );
if (logcat != NULL) {
char*p = logcat + sizeof(KERNEL_OPTION)-1;;
char*q = strpbrk( p, " \t\n\r" );;
if (q != NULL)
*q = 0;
forceFilters = p;
force_exit= 0;
}
/* if nothing found or invalid filters, exit quietly */
if (force_exit)
exit(0);
/* redirect our output to the emulator console */
if (console) {
char*p = console + sizeof(CONSOLE_OPTION)-1;
char*q = strpbrk( p, " \t\n\r" );
chardevname[64];
intlen;
if (q != NULL) {
len = q - p;
} else
len = strlen(p);
len = snprintf( devname, sizeof(devname), "/dev/%.*s", len, p );
fprintf(stderr, "logcat using %s (%d)\n", devname, len);
if (len < (int)sizeof(devname)) {
fd = open( devname, O_WRONLY );
if (fd >= 0) {
dup2(fd, 1);
dup2(fd, 2);
close(fd);
}
}
}
}
break;
default:
fprintf(stderr,"Unrecognized Option\n");
android::show_help(argv[0]);
exit(-1);
break;
}
}
if (!devices) {
devices = new log_device_t(strdup("/dev/"LOGGER_LOG_MAIN), false, 'm');
android::g_devCount = 1;
int accessmode =
(mode & O_RDONLY) ? R_OK : 0
| (mode & O_WRONLY) ? W_OK : 0;
// only add this if it's available
if (0 == access("/dev/"LOGGER_LOG_SYSTEM, accessmode)) {
devices->next = new log_device_t(strdup("/dev/"LOGGER_LOG_SYSTEM), false, 's');
android::g_devCount++;
}
}
if (android::g_logRotateSizeKBytes != 0
&& android::g_outputFileName == NULL
) {
fprintf(stderr,"-r requires -f as well\n");
android::show_help(argv[0]);
exit(-1);
}
android::setupOutput();
if (hasSetLogFormat == 0) {
const char* logFormat = getenv("ANDROID_PRINTF_LOG");
if (logFormat != NULL) {
err = setLogFormat(logFormat);
if (err < 0) {
fprintf(stderr, "invalid format in ANDROID_PRINTF_LOG '%s'\n",
logFormat);
}
}
}
if (forceFilters) {
err = android_log_addFilterString(g_logformat, forceFilters);
if (err < 0) {
fprintf (stderr, "Invalid filter expression in -logcat option\n");
exit(0);
}
} else if (argc == optind) {
// Add from environment variable
char *env_tags_orig = getenv("ANDROID_LOG_TAGS");
if (env_tags_orig != NULL) {
err = android_log_addFilterString(g_logformat, env_tags_orig);
if (err < 0) {
fprintf(stderr, "Invalid filter expression in"
" ANDROID_LOG_TAGS\n");
android::show_help(argv[0]);
exit(-1);
}
}
} else {
// Add from commandline
for (int i = optind ; i < argc ; i++) {
err = android_log_addFilterString(g_logformat, argv[i]);
if (err < 0) {
fprintf (stderr, "Invalid filter expression '%s'\n", argv[i]);
android::show_help(argv[0]);
exit(-1);
}
}
}
抛开前面那一堆局部变量的定义不看，main()函数是先调用 android_log_format_new()，创建了一个全局的AndroidLogFormat对象。我们可以看一下 android_log_format_new()的定义 (在文件system/core/liblog/logprint.c中)：
AndroidLogFormat *android_log_format_new()
{
AndroidLogFormat *p_ret;
p_ret = calloc(1, sizeof(AndroidLogFormat));
p_ret->global_pri = ANDROID_LOG_VERBOSE;
p_ret->format = FORMAT_BRIEF;
return p_ret;
}
主要就是为 AndroidLogFormat 对象分配了一块内存，然后初始化了其中的global_pri和format成员。
紧接着，main()函数是处理了两个单独的功能(两个if block)，一个是run test，另一个是show help。不知道Google写logcat的人是怎么考虑的，(1)、其实感觉将调用android_log_format_new()创建AndroidLogFormat对象的过程放在两个if block的后面可能会更好一点，反正两个if block是要exit的，而同时又都没有用到g_logformat对象。此外，(2)、我们搜遍logcat.cpp，是发现对android::show_help()的调用后面就总是紧接着对exit()的调用，其实将exit()放在android::show_help()的定义中也不错，这样就可以减少一些的重复code，可以外加一个exit code参数，并将函数改名为show_help_and_exit()。
接下来，main()函数是借助于 getopt()函数，将各个参数解析出来，然后保存在一些局部变量或全局变量中。可以稍微看一下那些略显复杂的参数的解析。我们将分析-s，-b，-v 这几个参数。
先是-s 参数，它会调用android_log_addFilterRule()函数，给log format添加一个filter，表示用户不想看到任何的log。我们可以看一下android_log_addFilterRule()函数的定义：
static FilterInfo * filterinfo_new(const char * tag, android_LogPriority pri)
{
FilterInfo *p_ret;
p_ret = (FilterInfo *)calloc(1, sizeof(FilterInfo));
p_ret->mTag = strdup(tag);
p_ret->mPri = pri;
return p_ret;
}
/*
* Note: also accepts 0-9 priorities
* returns ANDROID_LOG_UNKNOWN if the character is unrecognized
*/
static android_LogPriority filterCharToPri (char c)
{
android_LogPriority pri;
c = tolower(c);
if (c >= '0' && c <= '9') {
if (c >= ('0'+ANDROID_LOG_SILENT)) {
pri = ANDROID_LOG_VERBOSE;
} else {
pri = (android_LogPriority)(c - '0');
}
} else if (c == 'v') {
pri = ANDROID_LOG_VERBOSE;
} else if (c == 'd') {
pri = ANDROID_LOG_DEBUG;
} else if (c == 'i') {
pri = ANDROID_LOG_INFO;
} else if (c == 'w') {
pri = ANDROID_LOG_WARN;
} else if (c == 'e') {
pri = ANDROID_LOG_ERROR;
} else if (c == 'f') {
pri = ANDROID_LOG_FATAL;
} else if (c == 's') {
pri = ANDROID_LOG_SILENT;
} else if (c == '*') {
pri = ANDROID_LOG_DEFAULT;
} else {
pri = ANDROID_LOG_UNKNOWN;
}
return pri;
}
int android_log_addFilterRule(AndroidLogFormat *p_format,
const char *filterExpression)
{
size_t i=0;
size_t tagNameLength;
android_LogPriority pri = ANDROID_LOG_DEFAULT;
tagNameLength = strcspn(filterExpression, ":");
if (tagNameLength == 0) {
goto error;
}
if(filterExpression[tagNameLength] == ':') {
pri = filterCharToPri(filterExpression[tagNameLength+1]);
if (pri == ANDROID_LOG_UNKNOWN) {
goto error;
}
}
if(0 == strncmp("*", filterExpression, tagNameLength)) {
// This filter expression refers to the global filter
// The default level for this is DEBUG if the priority
// is unspecified
if (pri == ANDROID_LOG_DEFAULT) {
pri = ANDROID_LOG_DEBUG;
}
p_format->global_pri = pri;
} else {
// for filter expressions that don't refer to the global
// filter, the default is verbose if the priority is unspecified
if (pri == ANDROID_LOG_DEFAULT) {
pri = ANDROID_LOG_VERBOSE;
}
char *tagName;
// Presently HAVE_STRNDUP is never defined, so the second case is always taken
// Darwin doesn't have strnup, everything else does
#ifdef HAVE_STRNDUP
tagName = strndup(filterExpression, tagNameLength);
#else
//a few extra bytes copied...
tagName = strdup(filterExpression);
tagName[tagNameLength] = '\0';
#endif /*HAVE_STRNDUP*/
FilterInfo *p_fi = filterinfo_new(tagName, pri);
free(tagName);
p_fi->p_next = p_format->filters;
p_format->filters = p_fi;
}
return 0;
error:
return -1;
}
这个函数所做的事情就是，将filer字串掰开成两部分，一部分是TAG，另外一部分是priority。priority的部分，可以用字符，如"v"、"d" 等来指定，也可以用"2"、"3"等这些数字来指定。而TAG的部分，则当TAG为"*"时，表示要设置全局的priority，否则，需要新建一个 filter info，并插入到p_format对象filter info链表的头部。
然后是 -b参数，它是要添加一个log 设备。由此实现可以看到，如果我们想要指定多个devices，比如events、radio和system，则方法应该为-b events -b radio -b system，而不是-b events radio system。另外，device会被添加到device链表的尾部。
最后是 -v 参数。我们主要来看一下setLogFormat()函数的定义：
static int setLogFormat(const char * formatString)
{
static AndroidLogPrintFormat format;
format = android_log_formatFromString(formatString);
if (format == FORMAT_OFF) {
// FORMAT_OFF means invalid string
return -1;
}
android_log_setPrintFormat(g_logformat, format);
return 0;
}
很简单，将string形式的format做一个转换，然后设给g_logformat。我们也顺便看一下 android_log_formatFromString()和android_log_setPrintFormat()的定义：
void android_log_setPrintFormat(AndroidLogFormat *p_format,
AndroidLogPrintFormat format)
{
p_format->format=format;
}
/**
* Returns FORMAT_OFF on invalid string
*/
AndroidLogPrintFormat android_log_formatFromString(const char * formatString)
{
static AndroidLogPrintFormat format;
if (strcmp(formatString, "brief") == 0) format = FORMAT_BRIEF;
else if (strcmp(formatString, "process") == 0) format = FORMAT_PROCESS;
else if (strcmp(formatString, "tag") == 0) format = FORMAT_TAG;
else if (strcmp(formatString, "thread") == 0) format = FORMAT_THREAD;
else if (strcmp(formatString, "raw") == 0) format = FORMAT_RAW;
else if (strcmp(formatString, "time") == 0) format = FORMAT_TIME;
else if (strcmp(formatString, "threadtime") == 0) format = FORMAT_THREADTIME;
else if (strcmp(formatString, "long") == 0) format = FORMAT_LONG;
else format = FORMAT_OFF;
return format;
}
由这个实现，我们又发现一个可以对logcat做增强的地方。即，在main() 函数里，调用setLogFormat()函数的地方，或者在setLogFormat() 函数中，将formatString参数先做一点点处理，即把所有的字符都转为小写字符，大概会稍稍提升一点logcat的可用性。
参数解析完了之后呢？当然就是要看一下用户所指定的参数是否合适了。首先是devices参数。这个参数是必须的，但用户可能漏设了这个参数，于是要添加几个默认的device，就是"main"和"system"两个device了。然后是-r参数的问题。即"-r"参数是在设置了"-f"参数的情况下才有效，否则应用退出。接着是调用setupOutput()函数，来为输出的文件做进一步的设置。setupOutput()函数定义如下：
static void setupOutput()
{
if (g_outputFileName == NULL) {
g_outFD = STDOUT_FILENO;
} else {
struct stat statbuf;
g_outFD = openLogFile (g_outputFileName);
if (g_outFD < 0) {
perror ("couldn't open output file");
exit(-1);
}
fstat(g_outFD, &statbuf);
g_outByteCount = statbuf.st_size;
}
}
主要就是初始化输出文件的文件描述符。紧接着是log format参数。如果用户没有设置此参数，则logcat将尝试从环境变量中读取，如果还是读不到，那就只好用默认的FORMAT_BRIEF了。最后是filter参数。我们不管跟-Q参数有关的部分。要给logcat指定filters，都是要在参数列表的最后，用TAG:PRI这种形式来指定。如果用户没有指定filters参数，则logcat会尝试从环境变量中读取，如果指定了一个或多个filter参数，则将它们一个个的添加给 g_logformat。如果我们只想看到TAG为某些值的log，则总是需要在最后添加上"*:s" filter，否则，logcat就还是会把所有的log都显示出来。此处似乎有改进的空间，其实可以在发现用户有指定filters时，就先把"*:s"这个filter添加给g_logformat。
-c及-g功能的实现
在 main()函数里：
dev = devices;
while (dev) {
dev->fd = open(dev->device, mode);
if (dev->fd < 0) {
fprintf(stderr, "Unable to open log device '%s': %s\n",
dev->device, strerror(errno));
exit(EXIT_FAILURE);
}
if (clearLog) {
int ret;
ret = android::clearLog(dev->fd);
if (ret) {
perror("ioctl");
exit(EXIT_FAILURE);
}
}
if (getLogSize) {
int size, readable;
size = android::getLogSize(dev->fd);
if (size < 0) {
perror("ioctl");
exit(EXIT_FAILURE);
}
readable = android::getLogReadableSize(dev->fd);
if (readable < 0) {
perror("ioctl");
exit(EXIT_FAILURE);
}
printf("%s: ring buffer is %dKb (%dKb consumed), "
"max entry is %db, max payload is %db\n", dev->device,
size / 1024, readable / 1024,
(int) LOGGER_ENTRY_MAX_LEN, (int) LOGGER_ENTRY_MAX_PAYLOAD);
}
dev = dev->next;
}
if (getLogSize) {
exit(0);
}
if (clearLog) {
exit(0);
}
再看那几个辅助函数：
static int clearLog(int logfd)
{
return ioctl(logfd, LOGGER_FLUSH_LOG);
}
/* returns the total size of the log's ring buffer */
static int getLogSize(int logfd)
{
return ioctl(logfd, LOGGER_GET_LOG_BUF_SIZE);
}
/* returns the readable size of the log's ring buffer (that is, amount of the log consumed) */
static int getLogReadableSize(int logfd)
{
return ioctl(logfd, LOGGER_GET_LOG_LEN);
}
总的来说，就是利用logger驱动提供的ioctl功能，将一个log设备中的log内容清空，或者获取log设备的一些信息。
读取日志文件
在 main()函数里：
//LOG_EVENT_INT(10, 12345);
//LOG_EVENT_LONG(11, 0x1122334455667788LL);
//LOG_EVENT_STRING(0, "whassup, doc?");
if (needBinary)
android::g_eventTagMap = android_openEventTagMap(EVENT_TAG_MAP_FILE);
android::readLogLines(devices);
return 0;
}
主要就是调用readLogLines()了。该函数定义为：
static void readLogLines(log_device_t* devices)
{
log_device_t* dev;
int max = 0;
int ret;
int queued_lines = 0;
bool sleep = false;
int result;
fd_set readset;
for (dev=devices; dev; dev = dev->next) {
if (dev->fd > max) {
max = dev->fd;
}
}
while (1) {
do {
timeval timeout = { 0, 5000 /* 5ms */ }; // If we oversleep it's ok, i.e. ignore EINTR.
FD_ZERO(&readset);
for (dev=devices; dev; dev = dev->next) {
FD_SET(dev->fd, &readset);
}
result = select(max + 1, &readset, NULL, NULL, sleep ? NULL : &timeout);
} while (result == -1 && errno == EINTR);
if (result >= 0) {
for (dev=devices; dev; dev = dev->next) {
if (FD_ISSET(dev->fd, &readset)) {
queued_entry_t* entry = new queued_entry_t();
/* NOTE: driver guarantees we read exactly one full entry */
ret = read(dev->fd, entry->buf, LOGGER_ENTRY_MAX_LEN);
if (ret < 0) {
if (errno == EINTR) {
delete entry;
goto next;
}
if (errno == EAGAIN) {
delete entry;
break;
}
perror("logcat read");
exit(EXIT_FAILURE);
}
else if (!ret) {
fprintf(stderr, "read: Unexpected EOF!\n");
exit(EXIT_FAILURE);
}
else if (entry->entry.len != ret - sizeof(struct logger_entry)) {
fprintf(stderr, "read: unexpected length. Expected %d, got %d\n",
entry->entry.len, ret - sizeof(struct logger_entry));
exit(EXIT_FAILURE);
}
entry->entry.msg[entry->entry.len] = '\0';
dev->enqueue(entry);
++queued_lines;
}
}
if (result == 0) {
// we did our short timeout trick and there's nothing new
// print everything we have and wait for more data
sleep = true;
while (true) {
chooseFirst(devices, &dev);
if (dev == NULL) {
break;
}
if (g_tail_lines == 0 || queued_lines <= g_tail_lines) {
printNextEntry(dev);
} else {
skipNextEntry(dev);
}
--queued_lines;
}
// the caller requested to just dump the log and exit
if (g_nonblock) {
return;
}
} else {
// print all that aren't the last in their list
sleep = false;
while (g_tail_lines == 0 || queued_lines > g_tail_lines) {
chooseFirst(devices, &dev);
if (dev == NULL || dev->queue->next == NULL) {
break;
}
if (g_tail_lines == 0) {
printNextEntry(dev);
} else {
skipNextEntry(dev);
}
--queued_lines;
}
}
}
next:
;
}
}
由于可能要同时输出多个log设备文件中的log，这里使用了select机制来监控这些log设备文件中的哪个设备可读：
do {
timeval timeout = { 0, 5000 /* 5ms */ }; // If we oversleep it's ok, i.e. ignore EINTR.
FD_ZERO(&readset);
for (dev=devices; dev; dev = dev->next) {
FD_SET(dev->fd, &readset);
}
result = select(max + 1, &readset, NULL, NULL, sleep ? NULL : &timeout);
} while (result == -1 && errno == EINTR);
如果result>=0，则表示有 log设备文件可读或者超时。接着就用一个for循环检查哪个设备可读，即FD_ISSET(dev->fd, &readset)是否为true，如果为true，则表明该log设备可读，则需要进一步将该log设备中的日志读出。对于select系统调用而言，readset既是一个传入参数，又是一个传出参数，作为传入参数时，它表示需要监听哪些文件描述符上的事件，做为传出参数时，则表示 select()返回，是由于哪些文件上的事件。这也即是在for循环的开始处，每次都要重新构造readset的原因。为何此处是用的select机制，而没有使用opoll机制，此处是否有可能使用 epoll机制呢？留待后续分析。
logcat在读取log时，是轮流读取每一个可以读取的log设备，每个设备中读取一条。此处的一条log的含义，有必要通过实验或从log设备驱动的角度来进一步厘清，一个struct queued_entry_t竟然都要占用5KB的内存空间呢。这个部分对应于如下的code：
if (result >= 0) {
for (dev=devices; dev; dev = dev->next) {
if (FD_ISSET(dev->fd, &readset)) {
queued_entry_t* entry = new queued_entry_t();
/* NOTE: driver guarantees we read exactly one full entry */
ret = read(dev->fd, entry->buf, LOGGER_ENTRY_MAX_LEN);
if (ret < 0) {
if (errno == EINTR) {
delete entry;
goto next;
}
if (errno == EAGAIN) {
delete entry;
break;
}
perror("logcat read");
exit(EXIT_FAILURE);
}
else if (!ret) {
fprintf(stderr, "read: Unexpected EOF!\n");
exit(EXIT_FAILURE);
}
else if (entry->entry.len != ret - sizeof(struct logger_entry)) {
fprintf(stderr, "read: unexpected length. Expected %d, got %d\n",
entry->entry.len, ret - sizeof(struct logger_entry));
exit(EXIT_FAILURE);
}
entry->entry.msg[entry->entry.len] = '\0';
dev->enqueue(entry);
++queued_lines;
}
}
调用read函数之前，先创建一个 queued_entry_t的entry，接着调用read函数将log读到entry->buf中，最后调用 dev->enqueue(entry)将log记录加入到对应的log设备的log项队例中去。回想我们前面看到的struct log_device_t的定义中的enqueue()函数，在插入一条log之后，它总是会保证log设备的log队列中的各log是有序的，有 cmp()函数来决定各条log的前后。我们来看cmp()函数的定义：
static int cmp(queued_entry_t* a, queued_entry_t* b) {
int n = a->entry.sec - b->entry.sec;
if (n != 0) {
return n;
}
return a->entry.nsec - b->entry.nsec;
}
即log队列中各条log将以时间顺序由早到晚的排列。将一条log插入队列的同时，还会把当前的log记录数保存在 queued_lines变量中。
继续进一步处理log：
if (result == 0) {
// we did our short timeout trick and there's nothing new
// print everything we have and wait for more data
sleep = true;
while (true) {
chooseFirst(devices, &dev);
if (dev == NULL) {
break;
}
if (g_tail_lines == 0 || queued_lines <= g_tail_lines) {
printNextEntry(dev);
} else {
skipNextEntry(dev);
}
--queued_lines;
}
// the caller requested to just dump the log and exit
if (g_nonblock) {
return;
}
} else {
// print all that aren't the last in their list
sleep = false;
while (g_tail_lines == 0 || queued_lines > g_tail_lines) {
chooseFirst(devices, &dev);
if (dev == NULL || dev->queue->next == NULL) {
break;
}
if (g_tail_lines == 0) {
printNextEntry(dev);
} else {
skipNextEntry(dev);
}
--queued_lines;
}
}
此处有一点让人弄不明白，result>0才意味着log设备中有内容需要读龋为何此处没有把读取log设备文件的代码直接放在处理result>0这一case的block中，而要单独拎出来呢？留待后续分析。接着看此处大代码。
如果result == 0，就表明select的返回是由于超时，目前没有新的log可以读取，这时就要先处理之前已经读取的log。会用一个while循环来输出所需数量的log。在while 循环中，调用 chooseFirst()函数选择log队列不为空，且log队列中第一条日志记录时间戳最小的设备，即先输出最旧的日志：
static void chooseFirst(log_device_t* dev, log_device_t** firstdev) {
for (*firstdev = NULL; dev != NULL; dev = dev->next) {
if (dev->queue != NULL && (*firstdev == NULL || cmp(dev->queue, (*firstdev)->queue) < 0)) {
*firstdev = dev;
}
}
}
如果存在这样的设备，则依据于 g_tail_lines(-t参数可以指定)的值，初步判断一条log是应该要输出还是要丢弃：
if (g_tail_lines == 0 || queued_lines <= g_tail_lines) {
printNextEntry(dev);
} else {
skipNextEntry(dev);
}
g_tail_lines表示显示最新 log记录的条数，如果为0，就表示全部显示。如果g_tail_lines == 0或者queued_lines <= g_tail_lines，就表示这条log记录应该输出，否则就要丢弃。每处理完一条日志记录，queued_lines就减1，这样，最新的 g_tail_lines就可以输出出来了。
如果result > 0，就表示select是由于有设备文件可读而返回。分两种情况，一是用户指定了要显示的最新log的数量，则会检查已经读取的log条数是否已经超越了用户想要的log数量，如果是，则要丢弃最老的那些log记录；二是没有指定要显示的最新log的数量的情况，即都要显示，则会输出前面读取的所有的log。可以看一下这段代码：
while (g_tail_lines == 0 || queued_lines > g_tail_lines) {
chooseFirst(devices, &dev);
if (dev == NULL || dev->queue->next == NULL) {
break;
}
if (g_tail_lines == 0) {
printNextEntry(dev);
} else {
skipNextEntry(dev);
}
--queued_lines;
}
丢弃一条日志，用 skipNextEntry()函数，我们可以看一下它的定义：
static void maybePrintStart(log_device_t* dev) {
if (!dev->printed) {
dev->printed = true;
if (g_devCount > 1 && !g_printBinary) {
char buf[1024];
snprintf(buf, sizeof(buf), "--------- beginning of %s\n", dev->device);
if (write(g_outFD, buf, strlen(buf)) < 0) {
perror("output error");
exit(-1);
}
}
}
}
static void skipNextEntry(log_device_t* dev) {
maybePrintStart(dev);
queued_entry_t* entry = dev->queue;
dev->queue = entry->next;
delete entry;
}
在首次被调用同时之前没有输出国任何 log时，会先输出一行提示信息。然后就是从log设备log队列的头部取下一个log记录，并将其空间释放掉。后面会再来看输出一条log记录的 printNextEntry()函数。
输出一个设备文件中的一条log记录
从前面的分析中看出，最终log设备文件内容的输出是通过printNextEntry()函数进行的，这个函数的定义如下：
static void printNextEntry(log_device_t* dev) {
maybePrintStart(dev);
if (g_printBinary) {
printBinary(&dev->queue->entry);
} else {
processBuffer(dev, &dev->queue->entry);
}
skipNextEntry(dev);
}
maybePrintStart(dev)，这个我们前面看过了，就是用来打印我们每次用logcat打log都会看到的最开始的那几行的。然后就是分两种情况来打印log。当 g_printBinary为true时，以二进制方式输出log到指定的文件中：
void printBinary(struct logger_entry *buf)
{
size_t size = sizeof(logger_entry) + buf->len;
int ret;
do {
ret = write(g_outFD, buf, size);
} while (ret < 0 && errno == EINTR);
}
所谓的二进制格式，主要是相对于纯文本格式而言的，指的是，log驱动在存储log时是有一定的结构的，也就是struct logger_entry的结构，log驱动返回给logcat的也将是这种结构，这种结构并不是很适合直接给用户来查看。而二进制格式的输出则指，将会不做任何修饰及转换地将struct logger_entry写入文件中。
我们再来看g_printBinary为false的情况。此时，logcat在输出log之前，会先对一个log记录做一个到文本形式的格式化。其处理过程都在processBuffer()函数中：
static void processBuffer(log_device_t* dev, struct logger_entry *buf)
{
int bytesWritten = 0;
int err;
AndroidLogEntry entry;
char binaryMsgBuf[1024];
if (dev->binary) {
err = android_log_processBinaryLogBuffer(buf, &entry, g_eventTagMap,
binaryMsgBuf, sizeof(binaryMsgBuf));
//printf(">>> pri=%d len=%d msg='%s'\n",
//entry.priority, entry.messageLen, entry.message);
} else {
err = android_log_processLogBuffer(buf, &entry);
}
if (err < 0) {
goto error;
}
if (android_log_shouldPrintLine(g_logformat, entry.tag, entry.priority)) {
if (false && g_devCount > 1) {
binaryMsgBuf[0] = dev->label;
binaryMsgBuf[1] = ' ';
bytesWritten = write(g_outFD, binaryMsgBuf, 2);
if (bytesWritten < 0) {
perror("output error");
exit(-1);
}
}
bytesWritten = android_log_printLogLine(g_logformat, g_outFD, &entry);
if (bytesWritten < 0) {
perror("output error");
exit(-1);
}
}
g_outByteCount += bytesWritten;
if (g_logRotateSizeKBytes > 0
&& (g_outByteCount / 1024) >= g_logRotateSizeKBytes
) {
rotateLogs();
}
error:
//fprintf (stderr, "Error processing record\n");
return;
}
整体的流程大致为：1、依据一个log 设备是否是binary的，将一条log做进一步的处理，结果存放在一个AndroidLogEntry entry中；2、依据于用户添加的那些filters，判断一条log是否是用户所想要的，如果是，则其写入文件中；3、更新g_outByteCount，使其依然能够正确的表示已经写入文件的log的字节数；4、检查已经写入文件的log字节数，如果该值超出了用户设定的单个log文件最大大小时，则调用 rotateLogs()换一个log文件来写。
对于前述第1步，设备的binary不同于用户指定的全局的二进制格式的输出，该属性只与设备本身有关。回忆我们前面在参数解析部分(-b参数)看到的，只有"events" log设备是binary的。此外，我们再来看一下，所谓的对log的进一步处理，其结果究竟是什么样的，我们看一下AndroidLogEntry结构的定义(在system/core/include/cutils/logprint.h文件中)：
typedef struct AndroidLogEntry_t {
time_t tv_sec;
long tv_nsec;
android_LogPriority priority;
int32_t pid;
int32_t tid;
const char * tag;
size_t messageLen;
const char * message;
} AndroidLogEntry;
然后呢，我们暂时先抛开 binary设备的情况不管，先来看一下一般的device，其一条log的处理过程，即android_log_processLogBuffer() 函数(在system/core/liblog/logprint.c中)：
/**
* Splits a wire-format buffer into an AndroidLogEntry
* entry allocated by caller. Pointers will point directly into buf
*
* Returns 0 on success and -1 on invalid wire format (entry will be
* in unspecified state)
*/
int android_log_processLogBuffer(struct logger_entry *buf,
AndroidLogEntry *entry)
{
entry->tv_sec = buf->sec;
entry->tv_nsec = buf->nsec;
entry->pid = buf->pid;
entry->tid = buf->tid;
/*
* format: <priority:1><tag:N>\0<message:N>\0
*
* tag str
*starts at buf->msg+1
* msg
*starts at buf->msg+1+len(tag)+1
*
* The message may have been truncated by the kernel log driver.
* When that happens, we must null-terminate the message ourselves.
*/
if (buf->len < 3) {
// An well-formed entry must consist of at least a priority
// and two null characters
fprintf(stderr, "+++ LOG: entry too small\n");
return -1;
}
int msgStart = -1;
int msgEnd = -1;
int i;
for (i = 1; i < buf->len; i++) {
if (buf->msg[i] == '\0') {
if (msgStart == -1) {
msgStart = i + 1;
} else {
msgEnd = i;
break;
}
}
}
if (msgStart == -1) {
fprintf(stderr, "+++ LOG: malformed log message\n");
return -1;
}
if (msgEnd == -1) {
// incoming message not null-terminated; force it
msgEnd = buf->len - 1;
buf->msg[msgEnd] = '\0';
}
entry->priority = buf->msg[0];
entry->tag = buf->msg + 1;
entry->message = buf->msg + msgStart;
entry->messageLen = msgEnd - msgStart;
return 0;
}
用户写入log设备的部分，在 logcat读取时，全都会被放在struct logger_entry结构的msg成员中，在android_log_processLogBuffer()函数中，则会再次将一条log的 priority、TAG及message三个部分解析出来存放在AndroidLogEntry的单独成员里。其他倒还比较直观。
processBuffer()中处理log的第二步，即，依据于用户添加的那些filters，判断一条log是否是用户所想要的，如果是，则将其写入文件中。是通过android_log_shouldPrintLine()函数(在 system/core/liblog/logprint.c中)来做判断的：
static android_LogPriority filterPriForTag(
AndroidLogFormat *p_format, const char *tag)
{
FilterInfo *p_curFilter;
for (p_curFilter = p_format->filters
; p_curFilter != NULL
; p_curFilter = p_curFilter->p_next
) {
if (0 == strcmp(tag, p_curFilter->mTag)) {
if (p_curFilter->mPri == ANDROID_LOG_DEFAULT) {
return p_format->global_pri;
} else {
return p_curFilter->mPri;
}
}
}
return p_format->global_pri;
}
int android_log_shouldPrintLine (
AndroidLogFormat *p_format, const char *tag, android_LogPriority pri)
{
return pri >= filterPriForTag(p_format, tag);
}
当读出的一条log，其priority高于用户添加的TAG相同的filter的priority时，即认为是用户需要的。在发现一条log需要写入文件时，则会在要打印的log设备大于1时，先将log设备的label写入文件。然后则是调用 android_log_printLogLine()函数，将AndroidLogEntry的entry格式化并写入文件：
static char filterPriToChar (android_LogPriority pri)
{
switch (pri) {
case ANDROID_LOG_VERBOSE:return 'V';
case ANDROID_LOG_DEBUG:return 'D';
case ANDROID_LOG_INFO:return 'I';
case ANDROID_LOG_WARN:return 'W';
case ANDROID_LOG_ERROR:return 'E';
case ANDROID_LOG_FATAL:return 'F';
case ANDROID_LOG_SILENT:return 'S';
case ANDROID_LOG_DEFAULT:
case ANDROID_LOG_UNKNOWN:
default:return '?';
}
}
/**
* Formats a log message into a buffer
*
* Uses defaultBuffer if it can, otherwise malloc()'s a new buffer
* If return value != defaultBuffer, caller must call free()
* Returns NULL on malloc error
*/
char *android_log_formatLogLine (
AndroidLogFormat *p_format,
char *defaultBuffer,
size_t defaultBufferSize,
const AndroidLogEntry *entry,
size_t *p_outLength)
{
#if defined(HAVE_LOCALTIME_R)
struct tm tmBuf;
#endif
struct tm* ptm;
char timeBuf[32];
char headerBuf[128];
char prefixBuf[128], suffixBuf[128];
char priChar;
int prefixSuffixIsHeaderFooter = 0;
char * ret = NULL;
priChar = filterPriToChar(entry->priority);
/*
* Get the current date/time in pretty form
*
* It's often useful when examining a log with "less" to jump to
* a specific point in the file by searching for the date/time stamp.
* For this reason it's very annoying to have regexp meta characters
* in the time stamp.Don't use forward slashes, parenthesis,
* brackets, asterisks, or other special chars here.
*/
#if defined(HAVE_LOCALTIME_R)
ptm = localtime_r(&(entry->tv_sec), &tmBuf);
#else
ptm = localtime(&(entry->tv_sec));
#endif
//strftime(timeBuf, sizeof(timeBuf), "%Y-%m-%d %H:%M:%S", ptm);
strftime(timeBuf, sizeof(timeBuf), "%m-%d %H:%M:%S", ptm);
/*
* Construct a buffer containing the log header and log message.
*/
size_t prefixLen, suffixLen;
switch (p_format->format) {
case FORMAT_TAG:
prefixLen = snprintf(prefixBuf, sizeof(prefixBuf),
"%c/%-8s: ", priChar, entry->tag);
strcpy(suffixBuf, "\n"); suffixLen = 1;
break;
case FORMAT_PROCESS:
prefixLen = snprintf(prefixBuf, sizeof(prefixBuf),
"%c(%5d) ", priChar, entry->pid);
suffixLen = snprintf(suffixBuf, sizeof(suffixBuf),
"(%s)\n", entry->tag);
break;
case FORMAT_THREAD:
prefixLen = snprintf(prefixBuf, sizeof(prefixBuf),
"%c(%5d:%5d) ", priChar, entry->pid, entry->tid);
strcpy(suffixBuf, "\n");
suffixLen = 1;
break;
case FORMAT_RAW:
prefixBuf[0] = 0;
prefixLen = 0;
strcpy(suffixBuf, "\n");
suffixLen = 1;
break;
case FORMAT_TIME:
prefixLen = snprintf(prefixBuf, sizeof(prefixBuf),
"%s.%03ld %c/%-8s(%5d): ", timeBuf, entry->tv_nsec / 1000000,
priChar, entry->tag, entry->pid);
strcpy(suffixBuf, "\n");
suffixLen = 1;
break;
case FORMAT_THREADTIME:
prefixLen = snprintf(prefixBuf, sizeof(prefixBuf),
"%s.%03ld %5d %5d %c %-8s: ", timeBuf, entry->tv_nsec / 1000000,
entry->pid, entry->tid, priChar, entry->tag);
strcpy(suffixBuf, "\n");
suffixLen = 1;
break;
case FORMAT_LONG:
prefixLen = snprintf(prefixBuf, sizeof(prefixBuf),
"[ %s.%03ld %5d:%5d %c/%-8s ]\n",
timeBuf, entry->tv_nsec / 1000000, entry->pid,
entry->tid, priChar, entry->tag);
strcpy(suffixBuf, "\n\n");
suffixLen = 2;
prefixSuffixIsHeaderFooter = 1;
break;
case FORMAT_BRIEF:
default:
prefixLen = snprintf(prefixBuf, sizeof(prefixBuf),
"%c/%-8s(%5d): ", priChar, entry->tag, entry->pid);
strcpy(suffixBuf, "\n");
suffixLen = 1;
break;
}
/* snprintf has a weird return value.It returns what would have been
* written given a large enough buffer.In the case that the prefix is
* longer then our buffer(128), it messes up the calculations below
* possibly causing heap corruption.To avoid this we double check and
* set the length at the maximum (size minus null byte)
*/
if(prefixLen >= sizeof(prefixBuf))
prefixLen = sizeof(prefixBuf) - 1;
if(suffixLen >= sizeof(suffixBuf))
suffixLen = sizeof(suffixBuf) - 1;
/* the following code is tragically unreadable */
size_t numLines;
size_t i;
char *p;
size_t bufferSize;
const char *pm;
if (prefixSuffixIsHeaderFooter) {
// we're just wrapping message with a header/footer
numLines = 1;
} else {
pm = entry->message;
numLines = 0;
// The line-end finding here must match the line-end finding
// in for ( ... numLines...) loop below
while (pm < (entry->message + entry->messageLen)) {
if (*pm++ == '\n') numLines++;
}
// plus one line for anything not newline-terminated at the end
if (pm > entry->message && *(pm-1) != '\n') numLines++;
}
// this is an upper bound--newlines in message may be counted
// extraneously
bufferSize = (numLines * (prefixLen + suffixLen)) + entry->messageLen + 1;
if (defaultBufferSize >= bufferSize) {
ret = defaultBuffer;
} else {
ret = (char *)malloc(bufferSize);
if (ret == NULL) {
return ret;
}
}
ret[0] = '\0';/* to start strcat off */
p = ret;
pm = entry->message;
if (prefixSuffixIsHeaderFooter) {
strcat(p, prefixBuf);
p += prefixLen;
strncat(p, entry->message, entry->messageLen);
p += entry->messageLen;
strcat(p, suffixBuf);
p += suffixLen;
} else {
while(pm < (entry->message + entry->messageLen)) {
const char *lineStart;
size_t lineLen;
lineStart = pm;
// Find the next end-of-line in message
while (pm < (entry->message + entry->messageLen)
&& *pm != '\n') pm++;
lineLen = pm - lineStart;
strcat(p, prefixBuf);
p += prefixLen;
strncat(p, lineStart, lineLen);
p += lineLen;
strcat(p, suffixBuf);
p += suffixLen;
if (*pm == '\n') pm++;
}
}
if (p_outLength != NULL) {
*p_outLength = p - ret;
}
return ret;
}
/**
* Either print or do not print log line, based on filter
*
* Returns count bytes written
*/
int android_log_printLogLine(
AndroidLogFormat *p_format,
int fd,
const AndroidLogEntry *entry)
{
int ret;
char defaultBuffer[512];
char *outBuffer = NULL;
size_t totalLen;
outBuffer = android_log_formatLogLine(p_format, defaultBuffer,
sizeof(defaultBuffer), entry, &totalLen);
if (!outBuffer)
return -1;
do {
ret = write(fd, outBuffer, totalLen);
} while (ret < 0 && errno == EINTR);
if (ret < 0) {
fprintf(stderr, "+++ LOG: write failed (errno=%d)\n", errno);
ret = 0;
goto done;
}
if (((size_t)ret) < totalLen) {
fprintf(stderr, "+++ LOG: write partial (%d of %d)\n", ret,
(int)totalLen);
goto done;
}
done:
if (outBuffer != defaultBuffer) {
free(outBuffer);
}
return ret;
}
格式化log的部分不再细述。
processBuffer 函数的最后，还有一个rotateLogs的操作：
static int openLogFile (const char *pathname)
{
return open(pathname, O_WRONLY | O_APPEND | O_CREAT, S_IRUSR | S_IWUSR);
}
static void rotateLogs()
{
int err;
// Can't rotate logs if we're not outputting to a file
if (g_outputFileName == NULL) {
return;
}
close(g_outFD);
for (int i = g_maxRotatedLogs ; i > 0 ; i--) {
char *file0, *file1;
asprintf(&file1, "%s.%d", g_outputFileName, i);
if (i - 1 == 0) {
asprintf(&file0, "%s", g_outputFileName);
} else {
asprintf(&file0, "%s.%d", g_outputFileName, i - 1);
}
err = rename (file0, file1);
if (err < 0 && errno != ENOENT) {
perror("while rotating log files");
}
free(file1);
free(file0);
}
g_outFD = openLogFile (g_outputFileName);
if (g_outFD < 0) {
perror ("couldn't open output file");
exit(-1);
}
g_outByteCount = 0;
}
这个函数只有在执行logcat命令，指定了-f <filename>参数，即g_outputFileName不为NULL时才起作用。它的作用是将log循环输出到一组文件中。例如，指定-f参数为logfile，g_maxRotatedLogs为3，则这组文件分别为：logfile，logfile.1，logfile.2，logfile.3。
当当前输入到logfile文件中的log大小g_outByteCount大于等于g_logRotateSizeKBytes时，就要将 logfile.2的内容移至logfile.3中，同时将logfile.1的内容移至logfile.2中，同时logfle的内容移至 logfile.1中，再重新打开logfile文件进入后续输入，而原来的logfile.3文件则相当于被移除掉了。这样做的作用是不至于使得日志文件变得越来越来大，以至于占用过多的磁盘空间，而是只在磁盘上保存一定量的最新的日志记录。这样，旧的日志记录就会可能被新的日志记录所覆盖。
遗留的问题，binary device，即/dev/log/events，其log记录的格式化。
Done。
