Compute SHA-1 checksums of regular files to be stored with index data.

[cumulus.git] / scandir.cc

diff --git a/scandir.cc b/scandir.cc
index 64aec8a..81ddd4b 100644 (file)
--- a/scandir.cc
+++ b/scandir.cc
@@ -4,6 +4,7 @@
 #include <stdlib.h>
 #include <stdint.h>
 #include <dirent.h>
+#include <errno.h>
 #include <fcntl.h>
 #include <sys/types.h>
 #include <sys/stat.h>
@@ -13,20 +14,53 @@
 #include <string>
 #include <vector>
 
+#include "store.h"
+#include "sha1.h"
+
 using std::string;
 using std::vector;
 
+static OutputStream *info_dump = NULL;
+
 void scandir(const string& path);
 
-void dumpfile(int fd)
+/* Converts time to microseconds since the epoch. */
+int64_t encode_time(time_t time)
+{
+    return (int64_t)time * 1000000;
+}
+
+void dumpfile(int fd, dictionary &file_info)
 {
     struct stat stat_buf;
     fstat(fd, &stat_buf);
+    int64_t size = 0;
+
+    char buf[4096];
 
     if ((stat_buf.st_mode & S_IFMT) != S_IFREG) {
         printf("file is no longer a regular file!\n");
         return;
     }
+
+    SHA1Checksum hash;
+    while (true) {
+        ssize_t res = read(fd, buf, sizeof(buf));
+        if (res < 0) {
+            if (errno == EINTR)
+                continue;
+            printf("Error while reading: %m\n");
+            return;
+        } else if (res == 0) {
+            break;
+        } else {
+            hash.process(buf, res);
+            size += res;
+        }
+    }
+
+    file_info["sha1"] = string((const char *)hash.checksum(),
+                               hash.checksum_size());
 }
 
 void scanfile(const string& path)
@@ -34,24 +68,61 @@ void scanfile(const string& path)
     int fd;
     long flags;
     struct stat stat_buf;
+    char *buf;
+    ssize_t len;
+
+    dictionary file_info;
+
     lstat(path.c_str(), &stat_buf);
 
-    printf("%s:\n", path.c_str());
-    printf("  ino=%Ld, perm=%04o, uid=%d, gid=%d, nlink=%d, blksize=%d, size=%Ld\n",
-           (int64_t)stat_buf.st_ino, stat_buf.st_mode & 07777,
-           stat_buf.st_uid, stat_buf.st_gid, stat_buf.st_nlink,
-           (int)stat_buf.st_blksize, (int64_t)stat_buf.st_size);
+    printf("%s\n", path.c_str());
+
+    file_info["mode"] = encode_u16(stat_buf.st_mode & 07777);
+    file_info["atime"] = encode_u64(encode_time(stat_buf.st_atime));
+    file_info["ctime"] = encode_u64(encode_time(stat_buf.st_ctime));
+    file_info["mtime"] = encode_u64(encode_time(stat_buf.st_mtime));
+    file_info["user"] = encode_u32(stat_buf.st_uid);
+    file_info["group"] = encode_u32(stat_buf.st_gid);
+
+    char inode_type;
 
     switch (stat_buf.st_mode & S_IFMT) {
     case S_IFIFO:
+        inode_type = 'p';
+        break;
     case S_IFSOCK:
+        inode_type = 's';
+        break;
     case S_IFCHR:
+        inode_type = 'c';
+        break;
     case S_IFBLK:
+        inode_type = 'b';
+        break;
     case S_IFLNK:
-        printf("  special file\n");
+        inode_type = 'l';
+
+        /* Use the reported file size to allocate a buffer large enough to read
+         * the symlink.  Allocate slightly more space, so that we ask for more
+         * bytes than we expect and so check for truncation. */
+        buf = new char[stat_buf.st_size + 2];
+        len = readlink(path.c_str(), buf, stat_buf.st_size + 1);
+        if (len < 0) {
+            printf("error reading symlink: %m\n");
+        } else if (len <= stat_buf.st_size) {
+            buf[len] = '\0';
+            printf("    contents=%s\n", buf);
+        } else if (len > stat_buf.st_size) {
+            printf("error reading symlink: name truncated\n");
+        }
+
+        file_info["contents"] = buf;
+
+        delete[] buf;
         break;
     case S_IFREG:
-        printf("  regular file\n");
+        inode_type = '-';
+
         /* Be paranoid when opening the file.  We have no guarantee that the
          * file was not replaced between the stat() call above and the open()
          * call below, so we might not even be opening a regular file.  That
@@ -70,21 +141,29 @@ void scanfile(const string& path)
         flags = fcntl(fd, F_GETFL);
         fcntl(fd, F_SETFL, flags & ~O_NONBLOCK);
 
-        dumpfile(fd);
+        file_info["size"] = encode_u64(stat_buf.st_size);
+        dumpfile(fd, file_info);
         close(fd);
 
         break;
     case S_IFDIR:
-        printf("  directory\n");
+        inode_type = 'd';
         scandir(path);
         break;
+
+    default:
+        fprintf(stderr, "Unknown inode type: mode=%x\n", stat_buf.st_mode);
+        return;
     }
+
+    file_info["type"] = string(1, inode_type);
+
+    info_dump->write_string(path);
+    info_dump->write_dictionary(file_info);
 }
 
 void scandir(const string& path)
 {
-    printf("Scan directory: %s\n", path.c_str());
-
     DIR *dir = opendir(path.c_str());
 
     if (dir == NULL) {
@@ -106,7 +185,6 @@ void scandir(const string& path)
     for (vector<string>::iterator i = contents.begin();
          i != contents.end(); ++i) {
         const string& filename = *i;
-        printf("  d_name = '%s'\n", filename.c_str());
         scanfile(path + "/" + filename);
     }
 
@@ -115,7 +193,20 @@ void scandir(const string& path)
 
 int main(int argc, char *argv[])
 {
-    scandir(".");
+    FILE *dump = fopen("fileinfo", "w");
+    if (dump == NULL) {
+        fprintf(stderr, "Cannot open fileinfo: %m\n");
+        return 1;
+    }
+
+    FileOutputStream os(dump);
+    info_dump = &os;
+
+    try {
+        scanfile(".");
+    } catch (IOException e) {
+        fprintf(stderr, "IOException: %s\n", e.getError().c_str());
+    }
 
     return 0;
 }