Rewritten to incorporate the functionality of

sv_sync_www_flush and to make updates more safe.


git-svn-id: file:///svnroot/gsc/trunk@181 d2de0444-eb31-0410-8365-af798a554d48

1 files changed, 447 insertions, 120 deletions

diff --git a/cvs/sv_sync_www.c b/cvs/sv_sync_www.c
index e61aab1..3db8150 100644
--- a/cvs/sv_sync_www.c
+++ b/cvs/sv_sync_www.c
@@ -2,7 +2,7 @@
 
  Copyright (C) Gordon Matzigkeit <gord@fig.org>, Loic Dachary
  <loic@gnu.org>, 2001
- Copyright (C) 2005, Sergey Poznyakoff
+ Copyright (C) 2005, 2006 Sergey Poznyakoff
  
  This program is free software; you can redistribute it and/or
  modify it under the terms of the GNU General Public License
@@ -21,24 +21,24 @@
 
 */
 
-/* sf_sync_www.c - Allow unprivileged users to do a web update on www.gnu.org
+/* sv_sync_www.c - Allow unprivileged users to do a web update on www.gnu.org.ua
 
   In CVSROOT/loginfo file
   ALL      /bin/sv_sync_www_schedule ${USER} ${CVSROOT} %s
 
   In crontab (remove the backslash before /):
-  0 *\/1 * * * /bin/sv_sync_www_flush
+  0 *\/1 * * * /bin/sv_sync_www -s /var/spool/savane
 
   Or, to synchronize right after commit, in CVSROOT/loginfo:
   
-  ALL     (date; cat; (sleep 2; /bin/sf_sync_www %{s} ) &\ ) >> /var/log/sf_sync_www 2>&1
+  ALL     (date; cat; (sleep 2; /bin/sv_sync_www %{s} ) &\ ) >> /var/log/sv_sync_www 2>&1
 
   For test purpose:
-  ./sf_sync_www -n ' bla bla'
-  ./sf_sync_www -n 'foo/bar bla bla'
-  ./sf_sync_www -n 'f'\''oo bla bla'
-  ./sf_sync_www -n 'f'\''oo - New directory'
-  ./sf_sync_www -n 'bla - Imported sources'
+  ./sv_sync_www -n ' bla bla'
+  ./sv_sync_www -n 'foo/bar bla bla'
+  ./sv_sync_www -n 'f'\''oo bla bla'
+  ./sv_sync_www -n 'f'\''oo - New directory'
+  ./sv_sync_www -n 'bla - Imported sources'
 
   The idea is that it runs a cvs update -l (to prevent recursion) in the 
   directory where the commit was done. Since the command will be called
@@ -87,6 +87,9 @@
   Introduce command line options, make program configurable and
   implement .symlinks functionality
   Sergey Poznyakoff <gray@gnu.org.ua>, 2005-07-31
+
+  Implement 'scan directory' mode
+  Sergey Poznyakoff <gray@gnu.org.ua>, 2005-02-09
   
 */
 
@@ -95,19 +98,24 @@
 #endif
 
 #include <sys/types.h>
+#include <sys/wait.h>
+#include <sys/stat.h>
 #include <ctype.h>
-#include <string.h>
+#include <unistd.h>
+#include <stdlib.h>
 #include <stdio.h>
-#include <sys/wait.h>
+#include <string.h>
+#include <dirent.h>
 
 #include <error.h>
 #include <errno.h>
 
-#include <sys/stat.h>
-#include <unistd.h>
-#include <stdlib.h>
 #include <pwd.h>
 
+#define obstack_chunk_alloc malloc
+#define obstack_chunk_free free
+#include <obstack.h>
+
 /*
  * Return a malloc'ed copy of the string argument with all '
  * substituted by '\'' and a trailing and leading '.
@@ -115,7 +123,7 @@
  *                fib'ou -> 'fib'\''ou'
  */
 char *
-quote (char *string)
+quote (const char *string)
 {
   int i;
   int length;
@@ -135,7 +143,7 @@ quote (char *string)
   out = (char *) malloc (length + 1 + count * 3 + 2);
   if (count > 0)
     {
-      char *from_p = string;
+      const char *from_p = string;
       char *to_p = out;
       *to_p++ = '\'';
       while (*from_p)
@@ -157,37 +165,45 @@ quote (char *string)
   return out;
 }
 
-char *user;
 char *rsh_program = "/usr/bin/ssh";
 char *hostname;
-char *repository = "/webcvs";
+char *default_repository = "/webcvs";
+char *default_spooldir = "/var/spool/savane";
 char *access_method = NULL;
 char *www_directory = "/home/puszcza/software";
 char *cvs_command = "CVS_RSH=ssh cvs -q -z3";
+size_t min_uid = 100;
+size_t min_gid = 100;
 int dry_run = 0;
 int debug = 0;
 
 void
 usage ()
 {
-  printf ("sv_sync_www [OPTIONS] directory\n\n");
+  printf ("sv_sync_www [OPTIONS] DIRECTORY\n\n");
   printf ("OPTIONS are:\n");
+  printf ("\nMode selection:\n");
+  printf (" -h                Display this help summary\n");
+  printf (" -s                Scan mode: process each file in DIRECTORY\n");
+  printf ("\nMode modification:\n");
+  printf (" -n                Dry run: do nothing, only print what would have been done\n");
+  printf (" -x                Enable debugging output\n");
+  printf ("\nConfiguration:\n");
   printf (" -c COMMAND        Set cvs command to use (default %s)\n",
 	  cvs_command);
   printf (" -H HOSTNAME       Set hostname or IP address of the host where WWW directory\n");
   printf ("                   resides (default: none, WWW and CVS reside on the same host)\n");
-  printf (" -h                Display this help summary\n");
-  printf (" -n                Dry run: do nothing, only print what would have been done\n");
   printf (" -m METHOD         Set default access method. METHOD will be prepended to\n");
   printf ("                   REPOSITORY (see -R option below) before accessing the\n");
   printf ("                   repository\n");
   printf (" -r RSH_PROGRAM    Set full name of RSH program\n");
   printf ("                   (default: %s)\n", rsh_program);
   printf (" -R REPOSITORY     Set full name of the CVS repository\n");
-  printf ("                   (default: %s)\n", repository);
+  printf ("                   (default: %s)\n", default_repository);
   printf (" -w WWW_DIRECTORY  Set full name of the WWW directory\n");
   printf ("                   (default: %s)\n", www_directory);
-  printf (" -x                Enable debugging output");
+  printf (" -U UID            Set minimum allowable UID value\n");
+  printf (" -G UID            Set minimum allowable GID value\n");
 }
 
 static int
@@ -220,7 +236,7 @@ run_cmd (char *cmd)
       else if (WIFSIGNALED (status))
         {
           error (0, 0, "subprocess terminated on signal %d",
-                   WTERMSIG (status));
+		 WTERMSIG (status));
           return -1;
         }
       else
@@ -266,24 +282,377 @@ run_cmd (char *cmd)
 }  
 
 int
+has_file_list (const char *arg)
+{
+  /*
+   * If this is a new directory creation or an import command,
+   * there are no files to collect in the argument list.
+   * Otherwise, there are files to collect. 
+   */
+  return (strcmp (arg, "- New directory")
+	  && strcmp (arg, "- Imported sources"));
+}
+
+static int
+split_arg (const char *arg, char **pdir, char **pfiles)
+{
+  char *tmp;
+  /*
+   * First comes the relative directory name or empty string if current.
+   * directory = 0 if no subdirectory
+   * directory = relative path if subdirectory
+   */
+  if ((tmp = strchr (arg, ' ')))
+    {
+      size_t len = tmp - arg;
+      if (len > 0)
+	{
+	  *pdir = malloc (len + 1);
+	  if (!*pdir)
+	    return -1;
+	  memcpy (*pdir, arg, len);
+	  (*pdir)[len] = 0;
+	}
+      else
+	*pdir = NULL;
+      *pfiles = strdup (tmp + 1);
+      if (!*pfiles)
+	{
+	  free (*pdir);
+	  return -1;
+	}
+      return 0;
+    }
+  return 1;
+}
+
+
+int
+sync_www (const char *repository, const char *unquoted_directory, int files)
+{
+  int rc;
+  char *directory = quote (unquoted_directory);
+  char *cmd;
+  
+  /* Check that the local directory exists. */
+  if (!dry_run && files && unquoted_directory)
+    {
+      struct stat stbuf;
+      char *localdir;
+      size_t len = strlen (repository);
+
+      localdir = malloc (len + 1 + strlen (unquoted_directory) + 1);
+      if (!localdir)
+	{
+	  error (0, 0, "not enough memory");
+	  return 1;
+	}
+
+      strcpy (localdir, repository);
+      if (localdir[len - 1] != '/')
+        localdir[len++] = '/';
+      strcpy (localdir + len, unquoted_directory);
+
+      /* Check to see that the local unquoted_directory actually exists. */
+      if (stat (localdir, &stbuf) ||
+          (!S_ISDIR (stbuf.st_mode) && (errno = ENOTDIR)))
+	{
+	  error (0, errno, "Cannot find `%s'", localdir);
+	  return 1;
+	}
+    }
+
+  /* Update from CVS */
+  if (directory && !files)
+    asprintf (&cmd, "cd %s && ( %s -d %s%s checkout %s )",
+              www_directory, cvs_command, access_method ? access_method : "",
+	      repository, directory);
+  else if (!directory && files)
+    asprintf (&cmd, "cd %s && ( %s update -l )",
+              www_directory, cvs_command);
+  else if (directory && files)
+    asprintf (&cmd, "cd %s/%s && ( %s update -l )",
+              www_directory, directory, cvs_command);
+  else
+    {
+      error (0, 0, "Unexpected null directory and files");
+      return 1;
+    }
+  
+  if (run_cmd (cmd))
+    return 1;
+
+  free (cmd);
+
+  /* Restore symlinks */
+  asprintf (&cmd,
+            "cd %s/%s && test -r .symlinks && "
+            "(grep -v '^[ \t]*[;#]' .symlinks | "
+            "sed 's,^/,_,;s,\\.\\./,__/,g' |"
+            "while read S T; do ln -sf $S $T; done)",
+            www_directory, directory);
+  
+  rc = run_cmd (cmd);
+  free (cmd);
+  return rc;
+}  
+
+struct sync_entry
+{
+  char *repo;
+  char *dir;
+  int files;
+};
+
+static int
+dir_cmp (const char *a, const char *b)
+{
+  if (!a)
+    {
+      if (b)
+	return -1;
+      else
+	return 0;
+    }
+  else if (!b)
+    return 1;
+  return strcmp (a, b);
+}
+
+static int
+ent_cmp (const void *a, const void *b)
+{
+  const struct sync_entry *sa = a; 
+  const struct sync_entry *sb = b; 
+  int rc;
+
+  rc = strcmp (sa->repo, sb->repo);
+  if (rc)
+    return rc;
+  rc = dir_cmp (sa->dir, sb->dir);
+  if (rc)
+    return rc;
+  return sa->files - sb->files;
+}
+
+int
+sync_from_file (uid_t uid, const char *spooldir, const char *file)
+{
+  char *buf = NULL;
+  size_t size = 0;
+  size_t line = 0;
+  char *user, *dir, *filelist;
+  int err = 0;
+  struct obstack stk;
+  struct sync_entry *sp;
+  struct sync_entry ent;
+  struct sync_entry *sync_array;
+  FILE *fp = fopen (file, "r");
+  
+  if (!fp)
+    {
+      error (0, errno, "Cannot open file `%s/%s'", spooldir, file);
+      return 1;
+    }
+
+  obstack_init (&stk);
+  while (getline (&buf, &size, fp) > 0)
+    {
+      char *p;
+      struct passwd *pw;
+
+
+      p = buf + strlen (buf) - 1;
+      if (*p == '\n')
+	*p = 0;
+      
+      line++;
+
+      p = strtok (buf, " \t");
+      if (!p)
+	{
+	  error (0, 0, "%s/%s:%lu: Invalid line",
+		 spooldir, file,
+		 (unsigned long) line);
+	  err = 1;
+	  break;
+	}
+      
+      pw = getpwnam (p);
+      if (!pw)
+	{
+	  error (0, 0, "%s/%s:%lu: cannot get password entry for user %s", 
+		 spooldir, file,
+		 (unsigned long) line, p);
+	  err = 1;
+	  break;
+	}
+      if (uid != pw->pw_uid)
+	{
+	  error (0, 0, "%s/%s:%lu: invalid uid", 
+		 spooldir, file,
+		 (unsigned long) line);
+	  err = 1;
+	  break;
+	} 
+      
+      p = strtok (NULL, " \t");
+      if (!p)
+	{
+	  error (0, 0, "%s/%s:%lu: Invalid line, second field is missing",
+		 spooldir, file,
+		 (unsigned long) line);
+	  err = 1;
+	  break;
+	}
+      ent.repo = strdup (p);
+      if (!ent.repo)
+	{
+	  error (0, ENOMEM, "%s/%s:%lu",
+		 spooldir, file,
+		 (unsigned long) line);
+	  err = 1;
+	  break;
+	}
+
+      p = strtok (NULL, "");
+      if (!p)
+	{
+	  error (0, 0, "%s/%s:%lu: Invalid line, only two fields",
+		 spooldir, file,
+		 (unsigned long) line);
+	  err = 1;
+	  break;
+	}
+      err = split_arg (p, &ent.dir, &p);
+      if (err)
+	{
+	  error (0, err < 0 ? errno : 0, "%s/%s:%lu: Cannot split argument",
+		 spooldir, file,
+		 (unsigned long) line);
+	  break;
+	}
+      ent.files = has_file_list (p);
+      free (p);
+      
+      obstack_grow (&stk, &ent, sizeof (ent));
+    }
+  fclose (fp);
+
+  sync_array = obstack_finish (&stk);
+
+  if (err == 0)
+    {
+      struct sync_entry *sp, *lastp = NULL;
+      
+      /* Sort the array */
+      qsort (sync_array, line, sizeof ent, ent_cmp);
+
+      for (sp = sync_array; sp < sync_array + line; sp++)
+	{
+	  if (lastp)
+	    {
+	      if (strcmp (lastp->repo, sp->repo) == 0
+		  && dir_cmp (lastp->dir, sp->dir) == 0)
+		continue;
+	    }
+	  sync_www (sp->repo, sp->dir, sp->files);
+	  lastp = sp;
+	}
+    }
+
+  for (sp = sync_array; sp < sync_array + line; sp++)
+    {
+      free (sp->repo);
+      free (sp->dir);
+    }
+
+  obstack_free (&stk, NULL);
+  
+  return err;
+}
+
+int
+scan_spool (const char *spooldir)
+{
+  DIR *dir;
+  struct dirent *ent;
+  size_t total_count = 0;   /* Total number of jobs seen */
+  size_t success_count = 0; /* Number of successfully processed jobs */
+  
+  if (chdir (spooldir))
+    {
+      error (0, errno, "Cannot change to `%s'", spooldir);
+      return 1;
+    }
+
+  dir = opendir (".");
+  if (!dir)
+    {
+      error (0, errno, "Cannot open directory `%s'", spooldir);
+      return 1;
+    }
+
+  while ((ent = readdir (dir)))
+    {
+      struct stat st;
+      
+      if (ent->d_name[0] == '.')
+	continue;
+
+      total_count++;
+      if (stat (ent->d_name, &st))
+	{
+	  error (0, errno, "Cannot stat file `%s/%s'", spooldir, ent->d_name);
+	  continue;
+	}
+
+      if (S_ISREG (st.st_mode))
+	{
+	  if (st.st_uid < min_uid || st.st_gid < min_gid)
+	    {
+	      error (0, 0, "Ignoring file `%s/%s': UID or GID out of range",
+		     spooldir, ent->d_name);
+	      continue;
+	    }
+
+	  if (sync_from_file (st.st_uid, spooldir, ent->d_name) == 0)
+	    {
+	      if (!dry_run && unlink (ent->d_name))
+		error (0, errno, "Cannot unlink file `%s/%s'",
+		       spooldir, ent->d_name);
+		  
+	      success_count++;
+	    }
+	}
+    }
+
+  printf ("Total files: %lu; Jobs processed: %lu\n",
+	  (unsigned long) total_count,
+	  (unsigned long) success_count);
+  
+  closedir (dir);
+  return 0;
+}
+
+int
 main (int argc, char **argv)
 {
   int i;
-  char *unquoted_directory = 0;
-  char *directory = 0;
-  int files = 0;
-  char *cmd;
-  char *tmp;
   uid_t uid;
   gid_t gid;
+  char *p;
+  int spooldir_mode;
+  char *user = NULL;
+  char *repository = default_repository;
   
-  while ((i = getopt (argc, argv, "c:hH:m:nr:R:u:w:x")) != EOF)
+  while ((i = getopt (argc, argv, "c:hH:m:nr:R:su:U:G:w:x")) != EOF)
     switch (i)
       {
       case 'c':
 	cvs_command = optarg;
 	break;
-	
+
       case 'H':
         hostname = optarg;
         break;
@@ -292,6 +661,12 @@ main (int argc, char **argv)
         usage ();
         exit (0);
 
+      case 'G':
+	min_gid = strtoul (optarg, &p, 0);
+	if (p)
+	  error (1, 0, "invalid numeric value (near %s)", p);
+	break;
+	
       case 'm':
 	access_method = optarg;
 	break;
@@ -308,6 +683,16 @@ main (int argc, char **argv)
         repository = optarg;
         break;
 
+      case 's':
+	spooldir_mode = 1;
+	break;
+
+      case 'U':
+	min_uid = strtoul (optarg, &p, 0);
+	if (p)
+	  error (1, 0, "invalid numeric value (near %s)", p);
+	break;
+	  
       case 'u':
         user = optarg;
         break;
@@ -355,101 +740,43 @@ main (int argc, char **argv)
   if (setreuid (uid, uid))
     error (1, errno, "cannot assume effective user id");
   
-  if (argc == 1)
+  if (spooldir_mode)
     {
-      tmp = argv[0];
-      /*
-       * First comes the relative directory name or empty string if current.
-       * directory = 0 if no subdirectory
-       * directory = relative path if subdirectory
-       */
-      unquoted_directory = tmp;
-      if ((tmp = strchr (tmp, ' ')))
-        {
-          *tmp = '\0';
-          if (tmp - unquoted_directory <= 0)
-            unquoted_directory = 0;
-          tmp++;
-        }
+      char *spooldir = NULL;
+
+      if (argc > 1)
+	error (1, 0, "Extra arguments in scan mode");
+      else if (argc == 1)
+	spooldir = argv[0];
       else
-        error (1, 0, "Failed to find directory in first argument: `%s'",
-               argv[0]);
-    }
-  else if (argc == 2)
-    {
-      unquoted_directory = argv[0];
-      tmp = argv[1];
-    }
-  else
-    error (1, 0, "You must specify one or two arguments");
+	spooldir = default_spooldir;
 
-  directory = quote (unquoted_directory);
-  
-  if (!strcmp (tmp, "- New directory")
-      || !strcmp (tmp, "- Imported sources"))
-    {
-      /*
-       * If this is a new directory creation or an import command,
-       * there are no files to collect in the argument list.
-       */
+      return scan_spool (spooldir);
     }
   else
     {
-      /*
-       * There are files to collect in the argument list.
-       */
-      files = 1;
-    }
-
-  /* Check that the local directory exists. */
-  if (!dry_run && files && unquoted_directory)
-    {
-      struct stat stbuf;
-      char *localdir;
-
-      localdir = malloc (i + 1 + strlen (unquoted_directory) + 1);
-      if (!localdir)
-        error (1, 0, "not enough memory");
-
-      i = strlen (repository);
-
-      strcpy (localdir, repository);
-      if (localdir[i - 1] != '/')
-        localdir[i++] = '/';
-      strcpy (localdir + i, unquoted_directory);
+      char *unquoted_directory = NULL;
+      char *tmp;
+      
+      if (argc == 1)
+	{
+	  int rc = split_arg (argv[0], &unquoted_directory, &tmp);
+	  if (rc < 0)
+	    error (1, errno, "Cannot split argument");
+	  else if (rc > 0)
+	    error (1, 0, "Failed to find directory in first argument: `%s'",
+		   argv[0]);
+	}
+      else if (argc == 2)
+	{
+	  unquoted_directory = argv[0];
+	  tmp = argv[1];
+	}
+      else
+	error (1, 0, "You must specify one or two arguments");
 
-      /* Check to see that the local unquoted_directory actually exists. */
-      if (stat (localdir, &stbuf) ||
-          (!S_ISDIR (stbuf.st_mode) && (errno = ENOTDIR)))
-        error (1, errno, "Cannot find `%s'", localdir);
+      return sync_www (repository, unquoted_directory, has_file_list (tmp));
     }
-
-  /* Update from CVS */
-  if (directory && !files)
-    asprintf (&cmd, "cd %s && ( %s -d %s%s checkout %s )",
-              www_directory, cvs_command, access_method ? access_method : "",
-	      repository, directory);
-  else if (!directory && files)
-    asprintf (&cmd, "cd %s && ( %s update -l )",
-              www_directory, cvs_command);
-  else if (directory && files)
-    asprintf (&cmd, "cd %s/%s && ( %s update -l )",
-              www_directory, directory, cvs_command);
-  else
-    error (1, 0, "Unexpected null directory and files");
-  
-  if (run_cmd (cmd))
-    return 1;
-
-  free (cmd);
-
-  /* Restore symlinks */
-  asprintf (&cmd,
-            "cd %s/%s && test -r .symlinks && "
-            "(grep -v '^[ \t]*[;#]' .symlinks | "
-            "sed 's,^/,_,;s,\\.\\./,__/,g' |"
-            "while read S T; do ln -sf $S $T; done)",
-            www_directory, directory);
-  
-  return run_cmd (cmd);
+  /*NOTREACHED*/
+  return 0;
 }