--- Begin Message ---
|
Subject: |
RE: CVS lockdir bug with repository symlinks |
|
Date: |
Mon, 23 Jun 2003 15:58:25 -0400 |
Believe it or not, I'm not too familiar with diffs :)
Here's 'diff -u' output, hope this is what you're looking for.
--- lock.c.orig Mon Jun 23 15:56:39 2003
+++ lock.c Mon Jun 23 15:56:54 2003
@@ -158,6 +158,10 @@
mode_t save_umask;
int saved_umask = 0;
+ /*full path in actual repository, less relative portions and
+ symbolic links*/
+ char real_path[PATH_MAX];
+
if (lock_dir == NULL)
{
/* This is the easy case. Because the lock files go directly
@@ -176,9 +180,53 @@
assert (current_parsed_root->directory != NULL);
assert (strncmp (repository, current_parsed_root->directory,
strlen (current_parsed_root->directory)) == 0);
- short_repos = repository + strlen (current_parsed_root->directory) +
1;
- if (strcmp (repository, current_parsed_root->directory) == 0)
+ /*in order to avoid problems with symlinks in the repository, which
would
+ lead to putting locks in the wrong directories, expand the given
path
+ to give the REAL path to the directory we are accessing*/
+ if(realpath(repository, real_path)==NULL)
+ {
+ error(1, errno, "Error evaluating real path for %s",
repository);
+ }
+
+ /*if realpath != original given path, then some symlink or possibly
+ relative dirs were expanded, run extra checks on expanded path,
such
+ as making sure it exists and making sure it points to somewhere
within
+ the repository*/
+ if(strcmp(real_path, repository))
+ {
+
+ /*check if evaluated path actually exists, this may be
overkill
+ but doesn't hurt to check*/
+ if(access(real_path, F_OK))
+ {
+ error(1, 0, "path %s evaluated to non-existant
location %s",
+ repository, real_path);
+ }
+
+ /*check to make sure evaluated path still remains in the
repository,
+ simply check the path against the cvs root dir*/
+ if(strncmp (real_path, current_parsed_root->directory,
+
strlen(current_parsed_root->directory)))
+ {
+ if(islink(repository))
+ {
+ error(1, 0, "Error: path %s (symlink to %s)
leads out of repository",
+ repository, real_path);
+ }
+ else
+ error(1, 0, "Error: path %s leads out of
repository",
+ repository);
+ }
+ }
+ else
+ {
+ strcpy(real_path, repository);
+ }
+
+ short_repos = real_path + strlen(current_parsed_root->directory) +
1;
+
+ if (strcmp (real_path, current_parsed_root->directory) == 0)
short_repos = ".";
else
assert (short_repos[-1] == '/');
@@ -276,7 +324,8 @@
saved_umask = 0;
}
}
- return retval;
+
+ return retval;
}
/*
@@ -636,13 +685,18 @@
struct dirent *dp;
struct stat sb;
int ret = 0;
+ char *real_repository_path;
#ifdef CVS_FUDGELOCKS
again:
#endif
- if ((dirp = CVS_OPENDIR (repository)) == NULL)
- error (1, 0, "cannot open directory %s", repository);
+ /*use lock_name to give use the real path to look in, which is
+ dependent on whether a separate lockdir is in use or not*/
+ real_repository_path = lock_name (repository, "");
+
+ if ((dirp = CVS_OPENDIR (real_repository_path)) == NULL)
+ error (1, 0, "cannot open directory %s", real_repository_path);
errno = 0;
while ((dp = CVS_READDIR (dirp)) != NULL)
@@ -654,8 +708,8 @@
(void) time (&now);
#endif
- line = xmalloc (strlen (repository) + strlen (dp->d_name) + 5);
- (void) sprintf (line, "%s/%s", repository, dp->d_name);
+ line = xmalloc (strlen (real_repository_path) + strlen
(dp->d_name) + 5);
+ (void) sprintf (line, "%s/%s", real_repository_path,
dp->d_name);
if ( CVS_STAT (line, &sb) != -1)
{
#ifdef CVS_FUDGELOCKS
@@ -690,7 +744,7 @@
errno = 0;
}
if (errno != 0)
- error (0, errno, "error reading directory %s", repository);
+ error (0, errno, "error reading directory %s",
real_repository_path);
CVS_CLOSEDIR (dirp);
return (ret);
-----Original Message-----
From: Derek Robert Price [mailto:derek@ximbiot.com]
Sent: Monday, June 23, 2003 3:45 PM
To: Ambalu, Robert
Cc: 'bug-cvs@gnu.org'
Subject: Re: CVS lockdir bug with repository symlinks
Ambalu, Robert wrote:
>There is a bug in cvs when using a separate lock directory and the
>repository has symlinked files within it. If there are symlinks in the
>repository, the locks will get placed at the location of the symlink rather
>than the location of the actual file (which leads to inconsistent locking).
>
>
Could you send this patch again as a diff in context or universal
format? I prefer universal (`diff -u').
Derek
--
*8^)
Email: derek@ximbiot.com
Get CVS support at <http://ximbiot.com>!
--
If a man speaks in the forest, and there is no
woman around to hear him, is he still wrong?
--- End Message ---