While trying to figure out a way to import symblic links, I started
reading 1.10.7 code and discovered that the revision files can already
handle symbolic links in most routines (and most importantly
add_rcs_file()), using the PRESERVE_PERMISSIONS_SUPPORT code.
I'm curious why import ignores symlinks, then. From the import output
documentation:
`L FILE'
The file is a symbolic link; `cvs import' ignores symbolic
links. People periodically suggest that this behavior should
be changed, but if there is a consensus on what it should be
changed to, it doesn't seem to be apparent. (Various options
in the `modules' file can be used to recreate symbolic links on
checkout, update, etc).
Now looking at the PreservePermissions code, it seems a simple matter to
allow import to do symbolic links as well:
--- import.c.orig Sun May 7 13:18:41 2000
+++ import.c Sun May 7 13:20:35 2000
@@ -451,17 +451,16 @@
n->key = xstrdup (dp->d_name);
addnode(dirlist, n);
}
- else if (
+ else
+ {
+ if (
#ifdef DT_DIR
dp->d_type == DT_LNK || dp->d_type == DT_UNKNOWN &&
#endif
islink (dp->d_name))
{
add_log ('L', dp->d_name);
- err++;
}
- else
- {
#ifdef CLIENT_SUPPORT
if (client_active)
err += client_process_import_file (message, dp->d_name,
This produces an L line, followed by an N line, on import of a symlink.
I have tried this patch and as long as PerservePermissions is on, it
seems to work fine on import; export already worked. I have observed
that HEAD of the devel tree doesn't have any symlink-related change to
import.c either.
I am using CVS for SA work (configuration management) and as such, have
a very high degree of interest in maintaining symlinks. Does anyone
know why the code that is already there isn't used on import? Or if
there is some reason I should not use this patch at our site?
--
Scott
