This is my first crack at this issue. I've still got some problems.
I'm going to review the parts of the patch I'm having trouble with, and
have attached the complete patch for your perusal.
I have also attached my test repository with the dump file.
r1:2 is a replace of trunk/file.c
Much thanks to `eh' on the IRC channel for his patience and guidance!
@@ -50,6 +50,12 @@
+#define ADD "add"
+#define DEL "del"
+#define NESS_DEBUG 1
+const char * hash_get_set (apr_hash_t * ht, const void * key,
+ const char * add_or_del);
Not sure what to do with my own private functions and defines. HACKING
seems to be quiet on the matter. But I haven't seen any other functions
in code - so does everything go into header files?
Comment out the NESS_DEBUG define to stop the verbose trace lines.
@@ -995,18 +1013,6 @@
logfile_path, parent_path, pool));
*log_number = 0;
- /* The passed-in `path' is relative to the anchor of the edit, so if
- * the operation was invoked on something other than ".", then
- * `path' will be wrong for purposes of notification. However, we
- * can always count on the parent_path being the parent of base_name,
- * so we just join them together to get a good notification path.
- if (eb->notify_func)
- svn_wc_create_notify (svn_path_join (parent_path, base_name, pool),
- svn_wc_notify_update_delete, pool), pool);
Looking at this comment now I didn't use the svn_path_join() method like
it warns against. Does this comment still apply today as it did in
Now to the part that is getting me.
+ for (hi = apr_hash_first(db->pool, db->deleted_targets); hi; hi = apr_hash_next(hi))
+ const void *vkey;
+ void *vval;
+ const char *key;
+ const svn_string_t *val;
+ apr_hash_this(hi, &vkey, NULL, &vval);
+ key = vkey;
+ val = vval;
+ fprintf(stderr, "[%d] hash key = '%s' was found with val = '%s'\n", i, key, (char *)val);
+ if ((char*)val == DEL && db->edit_baton->notify_func)
+ svn_wc_create_notify (key,
+ svn_wc_notify_update_delete, pool), pool);
I'm thinking the const char * is killing me. In my testing it is giving
the same entries (yes, I know what the constant keyword means), and not
seeming to loop though the hash table as I expected for the key values.
But this is what the function arp_hash_this() returns for the key!
How can I get around the const void * above and loop through the hash
table with proper keys returned? This seems rather common to me to want
to get back the keys you entered.
The next loop is bigger and uglier than needed but it tries to delete
the entries in the loop and probably fails to do so since the keys do
not appear to be updating.
Comments and hints on beating this apr_hash_this() problem please.
I'm helping a friend build a shed tomorrow so I won't be around for most
of the day.
PGP Public Key: http://www.nesser.org/pgp-key/
20:05:22 up 5:43, 3 users, load average: 0.53, 0.55, 0.33
Received on Mon Sep 12 04:11:08 2005