aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorRob Austein <sra@hactrn.net>2006-10-17 15:58:53 +0000
committerRob Austein <sra@hactrn.net>2006-10-17 15:58:53 +0000
commit82f67a12f7dbcbcd707cd4e376e32bd5b9cc3403 (patch)
tree3cd7bc56d18a07e0a01e64d55f2b838684719157
parent89872f3a558d4287850c17226944cefbefb00f0e (diff)
Update todo
svn path=/rcynic/README; revision=402
-rw-r--r--rcynic/README54
1 files changed, 21 insertions, 33 deletions
diff --git a/rcynic/README b/rcynic/README
index b03eb899..66c35718 100644
--- a/rcynic/README
+++ b/rcynic/README
@@ -501,36 +501,24 @@ compiler warnings when mixed with functions using size_t. Least bad
solution is probably to use size_t everywhere and cast to int when
calling into OpenSSL.
-- Statistics logging.
-
-Add a counter-fiddling function called inline, then digest and log
-results at end.
-
-Fiddling function would look something like:
-
- typedef enum {
- rsync_failed,
- accepted_current_cert,
- accepted_current_crl,
- accepted_backup_cert,
- accepted_backup_crl,
- rejected_current_cert,
- rejected_current_crl,
- rejected_backup_cert,
- rejected_backup_crl,
- COUNTER_T_MAX
- } counter_t;
-
- void record_stat(const rcynic_ctx_t *rc,
- const char *uri,
- counter_t counter,
- long increment);
-
- typedef struct {
- char hostname[URI_MAX];
- long counters[COUNTER_T_MAX];
- } host_counter_t;
-
-Most (all?) things would be hostname-based rather than URI-based, but
-this way we can localize the URI -> hostname transformation in one
-place and not worry about it.
+- Statistics and logging.
+
+Added a MIB counter mechanism, still needs work:
+
+-- Better summary at end of run, with a config file variable to
+ disable it, more useful formatting/organization, and an appropriate
+ log level (can be fairly high, since we assume separate config file
+ variable to enable the summary system).
+
+-- Could consolidate a lot of the junk with a new routine that was a
+ combination of mib_increment() and logmsg(). For things that
+ should be logged in verbose mode and summarized in summary mode,
+ the easiest thing would be a single routine that took all the
+ arguments necessary to do either or both, and did the right thing
+ for the current settings.
+
+ Yes, this involves creation of even more MIB counters, good thing
+ this is a general mechanism.
+
+-- Better names for the MIB counters we have, and clean up the ones
+ that we don't end up using.