[PATCH 00/17] HTTP response code improvements

[John Keeping]

On Sun, Apr 05, 2015 at 04:54:46PM +0100, John Keeping wrote:
> Following Nicolas' report[0] about CGit caching error pages I've been
> looking into ways of (optionally) avoiding caching error responses.
> 
> This series doesn't achieve that but it does give us two features that
> will help to do so (neither of which are the case without this series of
> patches):
> 
> 1) Always return an HTTP error code if something goes wrong
> 2) Always set "ctx.page.status" when returning an error page

The change to avoid caching errors is quite simple on top of this
series:

-- >8 --
diff --git a/cache.c b/cache.c
index cd99812..be41467 100644
--- a/cache.c
+++ b/cache.c
@@ -261,6 +261,7 @@ unsigned long hash_str(const char *str)
 static int process_slot(struct cache_slot *slot)
 {
 	int err;
+	int use_new_slot;
 
 	err = open_slot(slot);
 	if (!err && slot->match) {
@@ -333,7 +334,8 @@ static int process_slot(struct cache_slot *slot)
 	// Lets avoid such a race by just printing the content of
 	// the lock file.
 	slot->cache_fd = slot->lock_fd;
-	unlock_slot(slot, 1);
+	use_new_slot = !ctx.page.status || ctx.page.status == 200;
+	unlock_slot(slot, use_new_slot);
 	if ((err = print_slot(slot)) != 0) {
 		cache_log("[cgit] error printing cache %s: %s (%d)\n",
 			  slot->cache_name,
-- 8< --

But I think we probably want something a bit more configurable than that
since there are some errors that users might want to cache.