Stale Branches

[John Keeping]

On Wed, Aug 12, 2015 at 03:49:06PM +0200, Jason A. Donenfeld wrote:
> I'd like some review of a few branches that have been sitting for a
> while, and perhaps a rebased version:
> 
> jk/http-status-codes: Should be pretty much okay and ready to merge,
> but John -- anything that needs to be done here?

I have a fixup on my local copy of this, but I can't remember why it's
needed :-S

The change is:

-- >8 --
diff --git a/ui-shared.c b/ui-shared.c
index c11d9b6..5126962 100644
--- a/ui-shared.c
+++ b/ui-shared.c
@@ -770,6 +770,7 @@ void cgit_print_docend(void)
 void cgit_print_error_page(int code, const char *msg, const char *fmt, ...)
 {
        va_list ap;
+       ctx.page.expires = 5 * 60; /* 5 minutes */
        ctx.page.status = code;
        ctx.page.statusmsg = msg;
        cgit_print_http_headers();
-- 8< --

On top of that, I have another change to stop caching "temporary"
failures and I think this change is related to that one.  Since we write
the expiry time into the "Expires" HTTP header we want to avoid using
the 10 year default if the error is being generated for a transient
failure (e.g. a missing object if someone generates a link to something
that hasn't yet reached the repository mirror where CGit is hosted).

I think it makes sense to take the series as it is (that is, without the
above fixup) and we can build more improvements in this area on top of
it.

> jk/follow-renames: This is old, but a good idea. Can you get this in
> shape for modern cgit?

There aren't too many changes, I'll send the updated patches soon.

> lh/grep: I always liked this idea, but Lars never completed it.
> Anybody want to take over?
> 
> lh/pretty-blob-view: I'm not sure what the original intent of this
> branch was. It doesn't really look prettier. If anybody has some ideas
> here and wants to revive it, feel free.
> 
> lf/filter: This seems relatively straight forward. Anybody have
> objections or nits?

Given that prefixcmp() has gone (replaced by starts_with() in git.git),
I think this should now go straight to using skip_prefix() as the commit
message suggests.