fix and test Rack hijack support

Rack hijacking may close the socket before it yields control back to our
epoll event loop.  So it's not safe to attempt to use the socket (or
even get the .fileno) with :delete/EPOLL_CTL_DEL.

So change :delete to :ignore, and only decrement the FD counter to allow
yahns to do graceful shutdown.  This means we'll potentially waste ~200
bytes of kernel memory due to epoll overhead until the FD is closed by
the app hijacking.

I'm not sure how Rack servers handle graceful shutdown when
hijacking, but maybe they just retrap SIGQUIT...

1 files changed, 0 insertions, 8 deletions

diff --git a/lib/yahns/fdmap.rb b/lib/yahns/fdmap.rb
index 1e1677a..d83898b 100644
--- a/lib/yahns/fdmap.rb
+++ b/lib/yahns/fdmap.rb
@@ -54,14 +54,6 @@ class Yahns::Fdmap # :nodoc:
     @fdmap_mtx.synchronize { @count -= 1 }
   end
 
-  def delete(io) # use with rack.hijack (via yahns)
-    fd = io.fileno
-    @fdmap_mtx.synchronize do
-      @fdmap_ary[fd] = nil
-      @count -= 1
-    end
-  end
-
   # expire a bunch of idle clients and register the current one
   # We should not be calling this too frequently, it is expensive
   # This is called while @fdmap_mtx is held