Make channel tests more robust to stderr output

At the start of feb 2019 the tests started to fail on macos on Travis.  No
changes were made to the repo, so the problem must be environmental.

It turns out that the server was emitting stderr output, and since the
tests were not consuming stderr, this made `channel.eof()` always
return false.

The resolution is to drain both stdout and stderr in the tests before
making assertions with `channel.eof()`.

I've also updated the doc comment on `channel.eof()` to reflect this,
as it is a bit of a sharp edge.

1 files changed, 8 insertions, 1 deletions

diff --git a/src/channel.rs b/src/channel.rs
index 03bbba0..7086f5d 100644
--- a/src/channel.rs
+++ b/src/channel.rs
@@ -328,7 +328,10 @@ impl Channel {
         self.read_limit = Some(limit);
     }
 
-    /// Check if the remote host has sent an EOF status for the selected stream.
+    /// Check if the remote host has sent an EOF status for the channel.
+    /// Take care: the EOF status is for the entire channel which can be confusing
+    /// because the reading from the channel reads only the stdout stream.
+    /// unread, buffered, stderr data will cause eof() to return false.
     pub fn eof(&self) -> bool {
         self.read_limit == Some(0) || unsafe { raw::libssh2_channel_eof(self.raw) != 0 }
     }
@@ -342,6 +345,10 @@ impl Channel {
     }
 
     /// Wait for the remote end to send EOF.
+    /// Note that unread buffered stdout and stderr will cause this function
+    /// to return `Ok(())` without waiting.
+    /// You should call the eof() function after calling this to check the
+    /// status of the channel.
     pub fn wait_eof(&mut self) -> Result<(), Error> {
         unsafe { self.sess.rc(raw::libssh2_channel_wait_eof(self.raw)) }
     }