/brz/remove-bazaar : changes

» Changes from revision 4797.85.31

To get this branch, use:

bzr branch
http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar

From Revision 4797.85.31 to 4797.85.12

expand all

Rev	Summary	Authors	Date
4797.85.31	Categorize all of the requests as safe/unsafe/semi/stream fo... Categorize all of the requests as safe/unsafe/semi/stream for retrying purposes.	John Arbash Meinel	14 years ago
4797.85.30	Document the functions in the new class. Document the functions in the new class.	John Arbash Meinel	14 years ago
4797.85.29	Move the hooks code over as well. Move the hooks code over as well. Basically, anything that needed all of these arguments should just be moved into the state class.	John Arbash Meinel	14 years ago
4797.85.28	Move all of the send logic into the _SmartClientRequest clas... Move all of the send logic into the _SmartClientRequest class. This shortens a lot of the function names, and avoids passing 6 arguments between lots of different functions so that we can retry requests, etc.	John Arbash Meinel	14 years ago
4797.85.27	I've started passing around too many parameters to too many ... I've started passing around too many parameters to too many functions. Encapsulate the state into a Request class.	John Arbash Meinel	14 years ago
4797.85.26	If we get ConnectionReset and we are aborting the call reset... If we get ConnectionReset and we are aborting the call reset the request. Even if body_stream is set, we just kill the connection because we can't do anything with it. We may as well allow a future request to try to reconnect.	John Arbash Meinel	14 years ago
4797.85.25	_send_request can't handle retrying during a read failure _send_request can't handle retrying during a read failure	John Arbash Meinel	14 years ago
4797.85.24	Show that we try to retry the request if the first attempt f... Show that we try to retry the request if the first attempt fails.	John Arbash Meinel	14 years ago
4797.85.23	Start direct testing of the reconnection attempts. Start direct testing of the reconnection attempts. Turns out that SSHConnection.close() also needs to close the read/write handles. It is that way in bzr.dev, just didn't notice it. Pipes notice that they are closed right away, sockets don't.	John Arbash Meinel	14 years ago
4797.85.22	Pull out portable_socket_pair. Pull out portable_socket_pair. Add direct tests for SmartClientStreamMedium.reset.	John Arbash Meinel	14 years ago
4797.85.21	Finally, we have something that can at least handle simple r... Finally, we have something that can at least handle simple reconnects. I've tested that it works to 'bzr branch qbzr' locally, and interrupt the server with SIGHUP during the streaming-of-content phase. After which the next request notices that the ssh process is no longer talking to us, and we reconnect to make the next request. That all works well. Though if I SIGHUP during get_parent_map_rpc, I've seen it fail during the 'read_response' rather than the write phase. But a decent first step.	John Arbash Meinel	14 years ago
4797.85.20	Pull in the code from bzr.dev that changes SmartSSHClientMed... Pull in the code from bzr.dev that changes SmartSSHClientMedium to proxy to SmartSimplePipeStreamMedium, rather than re-implementing the logic. This makes it a lot easier to just test PipeStream and not worry that the logic isn't being used in the SSH version. Currently fails to retry because the current request is still active.	John Arbash Meinel	14 years ago
4797.85.19	It helps if we actually pass the arguments to the helper fun... It helps if we actually pass the arguments to the helper function.	John Arbash Meinel	14 years ago
4797.85.18	Start trying to reconnect to the server if we get a Connecti... Start trying to reconnect to the server if we get a ConnectionReset during the transmission.	John Arbash Meinel	14 years ago
4797.85.17	Re-enable one of the subprocess tests. Re-enable one of the subprocess tests.	John Arbash Meinel	14 years ago
4797.85.16	We don't actually care to import msvcrt always. We don't actually care to import msvcrt always.	John Arbash Meinel	14 years ago
4797.85.15	Do the same test using simple pipes Do the same test using simple pipes	John Arbash Meinel	14 years ago
4797.85.14	Test how _read_bytes works. Test how _read_bytes works. It appears that when the process dies, you just get EOF (empty string) from read, which is nice, given that writing gives problems.	John Arbash Meinel	14 years ago
4797.85.13	Document the behavior of flush. Document the behavior of flush.	John Arbash Meinel	14 years ago
4797.85.12	Try to test _flush. Try to test _flush. On Windows, with buffered files, things hang, so don't do that. With an unbuffered pipe, we seem to get the same results as subprocess.Popen, which is that handle.flush() is a no-op. (Even though the reader is closed by that point, and we haven't read the bites that were already written, flush doesn't raise an error.)	John Arbash Meinel	14 years ago

Older »