Details
-
Type:
Bug
-
Status: Closed
-
Priority:
Blocker
-
Resolution: Fixed
-
Affects Version/s: 3.2.1
-
Fix Version/s: 3.3.0
-
Component/s: None
-
Labels:None
Description
This is the Master Ticket for file transfer bugs.
For reference, these are the links to the changes by Henning:
http://community.igniterealtime.org/message/200840#200840
http://community.igniterealtime.org/message/202593#202593
It was committed as rev 11815 and rev. 11822
The code of Henning was extensively reviewed by Tim Jentz and tested by Walter Ebeling. Several changes of the community were combined and additional changes were made to make file transfer via bytestream, bytestream via proxy and IBB working in Spark 2.6.3 again.
Especially IBB file transfer needed to change from IQ stanzas to message stanzas to allow IBB again. This is breaks the attempted full implementation of XEP-0047 that is not working without errors when using IQ Stanzas.
The patch combined fixes for a couple of sub tasks to this master ticket. Namely
smack-263
smack-335
smack-349
smack-334
Activity
Please provide von unit and integration tests to the patch
Following the discussion in http://community.igniterealtime.org/blogs/ignite/2011/09/02/smack-322-beta-fixes-ibb-transfer#comments I would ask to change the patch to keep list-single. The impact is "only" a breakage of IBB file transfers for Spark 2.5.8 client. Users of 2.5.8 can do file transfer via bytestream. It's a clearly visible issue in mixed environment and has to be documented in the release notes of Spark 2.7.0
reworked patch with list-single instead of list-multi as discussed in http://community.igniterealtime.org/message/216450#216450
includes patch for smack-346
smack-351.patch is the latest version of the patch
When Receiver of Bytestream answer the disc#info message, two messages are send.
The first message is the normal result message, the second message is the message which includes the needed information for the sender.
If the second message arrives to late, the sender decides that the receiver cant use bytestream and sends data in ibb.
This patch adds a 200ms sleep, so the sender has enough time to decide.
It's currently a bit quick&dirty, i will rework but for now, its good for testing purposes.
If a user try to run two clients on the same machine with local proxy, the second client can't establish a socket to the same port the first client is using (7777 per default).
If the default port is less then zero, smack will try to find a port which is not used. If the default port is above zero, smack want to use the specific port which can throw a IO Exception.
This patch make smack to search for a free port if the default port cant be used.
For future note, patches belong with the specific issue that they resolve, not the parent task. In addition, most comments are relevant to specific issues and also belong there.
Please review the patch and consider a commit.