- 05 Mar, 2010 6 commits
-
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1891 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1890 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
It is possible to have a transactional undo of object creation without having the object in cache. For some reason, it happens in 6.9GB FileStorage import. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1889 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1888 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
This makes Packet.__repr__ mad when used to trace what the test is doing. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1887 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1886 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
- 01 Mar, 2010 12 commits
-
-
Grégory Wisniewski authored
Remove MT* tests as then are just copy/paste of non-MT tests. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1885 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
Merge both into 'connector' only to simplify API. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1884 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
Just to avoid complains from pylint :) git-svn-id: https://svn.erp5.org/repos/neo/trunk@1883 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
Master nodes don't exchange node informations during election, this packet handler is not implemented and so unexpected. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1882 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1881 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1880 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1879 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1878 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
This fixes the inability to undo T-1 after undoing T. For reference, this is what is done in RelStorage. It is not as restrictive as FileStorage, but implementing the same support would/will take more time. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1877 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1876 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1875 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
The main purpose is to apply and handler after all pending requests are satisfied and allow process answer packets of current state out of order. This fix a bug that appears when a storage tries to get a lock on an object that is already held by a previous transaction. Is this case the store is delayed causing the answer packet be sent out of order (packet sequence breakage), so unexpected from client's point of view. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1874 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
- 27 Feb, 2010 2 commits
-
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1873 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1872 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
- 26 Feb, 2010 9 commits
-
-
Vincent Pelletier authored
Strangely, nothing in FileStorage forces a new transaction to start with a TID higher than previous transaction. Supporting this is required for a ZODB test to pass, so just do that - with a warning. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1871 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
If EPOLLERR or EPOLLHUP is asserted, the event will be appended to readable_fd_list, so it is useless to test their value in the second test too. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1870 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1869 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
This should improve scalability when the number of non-client nodes grows. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1868 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
The __undoLog() method is used by undoLog() and transactionLog() with different parameters. When a recursive call is made, keep initial arguments instead of always call as if it was undoLog(). git-svn-id: https://svn.erp5.org/repos/neo/trunk@1867 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
When called, queued events were using the last peer message id instead of having the good one on their connection when recalled, so answers made from queued events was using the wrong message id. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1866 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
This is not fixed, IdleEvent is not associated to the packet that create it. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1865 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1864 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1863 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
- 25 Feb, 2010 5 commits
-
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1862 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1861 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
This fixes undo support in client without changing the way replication is implemented. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1860 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Store does not handle conflicts (for reliability, it can only be done once all store responses are received), so this parameter must go away. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1859 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Aborting might happen even before vote is called, wot there might be pending store responses. Wait for them without doing conflict handling. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1858 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
- 24 Feb, 2010 5 commits
-
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1857 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1856 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1855 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
If anything went wrong, an exception must be raised before this point. And if everything has been fine, oid_list and data_dict.keys() are the same. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1854 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/trunk@1853 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
- 23 Feb, 2010 1 commit
-
-
Vincent Pelletier authored
Reduces the amount of duplicated (or deductible) information in packet generators and parsers. git-svn-id: https://svn.erp5.org/repos/neo/trunk@1852 71dcc9de-d417-0410-9af5-da40c76e7ee4
-