summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* Adding a forgotten type specMatthew Sackman2010-04-211-1/+2
|
* Merging default into bug 21673Matthew Sackman2010-04-211-0/+4
|\
| * merge bug22628 into defaultMatthias Radestock2010-04-201-53/+66
| |\
| | * Flatten out a tupleMatthew Sackman2010-04-201-4/+4
| | |
| | * Correct list comprehensionMatthew Sackman2010-04-201-1/+1
| | |
| | * Use a sequence ID when combining entries into the snapshot. This can then be ↵Matthew Sackman2010-04-201-51/+64
| | | | | | | | | | | | used to correctly sort the messages and remove the dependence on increasing guids
| * | cosmeticMatthias Radestock2010-04-201-2/+1
| |/
| * merge headsTony Garnock-Jones2010-04-201-0/+4
| |\
| | * Have a default config file for testing with make runTony Garnock-Jones2010-04-201-0/+4
| | |
* | | Switched the From to a Fun and hence CPSMatthew Sackman2010-04-204-16/+20
| | |
* | | Lots of fixes for handling of acktags, which was badly broken in some placesMatthew Sackman2010-04-191-37/+35
| | |
* | | WhoopsMatthew Sackman2010-04-191-1/+1
| | |
* | | Sort out handling of acktags in the queue_process. Use a set throughout - ↵Matthew Sackman2010-04-191-33/+28
| | | | | | | | | | | | this avoids the unpleasant lists:usort on every ack (which is what ordsets:from_list does). Also, don't bother with the intersection - just assume that we're only given acks we really know about, which the channel should be able to guarantee. This results in a performance improvement from MulticastMain -s 0 -r 12750 to MulticastMain -s 0 -r 13900 : i.e. 9%
* | | Beef up assertions on combine deltas, but it doesn't appear there was a bug ↵Matthew Sackman2010-04-191-9/+19
| | | | | | | | | | | | here anyway
* | | Switch four megs to a fraction of the file size limitMatthew Sackman2010-04-183-6/+6
| | |
* | | Correct mistake in indexing into pairs of GC candidatesMatthew Sackman2010-04-181-2/+3
| | |
* | | Make use of fhc:delete now that we have itMatthew Sackman2010-04-181-4/+2
| | |
* | | Another missing infinity on a callMatthew Sackman2010-04-181-1/+1
| | |
* | | A cosmetic and a missing infinity timeout on a callMatthew Sackman2010-04-182-3/+2
| | |
* | | Rewrote scanning of message files, to read in blocks of up to 4MB (hence ↵Matthew Sackman2010-04-182-58/+54
| | | | | | | | | | | | bounded) but can still cope with variable sized files. The advantage here is vastly reduced number of OS calls to position and read. The results is that in tests, GC time is reduced from around 35 seconds to about 2. Code is also 7 lines shorter - and arguably simpler - the only tricky bit is reusing left over data from one block with the next read block, hence different read and scan offsets
* | | merge bug22616 into bug21673Matthias Radestock2010-04-160-0/+0
|\ \ \
| * \ \ merge default into bug22616Matthias Radestock2010-04-163-31/+25
| |\ \ \ | | |/ /
| | * | tweak tx API on amqqueueMatthias Radestock2010-04-163-31/+25
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | commit_all and rollback_all take the channel pid as an additional arg. This brings these functions in line with deliver and ack, which also take both a txn and ch_pid. In the queue process this saves us some book keeping. Also, in the queue process we need to clear the txn field in the ch record on commit/rollback, since otherwise a subsequent channel 'DOWN' results in some suprising and unnecessary, though perfectly safe, control flow. Finally, there is no need to check that a commit relates to a channel the queue process knows about - this is always guaranteed to be the case. All these changes were cherry-picked from the bug21673 branch and bring default closer to that branch.
* | | | On commit we must know about the channelMatthew Sackman2010-04-161-9/+6
| | | |
* | | | set channel txn to none on rollbackMatthew Sackman2010-04-161-1/+2
| | | |
* | | | cosmeticsMatthew Sackman2010-04-161-4/+3
| | | |
* | | | Merging headsMatthew Sackman2010-04-160-0/+0
|\ \ \ \
| * \ \ \ merge bug22616 into bug21673Matthias Radestock2010-04-160-0/+0
| |\ \ \ \ | | |/ / /
| | * | | add memory_monitorMatthias Radestock2010-04-162-0/+300
| | | | |
* | | | | cosmeticMatthew Sackman2010-04-161-2/+1
|/ / / /
* | | | Merging headsMatthew Sackman2010-04-162-3/+5
|\ \ \ \
| * \ \ \ merge bug22616 into bug21673Matthias Radestock2010-04-160-0/+0
| |\ \ \ \ | | |/ / /
| | * | | oopsMatthias Radestock2010-04-161-0/+917
| | | | |
| * | | | merge bug22616 into bug21673Matthias Radestock2010-04-160-0/+0
| |\ \ \ \ | | |/ / /
| | * | | qa supervisor2 and its use in amqqueue/amqqueue_supMatthias Radestock2010-04-161-5/+7
| | |/ / | | | | | | | | | | | | | | | | | | | | Queue processes are now terminated nicely (and the supervisor waits for the termination), but this won't make any difference until queue processes trap exits and do something interesting on shutdown.
| * | | merge default into bug21673Matthias Radestock2010-04-162-3/+5
| |\ \ \ | | |/ /
| | * | minor refactoringMatthias Radestock2010-04-162-2/+5
| | | |
| | * | fix sigs on persister that have been wrong for agesMatthias Radestock2010-04-121-1/+1
| | | |
* | | | Capture more information for the next time the freaky GC error pops upMatthew Sackman2010-04-161-17/+23
|/ / /
* | | Whoops, wrong recordMatthew Sackman2010-04-161-2/+2
| | |
* | | Changes to the way we track messages that are pending acks to ensure that on ↵Matthew Sackman2010-04-161-59/+107
| | | | | | | | | | | | queue deletion we remove everything from the message stores that shouldn't be there
* | | Cope with queue death even though there are in-flight commits going onMatthew Sackman2010-04-163-27/+43
| | |
* | | Updated documentationMatthew Sackman2010-04-161-11/+2
| | |
* | | Merging headsMatthew Sackman2010-04-162-20/+20
|\ \ \
| * | | cosmeticMatthias Radestock2010-04-161-2/+2
| | | |
| * | | cosmetic and minor refactoringMatthias Radestock2010-04-151-18/+18
| | | |
* | | | Thorough reworking of API - the BQ is now responsible for hanging onto ↵Matthew Sackman2010-04-169-559/+526
|/ / / | | | | | | | | | unacked msgs and all details of transactions
* | | tx_commit takes a list of publishes, not a list of guidsMatthew Sackman2010-04-131-1/+1
| | |
* | | Whoops - redeliver was only used by the old persister and was removed from ↵Matthew Sackman2010-04-131-5/+1
| | | | | | | | | | | | _process months ago
* | | queue_index doesn't need to reimplement recursive deleteMatthew Sackman2010-04-131-9/+2
| | |
View Lorry Controller Status