- Make flush() block less, by waiting only for "accept" into the
streaming queue rather than "completion" (i.e. ACK from the far end).
This prevents complete window stalls when flushing, and should help performance
of apps that use flush(), like i2psnark (and SAM?).
close() still does a flush that waits for completion, as i2ptunnel
doesn't like a fast return from close().
- flush/close javadocs and comments
* i2ptunnel:
- Now that streaming flush() is fixed, use it in IRCClient, and
for initial data in I2PTunnelRunner, to avoid the 250 ms
passive flush delay
* Base64: comment out some unused methods
* Remove huge whitespace in CryptoConstants
* ElGamalAESEngine:
- Reduce rates
- Check number of tags earlier
Enter b32 or b64 hash or dest into list box, and
check enable for whitelist. Uncheck enable and enter
i2cp.enableBlackList=true in advanced i2cp options for
blacklist. Todo: make black/whitelists radio buttons.
Fix the window size increment logic so it does it much more often.
The code increased the window size by MSS * MSS / N, like
in RFC 2581, but it did it only once every N,
so that was like MSS * MSS / N**2.
Now do it all the time, except for isolated packets like keepalives
that aren't using more than one message of the window.
Seems to speed up outbound significantly, without any
noticable increase in stream.sendsBeforeAck.
- Detect and drop dup SYNs rather than create
a duplicate connection - will hopefully fix
"Received a syn with the wrong IDs"
- Send reset for a SYN ACK with the wrong IDs
- Don't send a reset to a null dest
- Logging tweaks
- Cleanups
- When an "immediate" ack is requested, do it within
250 ms (was 2000)
- Request immediate acks when < 1/3 of window remains,
or when < 3 packets remain in window,
and every 8 packets (was when < 2 packets in window remain)
- Change requested delay to RTT/2 (was RTO/2)
- Log cleanup and javadoc
- New I2PSocketEepGet fetches through existing tunnels
rather than through the proxy
- Use new eepget for i2psnark
- Add a fake user agent for non-proxied fetches
- Cleanups
- Fix BufferedStatsLog so it works at all
- Don't instantiate BufferedStatsLog unless stats.logFilters
property is defined (restart now required to enable logging)
This eliminates the StatLogWriter thread and a decent
amount of memory.
- Move two CLI classes out of the lib
- Commment out places where getStatLog() isn't checked for null
- Cleanups
* Code Janitor time! Many fixes and documenting fixes that should be
done in the future. for the most part, this is a general code cleanup.
* On smaller/embedded systems, the "final" keyword cleanups will have
more of an impact than on larger systems.
* Document missing hashCode() methods.
* Unhide more variables to make code easier to read.
- small pedantic fix
* streaming:
- Fix a deadly race condition.
- Some small pedantic fixes.
* core:
- Fix a deadly race condition.
* BOB:
- Fixed some races that occured from fixing races in streaming and core.
- Some badly needed code refactoring to depend less on the database.
* General cleanup on streaming and ministreaming.
This fixes some compile warnings, and prepares for a larger fix.
There is no code-flow changes, just lint. One warning remains as I am
unsure exactly how to solve the problem yet.
* more BOB fixes, complete with warnings when things go wrong, and
success messages when things turn around and go right. Terminates
early so that applications wait no more than 10 seconds or so.
* Reversed a few earlier patches that caused some odd behavior.
* Changed some core println()'s to debugging messages.
* More BOB threadgroup fixes, plus debug dump when things go wrong.
* Fixes to streaminglib, I2CP, which are related to the TG problem.
* JavaDocs fixups.
