MultiBit

[jim618]

There is a new test release of MultiBit.
The downloads URLs are at:

http://multibit.org/releases.html


Version 0.5.7beta (Encrypted wallets)

Enhancements:
+ Change is now spendable as soon as the transaction traverses the Bitcoin network.
+ Fix for "Could not generate key from password and salt" bug on send from unencrypted wallets.

Scan of release checklist



This fixes the feature of MultiBit that you could not spend change until the next block confirmation.
It would be very useful if a few people could test it out in the next few days before I put the change into the live code.

[1713587064]

1713587064

[1713587064]

1713587064

[Mike Hearn]

Got a link to the patch you made, so I can give it a quick review?

[jim618]

Hi Mike,

That would be very useful cheers.
There are three main changes, all of which I have put into the Wallet (though 3 should prob go into Transaction):

1) Wallet#completeTx now looks through the pending tx if there is not enough bitcoin in the unspent pool:
https://github.com/jim618/multibit/blob/v0.5-bitcoinj-0.7/src/main/java/com/google/bitcoin/core/Wallet.java

2) The Wallet#getBalance(BalanceType) works with another BalanceType (BalanceType.AVAILABLE_WITH_BOOMERANG_CHANGE):
(same URL)
This includes pending change if it has been seen by 2 peers.

3) I would be interested if you could think of a neater way of detecting your own change. In the Wallet I have a method:

transactionSpendsFromThisWalletAndHasBoomerangedBack

which feels a little clunky.

The rest of the changes are things like not allowing you to spend before you are connected to 2 peers (so as to give you a good chance of detecting the tx coming back) and UI updates when I get a onTransactionConfidenceChanged that I presume you are not so interested in.

[freemoney458]

Hi Jim,

Version 0.5.7beta (Encrypted wallets) downloaded and sent some coins to BitMarket.eu, a free bitcoin exchange market who now needs a bit financial help.

I sent coins two times in rapid succession, I did not notice any unexpected restrictions due to change problems.
I can not tell for sure if it was due to the boomerang rule though, because I have about 20 addresses in my wallet and MultiBit used different addresses to send from.
Anyway, I just wanted to tell you that the whole sending experience was very smooth and fast.

Great work!

[freemoney458]

I forgot to mention that in the latest beta version, I noticed that the transaction fee was set by default to 1 BTC.
I think this is way to high and can be really disgusting for novice users.

I am not sure if it is part of the binary download version.
If it is, I suggest you reduce the amount and maybe even add it as a check to your release checklist.

[jim618]

Hi Freemoney,

The transaction fee is stored in your multibit.properties - a default value of 0.001 is only used if it is missing ie it is the very first install. After the first install it picks up the setting you put in the Preferences.

I presume you have your MultiBit set to German ?
Perhaps it is a bug in the number parsing.
'0.001' parses in German as:

Zero thousands and one = 1 BTC

All the stored numbers *should* be stored in UK format but I will double check.

If you still have the '1 BTC' appearing (or can reproduce it) can you have a look in your multibit.properties as to the actual value stored please. It will be something like:

'fee=0.001'
Or
'fee=0,001' (note comma)

It *should* be the first one.  The locale specific number should appear on the UI but NOT in the config file.
As you can imagine it is very fiddly so I will double check the code.

[freemoney458]

Hi Jim,

yes, I have MultiBit set to German.

I looked things up more in detail:

The multibit.properties showed sendFee=1.
After changing in MultiBit to '0,001' BTC MultiBit used the correct fee during sending.
After quitting MultiBit and restarting it, it showed the correct fee on the screen: '0,001' BTC.
The multibit.properties showed sendFee=0.001.

So everything is like expected.

Maybe it was just a historic setting that now turned up when I seriously spent my first coins.
I will have a look at it when I update the next time to a newer MultiBit version.

On another note: I investigated the two rapid succession of 'send' transactions more in detail and found that the second transaction was actually done by spending the change from the first transaction. This was done within seconds after the first transaction.
So your boomerang rule worked out fine. 

[Nachtwind]

Version 0.5.7beta (Encrypted wallets)
Enhancements:
+ Fix for "Could not generate key from password and salt" bug on send from unencrypted wallets.

Confirmed

[jim618]

Hi Jim,

yes, I have MultiBit set to German.

I looked things up more in detail:

The multibit.properties showed sendFee=1.
After changing in MultiBit to '0,001' BTC MultiBit used the correct fee during sending.
After quitting MultiBit and restarting it, it showed the correct fee on the screen: '0,001' BTC.
The multibit.properties showed sendFee=0.001.

So everything is like expected.

Maybe it was just a historic setting that now turned up when I seriously spent my first coins.
I will have a look at it when I update the next time to a newer MultiBit version.

On another note: I investigated the two rapid succession of 'send' transactions more in detail and found that the second transaction was actually done by spending the change from the first transaction. This was done within seconds after the first transaction.
So your boomerang rule worked out fine. 

That is good to hear. The number parsing (especially commas and dots meaning different things in Europe) was fiddly.

I will double check the values on a fresh install just to be sure.

[lenny_]

Hello there,
I am very happy Multibit user and donor But one problem makes me worrying.
Very often I have my transfers won't ever confirm by network. They just stuck on:
"Seen by 1 peer. Not seen in chain."
And no progress at all. I cannot use by wallet, and I have to revert blockchain to earlier date or restore whole %APPDATA%\Multibit folder from recent backup.
Multibit latest version, 0.4.16, Windows 7 AMD64.
Any ideas?

[jim618]

Hi lenny_,

Thanks for your support !

Those 'seen by 1 peer' are the troublesome transactions for sure.
Once they have traversed the network (ie peers seen = 2 or more) you can be pretty sure they will confirm in my experience.

Reasons for the 'seen by 1 peer' that I know of:

+ flakey network (eg I have a 3G network that will sometimes just die). The network layer in bitcoinj (that i use) is pretty good at retransmitting at start up so if I have trouble I normally:
1) shut down MultiBit
2) reconnect my 3G dongle
3) start up MultiBit - it then (silently) retransmits any transactions it had problems sending

+ fees too low
the fee calcs are quite basic so if you have a transaction with many small inputs a bitcoind might not pass it on. Also if you have the fee set to the absolute minimum (0.0001 BTC) I think you will get slow confirms now. The symptoms looks similar.

+ something else to do with the Bitcoin network interaction I am not yet aware of.

Obviously the goal is to get every transaction out into the network 'happily' and confirmed so feedback from you if you think you are encountering one of the first two points above or if you think it is something else would be useful.

[lenny_]

Hello jim618,
Thanks for your reply!
Lets see reasons you suggest:

- slow network - it's not a case, that's wired broadband connection, I am running web server here and some other services without any problems on multiple devices
- fees too low - I tried fees up to 0.05 BTC and still nothing
- shutdown Multibit and try again - I tried it twice, shutdown and start again. I saw status changing several times from "Connecting..." to "Online", but finally it went "Online" and transaction with "1 peer" has not been sended.

Another clues:
- I just successfully sent 15 BTC to external address with 0.001 BTC fee.
- I am trying 5 times already and I cannot sent ~50 BTC from one wallet to another (both addresses in MultiBit, separate wallets). I tried a fee up to 0.05 BTC and no difference.
- This must be something else bugging. Related to my Bitcoins or addresses, or separate wallets, I guess.

Can I send you debug files?

EDIT:
Some more clues, they may be helpful:
- All transactions I am trying to send, with "Seen by 1 peer. Not seen in chain." are not visible in blockchain/blockexplorer. Transaction ID not found.
- I just received transaction from exchange, with 0 Fee. Multibit saying "Seen by 3 peers. Not seen in chain.", but transaction can be viewed in Blockchain.info without problems.

[jim618]

Yes please send me the log files (multibit_debug.log and multibit_console.log in the log directory - or just zip the log directory and send me the previous day's logs too) and email them to Jim at multibit dot org and I will have a look.

Another possibility might be if you have any firewall rules redirecting any port 8333 traffic (to say a bitcoind on a different machine). I don't think it is that as you are connecting ok but you mention you are running a web server so you might have ports locked down. 

[lenny_]

Indeed, ports 8333-8334 are redirected to different machine in my local network. But MultiBit seems to be working OK for months already, except those rare, broken transactions.
Logs sended.

[jim618]

For those of you who like poring over stats, I have put the MultiBit download logs on the multibit.org website.
There is a link at the bottom of the 'Downloads' section on the top right.

There are monthly totals and you can drill down to more detailed reports:

MultiBit download statistics

[jim618]

There is a new live release of MultiBit available at:

http://multibit.org


Version 0.4.17

Enhancements:
+ Change is now spendable as soon as the transaction traverses the Bitcoin network.

Scan of release checklist


This is rolling out the same functionality that was put in the test code on Friday.
It makes quite a difference to the usability of MultiBit in day to day use.

[jim618]

@Lenny

I have been looking through your debug logs and asking on the bitcoinj mailing list.
It is most likely due to the fee calculations in MultiBit/ bitcoinj not being right and hence the fee is too low for that particular transaction.

[lenny_]

@jim618:
Thank you! So it looks like transaction don't even get out from bitcoinj? So why it is reported as sended, and "seen by 1 peer"?
Is there anything I can do about it? What's the fee amount I should include? Can bitcoinj predict what transaction fee should be like?

Today, as always I had problem to send BTC to an address. Minimal Fee was included, but it stuck "seen by 1 peer", as usual.
So I went to BlockChain.info wallet, where I have same addresses added as backup. I manage to send same amount of BTC with 0 Fee.

https://blockchain.info/tx/e559ccd123d0250ee9b44fdf774a0e6eedb60c1f37dd67899446d2d34bd87cd0
Transaction was included in block relayed by OzCoin mining pool straight after 2 minutes.

Regards
lenny

[jim618]

Hi Lenny,

MultiBit sends out it's transactions to a bitcoind (who then relays it to the rest of the Bitcoin network).
Thus 'seen by 1 peer' is normally a result of a transaction being sent to a bitcoind but that node has not relayed it to other nodes. (it can also be due to network failure as I mentioned before). In your logs there were no errors to do with transmitting the transaction.

Bitcoinj (which MultiBit uses) currently does not calculate the fees as per the rather complicated Bitcoin fee calculation rules which is probably where the problem lies. If the node that receives the transaction 'does not like' the transaction then it won't relay it.

I am sure it is on Mike's todo list. There isn't really much the user can do about this at the moment as you cannot work out what the fee should be manually - it depends on how the transaction is constructed. (and how the transaction is constructed affects the fee - it's not trivial to work it out).

In effect, MultiBit/ bitcoinj has to 'persuade' the node it is transmitting to to pass on the transaction but at the moment the calculated fees aren't right.

I think blockchain.info runs bitcoinds actually behind the scenes hence the difference.

Sorry I cannot give you a better answer - it is just not done yet.

[jim618]

At the moment the multibit.org files are hosted in a personal webserver and the binaries are on Amazon S3.

To rationalise things a bit and give some performance headroom I have just ordered a dedicated server:
http://www.server4you.net/root-server/server-details.php?products=1

It will be housed in a datacentre in Strasbourg that has multiple fibre optic connections out. I figure the limiting factor will be the machine's 100 Mbit/s network adaptor.


The multibit download is about 26 MB so at 100% load that could provide:

100 * 1024 * 1024 * 3600 * 24 / (26 * 1024 * 1024 * 8 )  > 40,000 downloads a day.

Seeing as how downloads are about 200 a day now there is plenty of room for growth !

It will be a little while before everything gets moved over.