So, I just pasted the transaction details into blockchain.info/decode-tx and it came out correct. I also realized that armory had the correct PrevTxHash as input, but the sender address was still wrong. FYI: blockchains tool doesn't list the sender address, just the previous tx hash. I then broadcasted it, and it immediately showed up correctly in armory.

So I know the setup works, and I have my private keys, but it would be nice to know whether the sending issues were mine or the software. I did use RBF if that matters.


*Edit.
To keep from spamming too much, I'll just edit this post. I tried again with the same result. Online machine running 0.96.3 on windows didn't accept the transaction file as signed.

When I upgraded the online computer back to 0.96.3.992 and loaded the transaction it recognizes it as valid. I initially had the same problem with syncing, and it didn't work starting bitcond manually either. What did work was running bitcoin-qt before starting armory. It synced, recognized the transaction, and broadcast it successfully.

Hopefully this will help someone else someday. Thanks for help and patience.

I can't remember if it did or not, but I did check task manager and closed it manually if it did. This has solved some issues before, so I 'm aware of it.

Anyway, I just tried again and this time I was able to open the "broadcast raw transaction" option, paste the raw data, and get it parsed. As the outputs looked correct, I tried broadcasting, but nothing happened. I checked the parsed tx data again, and noticed that the sender address was not the one I had specified. Instead of the one with spendable UTXOs, it was one that shows up empty on blockchain.info. When I copy in the "signed" transaction data again to the online system, it still provides the correct transaction details, except for the fact that it comes in unsigned.

I really don't understand this...

***I don't think it should matter, but the offline signing wallet was made from a secure printed fragmentet backup in order to test it. The online watch-only wallet was created before I deleted and regenerated my wallet.

So I followed your advise, and installing the offline bundle (I used 0.92.2) on Ubuntu 14.04 worked perfectly, and I was able to upgrade to current version without any additional dependencies or tweeks. Thanks!

My online system was initially running on windows 10, with core 0.15.1 and armory 0.96.3, which I knew ran well, and received transactions. I then upgraded the online version to 0.96.3.992, and sent a test transaction to the watch only wallet I had saved on it. This led some issues as others have also described. First, armory stopped syncing, and despite saying it was online, it was frozen multiple blocks behind without. I only noticed this when it didn't pick up my test transaction, as 0.96.3 had did. After restarting it and rescanning, I quit armory and started core, allowing it to sync. Then starting armory and rescan/rebuilding, it still didn't pick up the transaction despite it now being confirmed within the block height armory had synced. It should be noted armory still didn't sync blocks newer than I had downloaded with core. I reverted back to 0.96.3 and it immediately worked, so it was not really a problem. Let me know if/what logs you would like to look at, and I can PM them.

To complete the test, I then created a transaction online, copied it to a usb, and imported to my offline signer. Everything looked ok, so I signed and it automatically saved it back on the USB. The transaction block became about twice as before signing. Importing it to the online computer, it claimed the transaction was unsigned. It could still read the details of the transaction. I tried again, and this time manually saving a text file of the signed transaction as well, but neither worked.

I tried a third time saving the signed transaction data as a hex text file, and electrum recognized it as signed and offered to broadcast (obviously, it didn't recognize armorys transaction format). I did not broadcast in order to check further. I then tried armory again, but it still would not accept the transaction as signed, nor did it recognize the raw tx data in "Offline Transaction". I then went tools -> broadcast raw transaction, prompting* the following error message:

"Not Online
Bitcoin Core is not available, so Armory will not be able to broadcast any transactions for you."

*This happened before I got to paste the tx data

At this point armory says it is connected and online, and has synced all new blocks on its own.

Although I haven't done it yet, it seems I can still broadcast the transaction using another service, so its more of an inconvenience than a big issue for me personally. Any ideas to what the issue is? I can also pm you the signed and unsigned armory data if you want it, but it seems like the problem lies with the online computer.

Not sure I follow regarding the offline bundle. The reason I asked was because of a post saying an offline release compatible with old CPUs was expected in october. I interpreted this as no offline bundle can be installed on old CPUs. Are you saying the older versions are compatible, so I can install 0.93 and upgrade?

I did try following the bitzuma guide with its 11 dependencies, but had already installed ubuntu 16.04. Both this and the offline bundle was lacking python-qt4. I grabbed that one, but it also failed due to dependencies and I read there were quite a lot of them for 16.04.

Anyways I'll try again on 14.04 both with aforementioned guide and a gcc4.7 release, and if it fails with an old offline package.

Thanks again for your help.

Not sure where to ask this question so I'm trying here:

I have tried installing Armory on an offline computer using Ubuntu 16.04, but failed due to the missing dependencies a couple other people have mentioned. I saw using 14.04 would help, but based on the age of my CPU, I guess I would stop at the same point as rothbart in the 0.96.2 release thread.

I'm sure you have better things to do, but are there any plans on making an offline bundle compatible with shitty old CPUs?

Thanks for all your work!

Not sure if this is worth even worth a mention, but I was playing around with fragmented backups to test and found something.

I figured I would try to make a backup with a mix of everything. First paper, w/o secureprint, then paper w secureprint, then file w/o secureprint and finally file with securerint. As this was only a test, I didn't take a very close look other than seeing the wallet name (obviously) and the secureprint code staying the same after toggling the latter on/off. After saving and printing the backups, I deleted the wallet and tried to restore them. When I did, the message "If any of the fragments require a SecurePrint code, you will only have to enter it once" (emphasis mine) made me confident the mix would not be a problem.

When I realized none of the backups fragment matched what was stated on the paper/file I realized something was wrong and went back to create a new one. Toggling secureprint on/off shows that this creates a completely new set. However, 3/4 fragmented backups I made have the same fragment-ID (not the 4th), and all 4 obviously the same wallet ID (this was with 0.96.3).

This was just me playing around, and I would have been a lot more careful before depositing anything. Still, there could perhaps be a warning that:
1. you can't mix and match secureprint
2. each time you toggle on/off you create a new set of fragments.

With 0.96.3.99, none of the fragments got the same ID, but the SecurePrint code stayed the same through each toggle.

Based on the initial text of the previous SSS being deterministic, I'm guessing this wasn't a problem before this fix was made?

--