Topic: RAM-Reduction & Backup Center Testing (version 0.89.99.16) - page 18. (Read 41296 times)

Yes, I installed dependencies:


This is the result of my 'make' command (I tried with no edits to the Makefile):



Python is in /usr/ correctly I think:


and this is the result of running

$ cd /usr
$ find . | grep libpython2


Sorry for so much code in a single post!

It's not obvious.  And I'm frustrated that the Makefile is still so terrible.  I have gone to great lengths to make sure Armory is easy to compile on all *nix systems, and it is ... once you get the Makefile worked out.  I really need a better Makefile, but my Makefile skills are crap. 

What you posted looks right.  Did you install all the dependencies first?  See here:

http://bitcoinarmory.com/download/building-armory-from-source/


The "python-dev" pacakge should put libpython2.7.a into the /usr/lib/python2.7/config-x86_64-linux-gnu/ directory.  If it's not there, please post the output of: 

Did you try it without the changes?  I didn't need to make any, I think I'm using 13.04 xubuntu.

I am trying to 'make' the Makefile in Ubuntu 13.04, but I have found this instruction from you Etotheipi:


What exactly does this mean? My makefile looks like this:


Which I believe does not have any comments on line LIBRARYOPTS, and also already has the line STATICPYTHON blank, so do I need to change this file before running 'make'?

Sorry if this is obvious.

There is still a problem where the app occasionally starts up like that and is filling the log with "Marking Orphan Chain" messages, but it looks like it's doing something or getting ready to do something.  This is one of the big bugs I have to fix before I can put out a real release.

After you've done the initial sync with Bitcoin-Qt, and Armory built the database, the most it should take is 10-20 minutes to do a rescan.  And if you've loaded it before and closed it in a good state, it won't even need to do that.  And when it is rescanning, it will tell you that it is.

Moral of the story:  if it stays at 1% for more than a couple minutes, it's hosed.  But restarting seems to fix it. 

I also have problems with this latest version. I have Win7 64 4 gb ram and some quadcore processor. I installed and ran. The blockchain downloaded fine. The scan took forever and I finaly ended with aborting the scan after 14h. Progress bar was still showing 1%. With version 0.89.99.3 The scan was finished after 7h or maybe even earlier (i went to bed so I don't know exactly). Any sugestions? Should I have deleted some files or something before starting?

Most likely an msvc11 build. Long story short, the tool used to build Armory for Windows doesn't natively creates .exe that can be run under WinXP. You'd have to select the specific build tool for XP to build XP compatible Armory. The task is simple, but that version of the binary is most likely not bundled in the package you got. You should ask for a WinXP build or build it yourself, although you'll need Win Vista or above to build it, since msvc11 needs .Net Framework 4.0, which is for Win 6.0 and above only

I downloaded the windows msi file which is ArmorySetup-0.89.99.5-beta_win32.exe

Ran it. Installed. Clicked OK OK, Next, Etc. Clicked "Run Armory". Then I get this:


Contents of that file:


My system specs:
Windows XP SP3 (almost fresh install) 32 bit. (pro or corp, if that matters, not home)
2 gig ram
Intel Pentium T4400

Bitcoin-Qt is running in the background, version 0.8.5
Bitcoin-Qt is also not "installed", I run it from a different drive with -datadir. The blockchain data is in another drive.

*edit* I tried doing -datadir and -offline (there are also shortcuts after installation) and both did not work. I can't even run it offline mode.

No, he is right.  The mathematics may be (almost) the same, but armory signing and bitcoin-qt signing are not compatible.  He is working on it, but unsurprisingly getting the RAM problems solved has much higher priority.

What do you mean?  As far as I know there's only one way to sign something.

I just installed 0.89.99.5-beta, and was a little disappointed to see that the Message Signing functionality is only compatible with other versions of Armory. Will there ever be an Armory/Bitcoin-Qt cross-compatible signing mechanism?

I have finally produced a script that creates a stand-alone OS X app of Armory.  Not a single line of higuys' script has been used (but I use his Info.plist), and I do many things in a different way.  Nevertheless, this is based on his work and it would have taken me at least ten times longer to do it without his work to build on! 

Some comments:

* I have only tested it slightly, and almost exclusively on Testnet.

* The building script is written in Python.  Perhaps that will make it easier for etotheipi to maintain it; it certainly made it easier for me to write it.   

* The app is 64-bit only.  I doubt that Armory can fit into a 32-bit memory model.

* Higuys' script delegated most of the compiling to Homebrew, and copied the result into the App.  This script compiles things directly (although mostly the same way as Homebrew would have done it).  It looks like this solves the problem with the App crashing on Macs that are older than the one used to build the App.  Homebrew assumes that compiled code will be executed on the local machine, and tries to get a bit of extra performance by enabling the full instruction set of the actual CPU.  The usual building scripts are more conservative, and build for all 64-bit CPUs.

* The app claims to work on OS X 10.6 or later, but in reality I have no idea if it works on 10.6 and 10.7.  I suspect that if it declared 10.8 or later that alone would prevent it from working on older versions.

* The app is huge, 136 M.  This is because it contains the majority of a Python installation (mostly .pyc files, but still 65M) and a large part of Qt (51M).  I suspect I can shave off another 25-50 M.

* Command line arguments are ignored (so even if you know how to pass command line arguments to OS X apps, you cannot ask Armory to use e.g. testnet).  This is because I cannot just pass on command line arguments to the Python script as OS X adds its own extra arguments.

I will send this script by PM to etotheipi.  I hereby donate the script to the Armory project and transfer copyright etc.

Looks like this fixed the crashing on exit bug on OS X!

Wow, I missed all the action. I will begin testing this week. I still use Win XP SP3 32 bit with 2 gig of ram. Two old laptops. And dozens of new desktops with 4 gig ram but still old XP on them for work. (I run a clothes business, no need for new OS.)

I do have an updated blockchain of bitcoin-qt.

Just upped to the latest. Quit OK now.

I have a similar problem. The building databases stage takes forever (about an hour or so), it doesn't bsod when I quit but on the restart the preparing/building databases stage starts all over again. Also similar specs - win 7 64 bit, i5-2500k, 12 g of ram. Logs don't indicate anything to be wrong either.

Nevermind, problem on my end, works as intended. Awesome stuff!

(This is Andy)

Alan, Glad to take some of the grunt work off of your hands.

Feel free to let me know any feedback you have about the windows installer.

Just pushed 0.89.99.5-testing.  Also updated the download link for Windows, which is now a NSIS installer courtesy of my new employee (Andy)   Try it out!  I did sign the .exe with my SSL Class 2 cert, but the signature is no shown when you double-click the .exe to run it.  I need to figure out why that is...  but you can still see it by right-clicking, going to "Properties" and then "Digital Signatures". 

By the way, the process of building the installer is now integrated into the MSVS build system, so every time I "Rebuild All", I get a properly named "ArmorySetup_.exe" file.  This will dramatically improve my release process.  Thanks Andy!

As for code, this is really just an efficiency update for Windows, with a few tweaks.  "goatpig" has been awesome: he built me a completely native port of LevelDB, which seems to be totally absent everywhere else on the internet... because it's really hard.  Please test it especially if you were testing before.  That way you can compare speed for me, and know what did and did not work before. 

Also updated the progress meter to give better feedback while building and scanning.  And a few tiny things mentioned on this thread over the past few days.

Meh.  It's not really a bug.  The backwards compatibility of the new backup features led to me breaking up the versions of the backups based on various factors.  Basically, the wallet is version 1.35, and the type of backup is "c" (vs "a" vs "").  I can see it being slightly confusing, but it's really because the wallet code hasn't changed, only the backup features did.

A little bug?