Bitcoin-Qt / bitcoind version 0.8.0 release candidate 1

[deepceleron]

The checkpoint for this release is 216116, the current block is 220944. Full verification is done on block contents after the checkpoint, so it is expected to run slower on the last 4000 blocks.

I just finished an upgrade of a previously synchronized Bitcoin 0.7.2 on Win7, it took 70 minutes from startup to "up to date".

The Win32 client reports being built with Qt 4.8.3, the current release is v4.8.4, and has a good many fixes.

[1715354968]

1715354968

[Mike Hearn]

Dan, I think there might be something wrong with your hard disk. Check /var/log/messages .... it shouldn't take a day to index 4000 blocks. Also if a hard disk is dying, getting very slow is one of the first symptoms.

[dansmith]

Here's the debug log's tail:

...
LoadBlockIndex(): transaction index disabled
LoadBlockIndex(): hashBestChain=0000000000000290e9c5ca7f9c4fb014c4f9f639dcffa6821c4616b7762509f7  height=220932 date=2013-02-13 09:02:08
init message: Verifying block database integrity...
Verifying last 288 blocks at level 3
LevelDB read failure: IO error: /home/default/.bitcoin/chainstate/029837.sst: No such file or directory


************************
EXCEPTION: 13leveldb_error       
Database I/O error       
bitcoin in Runaway exception       

Is there any way to recover from this error?
I double-checked that there's plenty of disk space left.

[deepceleron]

Here's the debug log's tail:

...
LoadBlockIndex(): transaction index disabled
LoadBlockIndex(): hashBestChain=0000000000000290e9c5ca7f9c4fb014c4f9f639dcffa6821c4616b7762509f7  height=220932 date=2013-02-13 09:02:08
init message: Verifying block database integrity...
Verifying last 288 blocks at level 3
LevelDB read failure: IO error: /home/default/.bitcoin/chainstate/029837.sst: No such file or directory


************************
EXCEPTION: 13leveldb_error       
Database I/O error       
bitcoin in Runaway exception       

Is there any way to recover from this error?
I double-checked that there's plenty of disk space left.

I have a feeling that the tech support answer for this will be "remove the index directories and reindex".

I would try fsck, and then look in lost+found for a 0-2MB file. Move it to the datadir as 029837.sst. See if Bitcoin doesn't stop crashing.

[kgo]

Working fine on a Macbook Pro OSX Snow Leopard.

[saddambitcoin]

upgraded on OSX 10.6.8, rescanning blocks took about 8 hrs but it seems to load faster now. connects to more peers than 0.7.2, before I would often be stuck at 8.

[nibor]

Ran with:

bitcoin-qt.exe -datadir=E:\bitcoin_database -dbcache=4000 -logtimestamps -txindex=1 -reindex=1

On i7 with 4 core+HT with 12gig ram on a nice new ssd on win7 64bit.

Took 30mins till started checking sigs. Then took another 30mins for last 4000ish blocks.


Used max of about 1.5-1.7gig.
Never read from disk for the chainstate - only wrote .logs
Only re-orged the blocks/index a few times.

Was entirely CPU bound. Initially on one thread then on all threads whilst checking sigs (ran at about 80% of the 8 virtual core busy which with hyperthreading is probably ideal).

Did a bit of checking in log file and are checking 1200 trans/second.

[sharky112065]

Upgrade process went great for me. Took 23 minutes (Debian Squeeze, 32GB RAM, 8 Core CPU, and Hardware Raid 6). Shut it down, backed up data, removed blk* files, and restarted. No problems thus far. P2pool seems to like it just fine.

[dooglus]

Before the upgrade, I had:

Code:

  -rw-------  1 chris chris 2097307549 Aug  8  2012 blk0001.dat
  -rw-------  1 chris chris 2097177246 Dec  4 18:48 blk0002.dat
  -rw-------  1 chris chris 1420134352 Feb 12 11:52 blk0003.dat

After the upgrade, I have:

Code:

  -rw-------  2 chris chris 2097307549 Aug  8  2012 blocks/blk00000.dat
  -rw-------  2 chris chris 2097177246 Dec  4 18:48 blocks/blk00001.dat
  -rw-------  2 chris chris 1420134352 Feb 12 11:52 blocks/blk00002.dat
  -rw-------  1 chris chris   50331648 Feb 13 19:21 blocks/blk00003.dat

ie. each file's number went down by one, and a new file was started even though the previous one (was #3, now #2) didn't get anywhere near to 2GB.

Is that normal?

[foo]

ie. each file's number went down by one, and a new file was started even though the previous one (was #3, now #2) didn't get anywhere near to 2GB.

Is that normal?

Yes. The new code creates 128 MB block files, but it will keep the old ones as is.

[tHash]

Around block 188,000, bitcoin threw an error that it "could not write block" and exited.   Continuing to download the blockchain without issue after a restart.

This is a fresh install on Windows 7 x64.

[grue]

Fully syncs up without a problem on Windows 7 x64. Haven't done any transactions on it though.

[fornit]

using win7 64bit.

during sync the client crashed twice claiming the db is corrupted. first time restarting the client helped, second time i had to reboot. now its running smooth.

[ArticMine]

...

Improvements
------------

Mac and Windows binaries are signed with certificates owned by the Bitcoin
Foundation, to be compatible with the new security features in OSX 10.8 and
Windows 8.

...

What happens if these certificates are revoked? Could this lead to a Treacherous Computing, http://www.gnu.org/philosophy/can-you-trust.html, attack against Bitcoin?

[grue]

Mac and Windows binaries are signed with certificates owned by the Bitcoin
Foundation, to be compatible with the new security features in OSX 10.8 and
Windows 8.

are these signatures only visible in windows 8? because i'm looking at the binary and it does not have a digital signature.

[Gavin Andresen]

The windows setup.exe is signed, as is the Mac .app bundle. The executables inside them are not signed (I can't think of a good reason to sign them, it would not increase download security at all).

You can also still check the download packages using the SHASUMS.asc file, which is signed with my gpg key.

And if you are running Linux or Windows you could check all of the files in the installer against other core developer's keys.

If the code signing certificate was revoked then we would go back to just using gpg keys. The code signing certificate is great because Windows and OSX know how to check it automatically when the download happens.

[ArticMine]

The windows setup.exe is signed, as is the Mac .app bundle. The executables inside them are not signed (I can't think of a good reason to sign them, it would not increase download security at all).

You can also still check the download packages using the SHASUMS.asc file, which is signed with my gpg key.

And if you are running Linux or Windows you could check all of the files in the installer against other core developer's keys.

If the code signing certificate was revoked then we would go back to just using gpg keys. The code signing certificate is great because Windows and OSX know how to check it automatically when the download happens.

Thank you very much for the clarification. My concern was actually would the .exe would continue to run under Windows 8 if the certificate was revoked after installation? I believe that under Windows XP/Vista/7 the .exe would continue to run, but I am not clear under Windows 8.  The possible danger here is Microsoft being able to shut-down Bitcoin-Qt / bitcoind after installation by revoking the certificate hence the reference to a Treacherous Computing attack.

I did run into a somewhat similar situation back in 2005 where an expired certificate on APC PowerChute not only prevented the software from installing but also caused significant damage to the Windows registry. I spent the better part of a Sunday afternoon fixing the Windows registry in order to bring the server back to health  The OS was Windows Server 2003. Granted this was an installation faliure but given the direction that Microsoft is taking with Windows 8 RT, I would be rather safe than sorry when it comes to Windows 8 on Intel/AMD.

[Mike Hearn]

I think we should be signing the Windows binaries as well. Binary signatures are not only used by browsers and the OS when running setup programs. Anti-virus scanners also use them to build binary reputations and avoid false positives. Bitcoin does things that can look a tiny bit malware like (connecting to P2P networks and exchanging lots of random-looking data), so signing the executables can't hurt.

I don't think Microsoft doing an Apple and trying to ban unsigned code is going to ever happen, at least not on regular desktop Windows. There are far, far, far too many legacy/in-house apps out there that are business critical and yet never been signed, never will be signed.

[grue]

The windows setup.exe is signed, as is the Mac .app bundle. The executables inside them are not signed (I can't think of a good reason to sign them, it would not increase download security at all).

You can also still check the download packages using the SHASUMS.asc file, which is signed with my gpg key.

And if you are running Linux or Windows you could check all of the files in the installer against other core developer's keys.

If the code signing certificate was revoked then we would go back to just using gpg keys. The code signing certificate is great because Windows and OSX know how to check it automatically when the download happens.

I don't use the setup because I want to avoid going through tedious steps with the installer. most software makers sign both their installer AND their binary, so I don't see why bitcoin should be any different.

[dansmith]

I again received:

************************
EXCEPTION: 13leveldb_error       
Database I/O error       
bitcoin in Runaway exception 

after I made a hard reset of my machine.
Harddisk is not to blame, since last time I got this error, I fsck'ed my ext4 root partition from a LiveCD and it came out clean.

(Good job I backed-up index,chainstate, and database dirs, so I restored them without having to reindex like I did last time).