[Qt] rescan progress #3928

cozz commented at 3:50 PM on March 21, 2014: contributor

Fixes #3832. If -rescan on startup, show progress in splashscreen, else show progress dialog.

Needed to implement a temporary notifications queue, otherwise the notifications get queued and block the ui from updating the progress dialog.

Tested linux only qt4 and qt5.

rescan2

rescan

in src/wallet.cpp:None in e638653ad4 outdated

 841 | +            if (pindex->nHeight % 100 == 0)
 842 | +            {
 843 | +                // show progress on splashscreen, if -rescan on startup
 844 | +                uiInterface.SplashMessage(_("Rescanning...") + strprintf("%d", (int)(Checkpoints::GuessVerificationProgress(pindex) * 100)) + "%");
 845 | +
 846 | +                // update progress dialog in GUI

laanwj commented at 4:14 PM on March 21, 2014:

Thanks for tackling this!

Nit: We should use the same ShowProgress in the splash screen, and avoid SplashMessage (which burdens the interface as it's basically just a copy of InitMessage) and the resulting duplication here.

laanwj commented at 4:49 PM on March 21, 2014:

I thought about that a bit and it will indeed give an implementation difficulty: how to get the SplashScreen to subscribe to the wallet in Init to get updates? Maybe add a uiInterface signal to announce a new CWallet to the GUI? Seems something we'll need eventually for MultiWallet support anyway.

in src/qt/walletmodel.h:None in e638653ad4 outdated

 210 | @@ -210,6 +211,7 @@ class WalletModel : public QObject
 211 |      EncryptionStatus cachedEncryptionStatus;
 212 |      int cachedNumBlocks;
 213 |  
 214 | +    QProgressDialog *progressDialog;

laanwj commented at 4:46 PM on March 21, 2014:

This dialog should be in the wallet view, not the model :)

cozz commented at 11:19 PM on March 21, 2014: contributor

update:

removed SplashMessage signal, now only 1 signal in CWallet::ScanForWalletTransactions
added LoadWallet signal to uiInterface
moved QProgressDialog to walletview

laanwj commented at 10:29 AM on March 23, 2014: member

ACK on code changes Still need to test

laanwj commented at 11:34 AM on March 24, 2014: member

Testing: With a fairly new wallet it goes from 1% to 2% then is finished. My guess is that progress reporting is not correct when not starting from the beginning?

cozz commented at 12:36 PM on March 24, 2014: contributor

I assume you have only downloaded 2% of the chain? In this case, rescan is finished after 2% yes. If not started from the beginning it should go to, lets say 98% pretty fast, and then scan the other 2% (assuming you are 100% synced).

#3696 introduced Checkpoints::GuessVerificationProgress for calculating the progress. so I just used it and didnt think about it any further. Checkpoints::GuessVerificationProgress takes fSigcheckVerificationFactor into account and shows the progress of syncing. It does not always work for rescan.

So I suggest writing a Checkpoints::GuessRescanProgress instead: Similar to GuessVerificationProgress, but:

consider that we may not be 100% synced
consider that rescan can skip blocks from the beginning
ignore fSigcheckVerificationFactor, but still consider that early blocks are faster to rescan, because less transactions

Or extend Checkpoints::GuessVerificationProgress with optional parameters, if possible.

sipa commented at 12:39 PM on March 24, 2014: member

For rscan progress, I think you can just use (GuessProgress(current) - GuessProgress(start)) / (GuessProgress(tip) - GuessProgress(start))?

laanwj commented at 12:42 PM on March 24, 2014: member

I have downloaded the entire chain. It just doesn't start the rescan from the beginning because the wallet birthdate is recent.

cozz commented at 1:23 PM on March 24, 2014: contributor

@sipa great idea, thanks. @laanwj To be honest, not sure how rescan can be finished after 2%, if you are synced. Even if blocks are skipped at the beginning, GuessVerificationProgress doesnt care about that.

Looks like this needs better testing of all the edge cases. I will

add an optional boolean parameter fIgnoreSigchecks to GuessVerificationProgress
implement @sipa formula
do some more testing and push an update

laanwj commented at 1:35 PM on March 24, 2014: member

@cozz I see. Yes, I have the full blockchain but it had started a reindex. So it makes sense that it stops at 2%, because it only had come that far with reindexing. False alarm.

cozz commented at 1:41 PM on March 24, 2014: contributor

@laanwj Ok. Still makes sense to consider skipped blocks at beginning or end for progress.

laanwj commented at 9:53 AM on March 28, 2014: member

@cozz you're going to do those changes inside this pull request?

cozz commented at 12:06 AM on March 30, 2014: contributor

update:

added bool fSigchecks to Checkpoints::GuessVerificationProgress(..)
consider dPrgressStart and dPrgressTip in progress calculation in CWallet::ScanForWalletTransactions(..)

sipa commented at 3:08 PM on March 30, 2014: member

ACK core changes; didn't test.

[Qt] rescan progress 392783697c

in src/qt/splashscreen.cpp:None in 311480666c outdated

   4 | @@ -5,8 +5,10 @@
   5 |  #include "splashscreen.h"
   6 |  
   7 |  #include "clientversion.h"
   8 | +#include "init.h"
   9 |  #include "ui_interface.h"
  10 |  #include "util.h"
  11 | +#include "wallet.h"

laanwj commented at 6:47 AM on March 31, 2014:

This include needs to be in #ifdef ENABLE_WALLET

cozz commented at 1:51 AM on April 2, 2014: contributor

update:

added #ifdef around #include "wallet.h"

BitcoinPullTester commented at 4:12 AM on April 2, 2014: none

Automatic sanity-testing: PASSED, see http://jenkins.bluematt.me/pull-tester/392783697c21a0c4cf3db6b0946d3d44d7fed537 for binaries and test log. This test script verifies pulls every time they are updated. It, however, dies sometimes and fails to test properly. If you are waiting on a test, please check timestamps to verify that the test.log is moving at http://jenkins.bluematt.me/pull-tester/current/ Contact BlueMatt on freenode if something looks broken.

laanwj commented at 9:46 AM on April 3, 2014: member

Works for me. Thanks for implementing this @cozz

laanwj referenced this in commit 9bd1bdd08c on Apr 3, 2014

laanwj merged this on Apr 3, 2014

laanwj closed this on Apr 3, 2014

Diapolo commented at 8:57 AM on April 4, 2014: none

A comment, when I supply -rescan with Qt we can click the splashscreen and it fades (as it's an own thread now)... this will result in a tray-icon but NO further user feedback during re-scan. Perhaps it should show the progress via tray pop-up also?

Btw. is there a RPC command to trigger a rescan during normal Qt runtime?

laanwj commented at 9:34 AM on April 4, 2014: member

Well if you don't want to see the splash screen, you also don't see the progress. I think that makes sense. I suppose a way to bring back the splash screen would be nice (for some reason it doesn't just minimize?) (see also #3282).

Some OSes (Windows http://www.codeproject.com/Articles/37451/Display-Progress-and-Overlay-Icons-for-Multiple-Vi , Ubuntu https://wiki.ubuntu.com/Unity/LauncherAPI ) have a way to show progress for an application in the tray, that would makes sense in this case. However, this was very OS specific last time I looked at it, and would pull in new dependencies, so I didn't bother with it.

Not sure what you mean, but just please do not abuse notification balloons for this.

BTW: not that I know of, I'm triggering rescans by generating a private key on the testnet bitaddress.org and importing it.

Diapolo commented at 9:49 AM on April 4, 2014: none

When you click the splashscreen it's gone, no way to bring it back currently. Would be nice if a click on the tray icon could re-activate it.

rebroad commented at 11:42 PM on May 15, 2014: contributor

Given that ShowProgress() could also be used to show block verification progress in the splash screen, is it ideally placed in the wallet related cpp files?

laanwj commented at 7:24 AM on May 16, 2014: member

It is used in the splash screen to display progress. The methods are on the wallet because the progress is wallet-specific. In theory multiple wallets could be rescanning in parallel.

DrahtBot locked this on Sep 8, 2021