I downloaded ebooks (1 in English, 1 in Hindi) with several different fonts selected in both EPUB and PDF formats.

I can no longer reproduce issues 1 and 2 from the description. I did not test issue 3 because apparently it has already been fixed.

If we are not going to be doing UI tallying in the future, we can safely close this ticket.

Most of the testing of this has been recorded in https://www.mediawiki.org/wiki/Anti-Harassment_Tools/SecurePoll_Improvements/Test_Results and other tickets.

@Niharika @phuedx A couple of few questions about which blocks we should count (I think related to work done in T270318 but that has already been resolved):

• We only count expired blocks as "past" blocks. Blocks which have been explicitly unblocked are not counted. For example, this IP has been blocked 3 times but only counts 1 past block (you'll need to open the network tab of your browser tools to see the response).
• We don't seem to count range blocks for "active" or "past" blocks. For example, this IP has a range block but IPInfo lists no active blocks.
• I couldn't get System Blocks to be counted (including $wgProxyList and $wgSoftBlockRanges, tested locally).

The bugs in the description and T289347#7299771 are no longer reproducible.

We now display numbers as small as 0.000001 (see round 13 of https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1691).

I rechecked all the examples in the description, all the stackoverflow moderator elections on beta and the Irish general election locally, concentrating on the numbers over 1000.

According to logstash, I see that these errors stopped on https://vote.wikimedia.org on 1st Sept at ~17:21, which I think corresponds to when the train deployed to votewiki.

Further testing of this will be done in T290000 and T290134.

After clicking "Create tally", you either see the results immediately or the This election is being (re-)tallied. Please check back later for the results message. You can then refresh the page without it trying to resubmit the tally.

QA has been using this to test the tallies on beta.

@STran Could you check @phuedx's comment in https://gerrit.wikimedia.org/r/c/mediawiki/extensions/SecurePoll/+/715196/4#message-14424b9e3a217a9baed0d1078b8c2a45212122e9?

Also raised T290134, numbers less than 0.0001 appear as 0.

In T290000#7322418, @Niharika wrote:

In T290000#7321307, @dom_walden wrote:

This does not affect the way results appear when they are written to the database or dump file.

Here is an example of how it looks when you extract the results from the dump file and pretty print them: https://www.mediawiki.org/wiki/Anti-Harassment_Tools/SecurePoll_Improvements/Test_Results/5_2_5000_956342267/dump

@dom_walden does this happen when the result is written to the database and the Tally page reads it from the database? Ref T289186

In T290027#7321300, @STran wrote:

I tested this against an updated definition of surplus (using number of votes * number of tied candidates + surplus) and we still see 9-12 eliminated at the same time https://gerrit.wikimedia.org/r/c/mediawiki/extensions/SecurePoll/+/715711/4

This does not affect the way results appear when they are written to the database or dump file.

Further notes and examples.

There is already a patch for this: https://gerrit.wikimedia.org/r/c/mediawiki/extensions/SecurePoll/+/715196

I raised T290000 (woo!) to track T288740#7311757.

We decided this wasn't important enough to spend anymore time on this. No one else could reproduce this.

@STran I don't think it likes numbers over 1000.

In T288748#7307176, @TThoabala wrote:

@dom_walden does the file you are using have gpg-sign-key ? after removing it I am getting an error though not exactly the same but I think it might be related

@STran The issue I mentioned in standup.

In T288748#7285931, @STran wrote:

At least it sounds like cleanup even if error is now working? 😅 If it's replicable, you can set an early return in GpgCrypt->cleanup() and that should give you a folder to investigate?

@Niharika We should check with T&S whether this behaviour is expected or not.

In T289186#7303044, @jrbs wrote:

In T289186#7297453, @dom_walden wrote:

After this change, the election results will be available in the dump file after the CLI tally has been run. I don't know whether or not you want people to be able to see this.

To clarify, your concern is that the results are potentially visible between decryption and posting? ...

I can no longer reproduce the bug in T280798#7214861 on beta.

In T289347#7298747, @Niharika wrote:

@STran helped me understand this better. We discussed in Slack about the possibility of adding BC and I am hesitant about us doing that at this point. We should go with whatever PHP's default is. If candidates 9, 10, 11 and 12 all achieved the same votes in a round according to PHP, we should eliminate them all. If 9 got more votes (by however much), we should not eliminate them.
We can probably not display this well in the UI but we can attempt to document this in all the necessary places and inform the election admins about this possibility.

In T289347#7298620, @gerritbot wrote:

Change 713963 had a related patch set uploaded (by STran; author: STran):

[mediawiki/extensions/SecurePoll@master] Fix the elimination check in the STV algo

https://gerrit.wikimedia.org/r/713963

@STran Checking out this change locally, I am not sure I am seeing the correct results.

In T289186#7295379, @jrbs wrote:

In T289186#7294114, @dom_walden wrote:

@Niharika @jrbs Just a note, I believe once the results are written to the database they will be visible in the dump file.

If the election does not have the voter-privacy property set, this dump file is visible to anyone.

Before we do this ticket, we might want to decide whether we want to:

1. Remove the results from the dump file (like was done with the private key in T288924), or
2. Rely on the election admins to appropriately set the voter-privacy property when creating an election

I think this is by design - the dump file is intended to be publicly available, iirc. Is there any sensitive information in the dump file beyond the private key?

@Niharika @jrbs Just a note, I believe once the results are written to the database they will be visible in the dump file.

In T288748#7282704, @STran wrote:

You might be able to investigate the temp folder (/tmp/securepoll-55dbaf2a56a8c7b2f947a4eaca87897c8c9fa39a/) if the election failed and didn't clean up after itself. You should see a key and an input file and maybe cating one of those would provide more data? It sounds like it might be a bad input somewhere if it took a long time (presumably successfully decrypting other inputs) to fail.

In T288748#7279715, @TThoabala wrote:

I ran this on local and the script completed without errors
Script finished in 402 s

• local docker
• gpg (GnuPG) 2.1.18
• libgcrypt 1.7.6-beta

We tested this on test.wikipedia.org.

In T287549#7274569, @MusikAnimal wrote:

QA notes: This hopefully should be easy to test. When you edit a page and add a link to one of the pages listed at Special:DisambiguationPages, it should have a tag next to it in the revision history that reads "Disambiguation links".

In T277222#7272875, @Niharika wrote:

In T277222#7271730, @dom_walden wrote:

I started 4 tallying jobs one after the other, but they all appear to be stuck:

• https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1427
• https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1449
• https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1471
• https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1493

(they are all unencrypted, 5000 votes).

I started them while https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/632 still had tally-job-enqueued=1 in the DB (tally-ongoing was not set). I have deleted that row from the DB but the jobs still appear stuck.

@dom_walden do they also fail to tally if we try to tally them from the cli?

@Zabe Further to your comment https://gerrit.wikimedia.org/r/c/mediawiki/extensions/SecurePoll/+/709533/1#message-dd40e9955064e53f6d6eaaea03533aaebed5d6e3, will this patch actually change anything? Are you planning a follow up patch? Thanks.

With the editor open on https://en.wikisource.org, the sidebar does not contain the "Download/print" options. I tested both Source and Visual editor, on every skin.

We list the Wikisource languages in their own language, or in the user's interface language if the language cannot be found on Wikidata.

I have tested this very briefly on https://test.wikipedia.org (e.g. https://test.wikipedia.org/w/index.php?title=Main_Page&action=historysubmit&type=revision&diff=470140&oldid=448939).

I started 4 tallying jobs one after the other, but they all appear to be stuck:

• https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1427
• https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1449
• https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1471
• https://vote.wikimedia.beta.wmflabs.org/wiki/Special:SecurePoll/tally/1493

(they are all unencrypted, 5000 votes).

I have been able to tally elections from the CLI on beta, including encrypted elections with 1000 votes and (small) encrypted STV elections.

For greater visibility, I am moving this back into In Progress.

For comparison, this election is also encrypted but only has 1 vote, and it finished almost immediately: https://vote.wikimedia.org/wiki/Special:SecurePoll/tally/1055

For now, I think this is good enough. I may raise another bug in the future, but not right now.

Testing on beta, I can now set the admins via the CLI and the election is immediately tallyable.

In T287738#7261592, @Niharika wrote:

@dom_walden Does it work for you now?

In the source editor, if I click the "syntax highlighting" button a row gets written to the user_properties table in the database to reflect the CodeMirror feature being enabled for my user. When I click it again, the row is removed.

In T287782#7251613, @Reedy wrote:

@dom_walden Want to try again please? :)

In T287780#7251693, @Zabe wrote:

I had the same error, but on a second try I was able to create https://test.wikipedia.org/wiki/SecurePoll:828.

@Reedy @Zabe @Majavah Trying to create an election with "All wiki" selected on votewiki and testwiki, get some sort of timeout:

[19e3148a-a7d9-4c40-9aa2-bf5e49d774e1] /wiki/Special:SecurePoll/create Wikimedia\Rdbms\DBTransactionSizeError: Transaction spent 4.2640490531921 second(s) in writes, exceeding the limit of 3

@Reedy Trying to submit a translation on testwiki, error:

[8fbb6604-7aca-43ee-bf6a-0ed067cb9c08] /w/index.php?title=Special:SecurePoll/translate/821/fr&action=submit TypeError: Argument 1 passed to MediaWiki\Storage\PageUpdater::saveRevision() must be an instance of CommentStoreComment, instance of MediaWiki\Extensions\SecurePoll\SecurePollContent given, called in /srv/mediawiki/php-1.37.0-wmf.16/extensions/SecurePoll/includes/Pages/TranslatePage.php on line 336

In T287780#7250375, @Reedy wrote:

Leaving open for now until fix is tested in WMF prod