@adamsmith That's good to know. I will test Zotero 7 with VoiceOver then as soon as it's released and report accessibility issues if applicable. Nice to hear that things are getting better

Test that out with a fresh document with a couple of citations and copying over some text. Does that work?

you need to wait longer. Otherwise, you need to test this with the data directory on your local disk. After moving the data directory, you can reset the data directory location from Advanced → Files

zotero(3)(+0000003): Proxy not required for https://zoteroproxycheck.s3.amazonaws.com/test

The purpose of that test was to copy the full path (from Zutilo) where Zotero *thinks* a linked file is located, and then try to open the file at that path outside Zotero ... which would test if it is

The test procedure is a bit extreme, but I still report the findings in case anything useful can be identified here. It does not sound right to see Zotero saturate the RAM more than 10 minutes after t

Can you test this out in a fresh document?

I've just submitted a fix for that. Can you give this a test ride?

Here is my test PDF: https://drive.google.com/file/d/1ZYVbLiVCAsZ-D0ZHbRMbnYk9QnD_nK9i/view?usp=sharing , and this is my recorded gif: https://drive.google.com/file/d/1iSHgN5uM4Zpxu21fT20q-e08BqsNCsw-

Yes, most likely. (And Troubleshooting Mode actually reduces PDF performance, so a better test would be disabling all plugins manually and restarting Zotero.)

I mean, if you follow the instructions to rename zotero.sqlite and then run sqlite3 zotero.sqlite < dump.sql, you'll either end up with a zotero.sqlite file or not, and if a file is created it will

You should use the Translator Editor in Zotero to create and test your translator, and you should follow the translator documentation and other translators to write your code.

I just tested and located my code error, and I couldn't reply to your message in a timely manner. I'm very sorry.

the browsers we support. You should always test on a supported browser before posting.)

So this feed would be a good test case if the feature can be added in the future.

I have also made the test in a VM on the Ubuntu 24.04 workstation:

I have tested on Ubuntu 24.04 LTS on a separate recent powerful workstation with the same library.

It takes for me about 1 second to completely to load the document (10.1007/s00348-023-03759-9) with text layer. Therefore I can't reproduce the issue so far. Can you test on Linux or macOS as well?

The hyphens problem is less important, but that paper gives a good number of test cases, with hyphens that should be either kept or removed.

I downloaded the beta version and completed the human test, and now it works.