Misuse of "Import" erases all history data

[jej]

Describe the bug
I see some closed issues about this, but maybe it worths to reopen it.
I suggested to a friend to use etherpad for a collaboration work. After few days and many modifications, she wanted to import an inline image in the text. She clicked on "Import/Export", picked an image file and didn't care about the warning... Of course the pad was scrambled by a binary text, and the image was not inserted inline as expected by the user. The problem is that now the historic does not show the revisions any more. The revisions are browsable with the slider, but current content only is displayed. All previous revisions seems recorded as usual, but not available.
I tested this on https://framapad.org/fr/ . I don't know what version of etherpad they use.

To Reproduce
Steps to reproduce the behavior:

1. Write/modify some text, in order to create some revisions.
2. Click on 'Import/Export'.
3. Pickup a binary image file (ex. png/jpg)
4. Import it, acknowledge the warning.
5. The binary data replace the text. Eventually: erase it and add some text.
6. Look at the history to see if previous revisions are still displayed.

Expected behavior
It's not a problem if binary data erase the current version, but at least it should be possible to retrieve the previous content of the pad.
Maybe the GUI should be more careful with import. A mandatory checkbox could make the user aware that all the content will be replaced, instead of a simple button.

Server (please complete the following information):

• Etherpad version: ?
• Here: https://framapad.org/

[JohnMcLear]

1. There is a plugin for forking pads to a new state from a point in time. ep_fork maybe? You access the previous revision through the timeslider. Plugins are visible on the plugin list page.

2. Re importing ux, I would say the expected ux for import in any text editor is to replace the whole document no?

[jej]

Thanks for your reply.

1. I didn't know this plugin. I will test for my own use.
2. The problem here is the bug about the timeslider. If the import replace the doc, why the slider allows to browse the revisions but don't show them? I would say that import a doc just create a new revision with the new content. But here the concept of the pad is broken...

[JohnMcLear]

1. Gimme 5 mins will find it for you.
2. Import should create a new revision, you are right! Can you provide exact steps to replicate then here because if you are saying it's not creating a new revision something is very broken! Please provide steps to replicate on a clean install / no plugins etc.

[JohnMcLear]

1. https://www.npmjs.com/package/ep_copypad -- give this a try, I'm not sure why it's not visible in npm / plugin list tho! You would need to install via npm install ep_copypad though as it isn't visible in the plugin list for some reason..

[jej]

To Reproduce
Steps to reproduce the behavior:

1. Write/modify some text, in order to create some revisions.
2. Click on 'Import/Export'.
3. Pickup a binary image file (ex. png/jpg)
4. Import it, acknowledge the warning.
5. The binary data replace the text. Eventually: erase it and add some text.
6. Look at the history to see if previous revisions are still displayed.

[JohnMcLear]

RE replicating btw, I tried to replicate locally but I get an error Import failed: The upload failed, please try again

I get warning [2021-10-02 10:47:01.610] [WARN] ImportHandler - Attempt to import non-ASCII file

My assumption here is you have an instance with a modified settings.json and something like abiword or soffice enabled?

[JohnMcLear]

Please provide STR in a clean environment of Etherpad, not a modified one.

[jej]

Yeah that's the problem, I don't manage the platform Framapad. I guess it's not so modified. My own version is full of plugins...
Do you know a clean fresh install I can try?

[JohnMcLear]

You can spin one up using the docs really quickly/easily, that's probably the best approach to ensure it's clean and we can replicate?

[JohnMcLear]

Just a thought. Image size could be maxing out buffer causing timeslider to not work? An idea, not researched

[jej]

Ok... I can reproduce with a clean install :

1. Write/modify some text, in order to create some revisions.
2. Click on 'Import/Export'.
3. Pickup a HTML file.
4. Import it, acknowledge the warning.
5. The HTML content replace the text. (Eventually: erase it and add some text.)
6. Look at the history to see if previous revisions are still displayed.

In my case, the slider can browse the revisions but nothing is displayed. If I CTRL+SHIFT+R the page, the current content is displayed for any revisions.

[JohnMcLear]

ack, gimme 5 will check

[JohnMcLear]

6. I see previous revisions so unable to replicate. Do you get any errors in console?

[JohnMcLear]

What size is the HTML file contents?

I was able to get curSplice[sline] not populated, actual curSplice contents is Array(3) [ 0, 2, undefined ] . Possibly related to https://github.com/ether/etherpad-lite/issues/2802

I think STR for that is insert content, import empty HTML file, type some stuff, visit timeslider, click play.