Fixes #3019 - Better import from LO / OO / soffice #3783

JohnMcLear · 2020-03-29T12:10:37Z

Discussion here: #3019

I don't like the return in contentcollector.js -- I'm +1 a better method.

This fixes copy/paste and import too, in theory. Not fully tested.

Once OO is working test abiword to see if it will take the same Filter for the html value.

sdf

update

Updating

This change is meant to ease using LibreOffice as converter. When LibreOffice converts a file, it adds some classes to the <title> tag. This is a quick & dirty way of matching the <title> and comment it out independently on the classes that are set on it.

…mporting

muxator · 2020-03-30T00:50:28Z

Crashes the import process. Etherpas itself survives. No logs.

See #3019 (comment).

JohnMcLear · 2020-03-30T09:14:17Z

Ugh, libreoffice get this wrong.

If libreoffice isn't fully installed IE you only install libreoffice-common it fails.

sudo apt-get install libreoffice

is required.

What should happen is instead of Returning Error: source file could not be loaded Libreoffice should explain why it's not loading and how to remedy. Quite disappointed!

JohnMcLear · 2020-03-30T09:38:35Z

When using NodeJS spawn

Error in option: --convert-to html:"XHTML Writer File:UTF8"

When using CLI

soffice --headless --convert-to html:"XHTML Writer File:UTF8" --outdir . /home/jose/testOneLiner.odt --invisible --nologo --nolockcheck --writer

Works

JohnMcLear · 2020-03-30T10:08:03Z

I dunno tbh, this fix is fine as far how it's designed, the problem is with nodejs not properly passing the arguments to spawn

I tried hardcoding and shell:true to no avail..

      var soffice = spawn(settings.soffice, [
        '--headless',
        '--invisible',
        '--nologo',
        '--nolockcheck',
        '--writer',
        '--convert-to', 'html:"XHTML Writer File:UTF8"',
        task.srcFile,
        '--outdir', tmpDir
      ],
        {
          shell : true
        }
      );

@muxator can you take a look and sanity check my work on this? If you set shell to false you can see that spawn isn't passing the correct command to soffice

JohnMcLear · 2020-04-01T16:19:46Z

Bump @muxator for help :D

muxator · 2020-04-04T21:48:08Z

@JohnMcLear I confirm your problem.

When invoked directly on the shell this command works flawlessly:

libreoffice6.4 --headless --convert-to html:"XHTML Writer File:UTF8" --outdir . /tmp/oneline.odt --invisible --nologo --nolockcheck --writer
convert /tmp/oneline.odt -> /tmp/oneline.html using filter : XHTML Writer File:UTF8

The same command wrapped in a spawn() call in pure Javascript, without Etherpad being involved at all, fails:

// file: test-libreoffice.js
const { spawn } = require('child_process');

//const ls = spawn('ls', ['-lh', '/usr']);
const ls = spawn('libreoffice6.4', [
  '--headless',
  '--convert-to',
  'html:"XHTML Writer File:UTF8"',
  '--outdir',
  '.',
  '--invisible',
  '--nologo',
  '--nolockcheck',
  '--writer',
  '/tmp/oneline.odt',
]);

ls.stdout.on('data', (data) => {
  console.log(`stdout: ${data}`);
});

ls.stderr.on('data', (data) => {
  console.log(`stderr: ${data}`);
});

ls.on('close', (code) => {
  console.log(`child process exited with code ${code}`);
});

Result:

$ node test-libreoffice.js 
convert /tmp/oneline.odt -> /tmp/oneline.html using filter : "XHTML Writer File:UTF8"
Error: Please verify input parameters... (SfxBaseModel::impl_store <file:///tmp/oneline.html> failed: 0x81a(Error Area:Io Class:Parameter Code:26))
child process exited with code 0

Tested on nodejs 12.13.1.

I'll try to profile the LibreOffice execution.

JohnMcLear · 2020-04-04T21:50:22Z

I'm glad you sanity checked this, glad it's not me being an idiot! Thanks! Please let me know if you solve and how, this one pissed me off good :D

muxator · 2020-04-04T22:12:48Z

This is not nodejs specific.

The same program above, converted in python, gives the exact same result:

#!/usr/bin/env python3
#
# file: /tmp/test.py

import subprocess

args = [
  'libreoffice6.4',
  '--headless',
  '--convert-to',
  'html:"XHTML Writer File:UTF8"',
  '--outdir',
  '.',
  '/tmp/oneline.odt',
  '--invisible',
  '--nologo',
  '--nolockcheck',
  '--writer',
]

result = subprocess.run(args, capture_output=True)

print('stdout:', result.stdout)
print('stderr:', result.stderr)

When executing:

$ ./test.py 
stdout: b'convert /tmp/oneline.odt -> /tmp/oneline.html using filter : "XHTML Writer File:UTF8"\n'
stderr: b'Warning: failed to read path from javaldx\nError: Please verify input parameters... (SfxBaseModel::impl_store <file:///tmp/oneline.html> failed: 0x81a(Error Area:Io Class:Parameter Code:26))\n'

muxator · 2020-04-04T22:32:54Z

Aaaand this is the reason for which everyone always needs to have a good diff tool at hand.

Here's the difference between the output of the succeeding (directly on the shell) and the failing (js via child_process.spawn) execution:

-convert /tmp/oneline.odt -> /tmp/oneline.html using filter : XHTML Writer File:UTF8
+convert /tmp/oneline.odt -> /tmp/oneline.html using filter : "XHTML Writer File:UTF8"

Do you see that in the js (& python) execution there are two double quotes more?

Now, in hindsight, it is easy to understand: the double quotes are needed only when executing via the shell, because the shell needs escaping for spaces.

# bash requires the double quotes BECAUSE of the spaces, to convey
# that "html:XHTML Writer File:UTF8" is really a single parameter
#
--convert-to "html:XHTML Writer File:UTF8"

When we explicitly pass parameters to a subprocess via execve() (which is what all these library calls are doing in the end), there is no need to escape anything, because the parameters are already cleanly separated among the array elements.

Leave out the double quotes and it should work.

On my toy Javascript & python programs it worked.

JohnMcLear · 2020-04-04T22:37:32Z

woop nice will add to my list of things to do 2mrw. Thanks!

muxator · 2020-04-04T22:39:08Z

After looking at this snippet I think that, we might be having one more problem.

LibreOffice exits with a 0 exit code even when the conversion fails. It did when we were passing a bad parameter.

If Etherpad just checks for the process exit code it may mistake a failed conversion for a successful one.

We should check for something else in addition.

the process stdout/stderr? That's fragile
that the converted file was really created where it was supposed to be <-- I would trust this more.

JohnMcLear · 2020-04-04T22:42:01Z

Fix spawn command to not use "
Ensure output file exists with fs.fileExists lookup.

JohnMcLear · 2020-04-05T15:06:25Z

Lol I just looked and I already WAS leaving out the double quotes btw :D

  if(type === "html") type = "html:XHTML Writer File:UTF8" // note how I'm not escaping the "

    queue.push({"srcFile": srcFile, "destFile": destFile, "type": type, "...

 var soffice = spawn(settings.soffice, [
        '--headless',
        '--invisible',
        '--nologo',
        '--nolockcheck',
        '--writer',
        '--convert-to', task.type,
        task.srcFile,

gonna test in isolation now to generate some logic that works/fails.

The below works:

var spawn = require("child_process").spawn;

var soffice = spawn("/usr/bin/soffice", [
  '--headless',
  '--invisible',
  '--nologo',
  '--nolockcheck',
  '--writer',
  '--convert-to', 'html:XHTML Writer File:UTF8',
  '/home/jose/test.doc',
  '--outdir', '/home/jose/'
]);

JohnMcLear · 2020-04-05T15:26:49Z

Yea, I'm fucking stumped.

Sanity check for me..

Checkout develop && specify soffice in settings.json
Import a word file, it works.
Edit src/node/utils/LibreOffice.js Insert if (type === 'html') type = "html:XHTML Writer File:UTF8"; @ Line 43
Restart Etherpad
Import a word file, it fails.

Yet running this isolated:

var soffice = spawn("/usr/bin/soffice", [
  '--headless',
  '--invisible',
  '--nologo',
  '--nolockcheck',
  '--writer',
  '--convert-to', 'html:XHTML Writer File:UTF8',
  '/home/jose/test.doc',
  '--outdir', '/home/jose/'
]);

Works fine...

Can you confirm? Am I losing my mind?

muxator · 2020-04-06T20:19:23Z

Sanity check for me..
[...]

I am going to try it now.

muxator · 2020-04-06T20:41:13Z

Confirmed the issue.

What nodejs version are you running? Is it 12?

Edit: the error happens on both node 10 and 12, but on 12 it is harder to spot

muxator · 2020-04-06T20:52:49Z

I have a suspect.

To see what happens, you'll have to run Etherpad under nodejs 10 (on nodejs 12 a deprecated feature was removed and you won't see anything useful in the logs). See #3834 and #3841.

The type parameter somehow ends up in the filename of a temporary file we generate, or we think we generated (a brittle choice in the first place 😄)).

When type === 'html', the temporary file name is /tmp/upload_XXX.html.

When type === 'html:XHTML Writer File:UTF8', this is what is generated in the logs:

[2020-04-06 22:39:31.270] [WARN] console - Converting Error: { [Error: ENOENT: no such file or directory, rename '/tmp/upload_f3354ec47b4d3989f15cd7bbaedd81de.html:XHTML Writer File:UTF8' -> '/tmp/etherpad_import_1137536363.html']
  errno: -2,
  code: 'ENOENT',
  syscall: 'rename',
  path:
 '/tmp/upload_f3354ec47b4d3989f15cd7bbaedd81de.html:XHTML Writer File:UTF8',
  dest: '/tmp/etherpad_import_1137536363.html' }
Error: ENOENT: no such file or directory, rename '/tmp/upload_f3354ec47b4d3989f15cd7bbaedd81de.html:XHTML Writer File:UTF8' -> '/tmp/etherpad_import_1137536363.html'

Moreover, in the next line, you'll see why we need to integrate Rewrite the customError module with class syntax #3841 (or upgrade log4js), and why this is not visible under nodejs 12:

[2020-04-06 22:39:31.274] [ERROR] console - (node:13895) [DEP0079] DeprecationWarning: Custom inspection function on Objects via .inspect() is deprecated

muxator · 2020-04-06T20:59:25Z

I think that the line that needs to be changed is this one:

etherpad-lite/src/node/utils/LibreOffice.js

Line 105 in e625168

    
           var sourceFilename = filename.substr(0, filename.lastIndexOf('.')) + '.' + task.type;

But I really would like to see this become more robust (without encoding information in the temporary file name).

Some more monkey patching on that file would be almost immoral. 😭

JohnMcLear · 2020-04-06T20:59:41Z

How the funk did I not see that?!

I'm running v13.10.1 and that might explain why but wow, what a crazy situation tho..

Solution, pass the padId+somerandomnonce() as the output filename string?

JohnMcLear · 2020-04-06T21:03:22Z

Yea, agreed. This whole import/export pass to third party binaries is a mess. Gonna fix this, make it stable then will do a good review for future versions.

muxator · 2020-04-06T21:11:06Z

Solution, pass the padId+somerandomnonce() as the output filename string?

You mean, ask libreoffice to generate a converted file with a name we control? Something like that, if LibreOffice's command line allows it.

I would prefer that.
or mess with task.type here:

etherpad-lite/src/node/utils/LibreOffice.js

Line 105 in e625168

var sourceFilename = filename.substr(0, filename.lastIndexOf('.')) + '.' + task.type;

Removing everything after the first ":", if exists.

But it stinks.

In the next commit, we are going to change the conversion method to "html:XHTML Writer File:UTF8". Without this change, that conversion method name would end up in the extension of the temporary file that is created as an intermediate step. In this way, the file extensione will always stay ".html". No functional changes, hopefully. Only the extension of the temporary file should change.

This yields better conversion results, but requires the previous change, otherwise there would have been difficulties in locating the temporary file name.

JohnMcLear · 2020-04-07T10:45:40Z

Right, let's get this moving. Please merge this PR as it is now and then I can keep moving forward with changes. I'm going to hit problems with contentcollector.js and my list work really soon!

muxator · 2020-04-08T20:51:29Z

Merged with some reordering of the commits.

Thanks.

JohnMcLear added 7 commits March 15, 2020 09:53

Merge pull request #2 from ether/develop

9ea5ad9

sdf

Merge pull request #3 from ether/develop

c696165

update

Merge pull request #4 from ether/develop

a59d48c

Updating

Merge branch 'develop' of github.com:ether/etherpad-lite into develop

0cf1692

Merge branch 'develop' of github.com:ether/etherpad-lite into develop

f5a47db

contentcollector: remove weird stuff LibreOffice adds to DOM before i…

1aabb06

…mporting

JohnMcLear requested a review from muxator March 29, 2020 12:10

muxator force-pushed the 3019 branch from 7b76861 to f85b48c Compare March 30, 2020 00:31

muxator mentioned this pull request Mar 30, 2020

import from OO document #3019

Closed

JohnMcLear assigned muxator Apr 1, 2020

JohnMcLear changed the title ~~Fixes #3019~~ Fixes #3019 - Better import from LO / OO / soffice - don't import 2 bullets in place of one - Need help Apr 1, 2020

JohnMcLear mentioned this pull request Apr 4, 2020

ImportHandler: instead of failing when a non Ascii code is uploaded do a roundtrip conversion - Is this mergeable on 03/04/2020? Seems buggy #3796

Closed

JohnMcLear assigned JohnMcLear and unassigned muxator Apr 4, 2020

JohnMcLear added 2 commits April 7, 2020 10:44

LibreOffice: use "html:XHTML Writer File:UTF8" export method

2b83227

This yields better conversion results, but requires the previous change, otherwise there would have been difficulties in locating the temporary file name.

JohnMcLear changed the title ~~Fixes #3019 - Better import from LO / OO / soffice - don't import 2 bullets in place of one - Need help~~ [PRIORITY #2] Fixes #3019 - Better import from LO / OO / soffice - READY TO MERGE Apr 7, 2020

muxator force-pushed the 3019 branch 2 times, most recently from 1351c22 to 2b83227 Compare April 8, 2020 20:49

muxator merged commit 08b83ae into ether:develop Apr 8, 2020

muxator changed the title ~~[PRIORITY #2] Fixes #3019 - Better import from LO / OO / soffice - READY TO MERGE~~ Fixes #3019 - Better import from LO / OO / soffice Apr 8, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fixes #3019 - Better import from LO / OO / soffice #3783

Fixes #3019 - Better import from LO / OO / soffice #3783

JohnMcLear commented Mar 29, 2020 •

edited

Loading

muxator commented Mar 30, 2020 •

edited

Loading

JohnMcLear commented Mar 30, 2020

JohnMcLear commented Mar 30, 2020

JohnMcLear commented Mar 30, 2020 •

edited

Loading

JohnMcLear commented Apr 1, 2020

muxator commented Apr 4, 2020

JohnMcLear commented Apr 4, 2020

muxator commented Apr 4, 2020

muxator commented Apr 4, 2020

JohnMcLear commented Apr 4, 2020 •

edited

Loading

muxator commented Apr 4, 2020

JohnMcLear commented Apr 4, 2020

JohnMcLear commented Apr 5, 2020 •

edited

Loading

JohnMcLear commented Apr 5, 2020

muxator commented Apr 6, 2020

muxator commented Apr 6, 2020 •

edited

Loading

muxator commented Apr 6, 2020 •

edited

Loading

muxator commented Apr 6, 2020

JohnMcLear commented Apr 6, 2020

JohnMcLear commented Apr 6, 2020

muxator commented Apr 6, 2020

JohnMcLear commented Apr 7, 2020

muxator commented Apr 8, 2020

Fixes #3019 - Better import from LO / OO / soffice #3783

Fixes #3019 - Better import from LO / OO / soffice #3783

Conversation

JohnMcLear commented Mar 29, 2020 • edited Loading

muxator commented Mar 30, 2020 • edited Loading

JohnMcLear commented Mar 30, 2020

JohnMcLear commented Mar 30, 2020

JohnMcLear commented Mar 30, 2020 • edited Loading

JohnMcLear commented Apr 1, 2020

muxator commented Apr 4, 2020

JohnMcLear commented Apr 4, 2020

muxator commented Apr 4, 2020

muxator commented Apr 4, 2020

JohnMcLear commented Apr 4, 2020 • edited Loading

muxator commented Apr 4, 2020

JohnMcLear commented Apr 4, 2020

JohnMcLear commented Apr 5, 2020 • edited Loading

JohnMcLear commented Apr 5, 2020

muxator commented Apr 6, 2020

muxator commented Apr 6, 2020 • edited Loading

muxator commented Apr 6, 2020 • edited Loading

muxator commented Apr 6, 2020

JohnMcLear commented Apr 6, 2020

JohnMcLear commented Apr 6, 2020

muxator commented Apr 6, 2020

JohnMcLear commented Apr 7, 2020

muxator commented Apr 8, 2020

JohnMcLear commented Mar 29, 2020 •

edited

Loading

muxator commented Mar 30, 2020 •

edited

Loading

JohnMcLear commented Mar 30, 2020 •

edited

Loading

JohnMcLear commented Apr 4, 2020 •

edited

Loading

JohnMcLear commented Apr 5, 2020 •

edited

Loading

muxator commented Apr 6, 2020 •

edited

Loading

muxator commented Apr 6, 2020 •

edited

Loading