Corrupted Microsoft Word files (docx)

[marlemion]

Microsoft Word files cannot be opened after being uploaded to owncloud. MS Office claims that they are damaged.

Reproduce the bug: Drop the file to the server via the web interface and let it sync to the file system via the windows client or upload it directly via the windows client. Downloading directly from the web interface does also result in a corrupted docx file. So apparently ANY DOCX FILE GETS DAMAGED.

The windows client is current 1.8.4, the server 8.1. Apache is at 2.4.16, encryption is not enabled, fastcgi is not installed, no redirection, apache PHP at 5.6. The log file does not show any error message.

<?php
$CONFIG = array (
'instanceid' => '#####',
'passwordsalt' => '#####',
'trusted_domains' =>
array (
0 => '#####',
),
'overwrite.cli.url' => '/owncloud',
'datadirectory' => '/var/www/localhost/htdocs/owncloud/data',
'dbtype' => 'mysql',
'version' => '8.1.0.8',
'dbname' => '#####',
'dbhost' => '#####',
'dbtableprefix' => 'oc_',
'dbuser' => '#####',
'dbpassword' => '#####',
'installed' => true,
'theme' => '',
'maintenance' => false,
'secret' => '#####',
'loglevel' => 0,
);

This is crucial. I already messed up some of my work due to this bug.

[guruz]

Can you upload a sample (correct) file somewhere that can be used for reproducing?

[Dianafg76]

I tried this , and does it well , I can open the .docx file. And if I believe in a .zip several .docx I can see them with no problem.

Desktop v ownCloud-2.0.0.5308-nightly20150729-setup.exe, In WIN
Server v {"installed":true,"maintenance":false,"version":"8.1.0.8","versionstring":"8.1 RC2","edition":"Enterprise"}
Desktop v ownCloud-2.0.0.2576-nightly20150730.pkg

[danimo]

@marlemion This is a problem with your local setup then, not a client bug anyway. Something seems to recompress zip files. Which apps are installed on your owncloud server?

[danimo]

@marlemion Also, if you upload zip files the same way, are they still identical? (same size/hash)

[marlemion]

Apps are pretty much standard: files, activities, calendar, contacts, images

It happens to any zipped MS office file. But I can assure you that it also affects xlsx files. Just messed up a very important Excel file right now. Versioning back does not help. It used to work beforehand. So I could work on already uploaded files without any problem. But if I upload them or change the version, it corrupts the zip container.

Normal zip files work.

md5sum is different after upload as is the size (it is bigger).

[danimo]

@marlemion Normal zip files work. Yes, they may work, but are they also different (hash-wise)?

[danimo]

@DeepDiver1975 @MorrisJobke any idea how to debug what repacks the file?

[marlemion]

Normal zip files remain original, i.e. no change in md5sums.

[guruz]

Any virus scanners or other security solutions running locally?

Are you syncing through a proxy?

Drop the file to the server via the web interface and let it sync to the file system via the windows client or upload it directly via the windows client. Downloading directly from the web interface does also result in a corrupted docx file.

So also docx files uploaded via web and then downloaded via web (no sync client involved) has this issue?

[marlemion]

No virus scanners.

No proxy (it runs on apache in the directory "owncloud" behind a router, which forwards the port).

Yes, the corruption also appears when uploading via the web client (just retested a minute ago).

[guruz]

Then we need to exclude local vs server problem.

Do you have a chance to try this with a different computer (e.g. try a Mac) from a different network (friend, work, etc) ?

Does your webserver have any fancy pagespeed or similar modules?

[marlemion]

I have the same problem from my Windows PC (at work) and my linux PC (at home).

apache2ctl modules:

Loaded Modules:
 core_module (static)
 so_module (static)
 http_module (static)
 mpm_event_module (static)
 access_compat_module (shared)
 actions_module (shared)
 alias_module (shared)
 auth_basic_module (shared)
 authn_anon_module (shared)
 authn_core_module (shared)
 authn_dbm_module (shared)
 authn_file_module (shared)
 authz_core_module (shared)
 authz_dbm_module (shared)
 authz_groupfile_module (shared)
 authz_host_module (shared)
 authz_owner_module (shared)
 authz_user_module (shared)
 autoindex_module (shared)
 cgi_module (shared)
 cgid_module (shared)
 deflate_module (shared)
 dir_module (shared)
 env_module (shared)
 expires_module (shared)
 ext_filter_module (shared)
 filter_module (shared)
 headers_module (shared)
 include_module (shared)
 info_module (shared)
 log_config_module (shared)
 logio_module (shared)
 mime_module (shared)
 mime_magic_module (shared)
 negotiation_module (shared)
 proxy_module (shared)
 proxy_http_module (shared)
 rewrite_module (shared)
 setenvif_module (shared)
 socache_shmcb_module (shared)
 speling_module (shared)
 ssl_module (shared)
 unique_id_module (shared)
 unixd_module (shared)
 usertrack_module (shared)
 vhost_alias_module (shared)
 proxy_html_module (shared)
 xml2enc_module (shared)
 php5_module (shared)

[guruz]

proxy_html_module

https://httpd.apache.org/docs/2.4/mod/mod_proxy_html.html

ext_filter_module

https://httpd.apache.org/docs/2.4/mod/mod_ext_filter.html

Looks suspcious..

[marlemion]

Well, I do proxy something through that server, but not owncloud. Shall I disable ext_filter and try again? (right now I can't remember why I enabled it)

[guruz]

Both modules look like they can rewrite things passed through, so who knows.. yes :)

[marlemion]

Nope. Same problem. Btw. the ext_filter module is enabled by default in the gentoo configuration.

[guruz]

Check your other modules.. ? Check if you really disabled the modules?

Use a binary diff tool to compsare the correct file vs the corrupt file, maybe there is a (plaintext) clue..

[marlemion]

What would be a certified module constellation to be working with owncloud? I don't have the time to disable each module.

I will compare the files when I am back home at my Linux PC.

[guruz]

Would first need to see if it is really the modules, but that's my only guess now..

In general if it is a module, then it is probably because it modified the file.

Maybe it's not a module.

don't ask me about certified though ;) https://owncloud.com/owncloud-server-or-enterprise-edition/

[marlemion]

Ok, some progress. I started owncloud with a very simplistic module environment of apache:

apache2ctl modules
Loaded Modules:
 core_module (static)
 so_module (static)
 http_module (static)
 mpm_event_module (static)
 access_compat_module (shared)
 auth_basic_module (shared)
 authz_core_module (shared)
 authz_host_module (shared)
 dir_module (shared)
 mime_module (shared)
 negotiation_module (shared)
 socache_shmcb_module (shared)
 ssl_module (shared)
 unixd_module (shared)
 proxy_html_module (shared)
 xml2enc_module (shared)
 php5_module (shared)

Now it works. So I have to find the module, which is responsible for the rewrite. :(

[guruz]

@marlemion Great! Good luck and comment here if you found out...

@dragotin @moscicki @DeepDiver1975 ...something which would have been prevented by transfer checksums FYI #2542

[marlemion]

Ok, I narrowed it down. It wasn't any module, but rather a virtual host, which is a redirection from my XBMC/KODI control web interface to my outer url/xbmc:

# -*- apache -*-

Servername servername
ProxyPreserveHost On

<Location /xbmc>
        ProxyPass http://ipadress:8081/xbmc
        ProxyPassReverse http://ipadress:8081/xbmc
</Location>

<Location /xbmc/jsonrpc>
        ProxyPass http://ipadress:8081/jsonrpc
        ProxyPassReverse http://ipadress:8081/jsonrpc
</Location>

SetOutputFilter INFLATE;proxy-html;DEFLATE

ProxyHTMLEnable On

ProxyHTMLDocType "<!DOCTYPE html>"

# these directives need to be duplicated since we're in a
# different scope. We actually only want to add "img data-src"
ProxyHTMLLinks  a       href
ProxyHTMLLinks  area    href
ProxyHTMLLinks  link    href
ProxyHTMLLinks  img     src longdesc usemap data-src
ProxyHTMLLinks  object  classid codebase data usemap
ProxyHTMLLinks  q       cite
ProxyHTMLLinks  blockquote      cite
ProxyHTMLLinks  ins     cite
ProxyHTMLLinks  del     cite
ProxyHTMLLinks  form    action
ProxyHTMLLinks  input   src usemap
ProxyHTMLLinks  head    profile
ProxyHTMLLinks  base    href
ProxyHTMLLinks  script  src for

# finally, rewrite URLs
ProxyHTMLURLMap http://ipadress:8081/xbmc /xbmc

I have disabled it so far. Maybe this is of any help for you. Thanks a lot. I was able to recover my lost data thanks to the versioning feature. :) Apparently, I copied the file from some forum, so I don't have any clue what it does.

[daxprather]

Had same data corruption problem with MS Word files. Completely removing recently installed Kodi from client to see if problem fixed.