unicode filenames (in backup)

[MtK]

Hey,
when a user uploads a file with a filename in something other then latin for example עברית.txt (Hebrew), everything is find, and the server lives well with the file's name & content.

upon backup of that user, that same file's name become: עברית.txt
(its content is still intact)

this is really bad, because if we want to restore the account, all those unicode filenames cannot be read anymore by the site storing them.




btw, not sure it's related, but DA doesn't like non-Latin filenames.
when trying to create one, like: עברית.txt

Unable to create create new file

Details

The new file path is invalid.

but the same file, can be uploaded with the same name, to the same path via FTP without any issues.

 

[nobaloney]

What happens if file name is entirely in hebrew, for example:

עברית

(in other words I'm wondering if mixing between a right-to-left and left-to-right characterset in the same word is causing the code to misunderstand the characterset)

Jeff

 

[DirectAdmin Support]

Hello,

DirectAdmin itself is going to be more strict with regards to what it allows for filenames...
However, if the file is in the public_html somewhere, the name shouldn't matter, since it's all up to tar to compress the "domains" directory.

Where exactly is the עברית.txt name being seen? Is it stored like that in the tar.gz file itself, or only after it's been extracted.

Which version of tar are you running?
It may be possible that a newer version has a fix... as some quick googling of the issue results in several posts about possible tar bugs.

tar --version

we've got a compile guide here for an unrelated issue:
http://help.directadmin.com/item.php?id=220

I believe 1.26 is still the latest version

John

 

[MtK]

What happens if file name is entirely in hebrew, for example:

עברית

(in other words I'm wondering if mixing between a right-to-left and left-to-right characterset in the same word is causing the code to misunderstand the characterset)

Jeff

Same result

Hello,

DirectAdmin itself is going to be more strict with regards to what it allows for filenames...
However, if the file is in the public_html somewhere, the name shouldn't matter, since it's all up to tar to compress the "domains" directory.

Where exactly is the עברית.txt name being seen? Is it stored like that in the tar.gz file itself, or only after it's been extracted.

Which version of tar are you running?
It may be possible that a newer version has a fix... as some quick googling of the issue results in several posts about possible tar bugs.

tar --version

we've got a compile guide here for an unrelated issue:
http://help.directadmin.com/item.php?id=220

I believe 1.26 is still the latest version

John

upgraded to 1.26, nothing changed.
even tried to manually create a tar.gz (not through backup) and got:

An error occured while creating the compressed file

Details

/domains/unicode.lighthost.co.il/public_html/עברית is not a valid path

and the same for

/domains/unicode.lighthost.co.il/public_html/עברית.txt is not a valid path

 

[nobaloney]

Seems to be an issue with tar, then.

Jeff

 

[MtK]

Seems to be an issue with tar, then.

Jeff

Maybe, but there's also a DA bug, as I mentioned - try uploading a file called עברית though the filemanager

 

[DirectAdmin Support]

DirectAdmin itself is going to be more strict with regards to what it allows for filenames.

This holds true for this case. What DA allows as a filename is going to be very strict.
The name you've mentioned won't fall under that category as it contains characters which DA may not correctly handle, thus does not fall under the category of allowed filenames.
I wouldn't consider this a DA bug, but rather a design choice.
The issue with tar however, may be a tar bug (or perhaps just a design choice by the tar developers, I'm not too sure)

Other services like proftpd, apache, etc.. will likely handle them correctly, so they can be used normally.

John