SMB2FS OS3 port

[BSzili]

There's already UTF-8 support in filesysbox.library, which is why I asked for some example file names to easier testing.

[DJBase]

Originally Posted by BSzili

I haven't tested any files or passwords with international characters yet. If possible, could you give me any examples of file names you had problems with?

Try something with '?' as character in example. Seems to not work in passwords.

[BSzili]

That's a known issue, as ? is used internally as a string separator. You can use the PASSWORD argument instead as a workaround.

[amigafreak68k]

Download LimpidClock, extract the Archive and copy the locale files to a share

http://aminet.net/package/util/wb/limpidclock

[BSzili]

Thanks. Should I try copying those files from the share or to the share?

[amigafreak68k]

to the share… From the share I didn‘t tested right now

[BSzili]

I was able to reproduce this, I'll look into what causes the problem.

[BSzili]

While filesysbox.library comes with UTF-8 support, it assumes that the underlying system can cope with UTF-8 strings, which is not the case for "classic" AmigaOS. To make it worse, there's no system-wide character set setting. This is why codesets.library has to do all the guesswork based on the system language and environment variables.
Fortunately ISO-8859-1 <-> UTF-8 conversion is relatively simple and fast, so I'll implement this in the handler. Of course this is not a 1:1 conversion as not every UTF-8 character can be represented in Latin-1, but it'll work for most users.

[E-Penguin]

This is premium-level technical support

[tygre]

Hi BSzili!

Originally Posted by BSzili

While filesysbox.library comes with UTF-8 support, it assumes that the underlying system can cope with UTF-8 strings, which is not the case for "classic" AmigaOS. To make it worse, there's no system-wide character set setting. This is why codesets.library has to do all the guesswork based on the system language and environment variables.
Fortunately ISO-8859-1 <-> UTF-8 conversion is relatively simple and fast, so I'll implement this in the handler. Of course this is not a 1:1 conversion as not every UTF-8 character can be represented in Latin-1, but it'll work for most users.

That would be awesome!

I cannot give you filenames in Cyrillic or others here because EAB actually only accepts Latin characters but let me know if you want some, I could send them to you by e-mail.

PS. I'll be curious to know your solution: mine consisted in replacing UTF16 "chars" by Xs and append a unique identifier to each file...

[Dale423]

@BSzili

I cannot access your site with Chrome on Win11. (https://bszili.morphos.me/stuff/smb2-handler). I am getting this error:

This site can’t provide a secure connectionbszili.morphos.me uses an unsupported protocol.
ERR_SSL_VERSION_OR_CIPHER_MISMATCH

[Thorham]

Same here with Firefox on Windows 10:

Secure Connection Failed

An error occurred during a connection to bszili.morphos.me. Cannot communicate securely with peer: no common encryption algorithm(s).

Error code: SSL_ERROR_NO_CYPHER_OVERLAP

The page you are trying to view cannot be shown because the authenticity of the received data could not be verified.

Please contact the website owners to inform them of this problem.

[BSzili]

Try a different browser that allows you to access http:// links without forcefully promoting them to https:// which currently isn't available on my website. I'm using Brave, but there must be other browsers that don't do that.

Originally Posted by tygre

Hi BSzili!
That would be awesome!

I cannot give you filenames in Cyrillic or others here because EAB actually only accepts Latin characters but let me know if you want some, I could send them to you by e-mail.

PS. I'll be curious to know your solution: mine consisted in replacing UTF16 "chars" by Xs and append a unique identifier to each file...

The mapping will probably be complicated by the FUSE API, as many functions expect a path. In the first round there will be no Cyrillc support because of the character set issue, only Latin-1. Later I'll add optional support for codesets.library for those who use a different locale.

[tygre]

Hi BSzili!

Originally Posted by BSzili

The mapping will probably be complicated by the FUSE API, as many functions expect a path. In the first round there will be no Cyrillc support because of the character set issue, only Latin-1. Later I'll add optional support for codesets.library for those who use a different locale.

Indeed!

My "solution" was to expose "AmigaOS-ready" names (unique Latin-1 filenames) only to the Amiga, while keeping the normal names (UTF16) internally, using a simple key-value dictionary to remember the mapping.

For example, AmigaAMP would list a directory containing a file "XXXXX-1234.mp3" and can play it because, internally, the file name would still be something with 5 UTF16 "chars" that the Samba server can understand and access.

Keep us posted on your implementation!

[BSzili]

It turns out salass00 is already working on a solution. His idea is to use URL encoding-style escaping which would allow for mapping the characters back without the need to store the original UTF-8 names.

[bni]

Thanks a lot for this.

Just tested it on my A1200 with PCMCIA network card. Works great to connect to my Linux/Samba machine. File listing and file transfer is fast.

Does not work with macOS Ventura built in file sharing however. Get the "No disk in drive". I tried to force SMB2 like described here https://www.droidwin.com/smb-file-sh...s-ventura-fix/ but didn't make a difference.

[BSzili]

In theory both SMB2 and SMB3 should work as the handler supports both, no disk means it couldn't connect to the share. To rule out problems on the Mac side, have you tried connecting from your Linux to it?

[bni]

Tried that now and SMB seems to be completely non-functional in macOS Ventura

Aborts after connect and login with
smb: \> SMBecho failed (NT_STATUS_CONNECTION_DISCONNECTED). The connection is disconnected now

So I won't clutter the thread with macOS problems further.

[patrik]

Originally Posted by BSzili

It turns out salass00 is already working on a solution. His idea is to use URL encoding-style escaping which would allow for mapping the characters back without the need to store the original UTF-8 names.

Smart solution!

[tygre]

Originally Posted by patrik

Smart solution!

PS. Consider also the the lengths of the file names: escaping will increase a lot the file names, which should be limited to 32 or 107 or... depending on the filesystem? (A unique identifier will still be needed in case two (original) file names different only beyond the max. length of the Amiga, else they would be undistinguishable! )