OpenBSD fix committed

[Shachar Shemesh <rsyncrypto@sh...>]

Hi all,

I just committed a fix to a problem that prevented rsyncrypto from 
working on OpenBSD. Apparently, either Linux or BSD has a somewhat 
strange behavior where mmap is used with PROT_WRITE but not PROT_READ. 
On BSD you can't do anything with the memory in question, while on Linux 
you can.

This is a first step toward porting rsyncrypto to Windows, making it 
totally multi-platform.

          Shachar

-- 
Shachar Shemesh
Lingnu Open Source Consulting ltd.
Have you backed up today's work? http://www.lingnu.com/backup.html

Re: Run rsyncrypto on windows

[Shachar Shemesh <rsyncrypto@sh...>]

DavidV wrote:

>Message body follows:
>
>Shachar,
>
>This project is very interesting.
>I think you are the only programmer working on a crypto
>program compatible with rsync and I want to congratulate you
>for that.
>  
>
Thanks. After having written it already it turns out there were other 
attempts as well. Both other technologies that I have seen seem to 
require you to keep around your entire previously synced copy, however, 
and so seem to me much less practical than rsyncrypto, which only 
requires you to keep around a ~64 bytes state file per encrypted file. 
As I have not actually worked with any of them, however, I cannot tell 
you that for sure. I'm ccing the mailing list in case someone else has 
any better experience to share.

>But I have an important question for you :
>Is it possible to run it on windows ?
>  
>
Possible, yes, I guess so.

>If so, how ?
>  
>
You need to port it.

Porting rsyncrypto to Windows IS on my "to do" list, so I will get 
around to it eventually, but I really can't tell you when that will 
happen. There are two major hurdles to clear. One is that only Debian  
GNU/{Linux, Hurd, FreeBSD} currently ship gzip with the rsyncable patch 
by default, which means that on any other platform you will have to 
compile gzip as well as rsyncrypto in order to use it. The idea is to 
replace gzip with zlib, and keep a local copy already patched, to be 
compiled along with rsyncrypto. That is the first hurdle to pass.

The second hurdle to pass is to switch rsyncrypto to using standard IO, 
or at least use wrappers. This is not a big problem, as it sort of uses 
wrappers already. It's just something I'll have to do sometimes soon. 
There is also the question of Unicode, but that's more of a rsync issue 
than a rsyncrypto issue.

>Thank you very much in advance for taking a few minutes of
>your precious time to reply.
>  
>
No problem, but please direct all future requests to the mailing list.

>DavidV.
>  
>
          Shachar

-- 
Shachar Shemesh
Lingnu Open Source Consulting ltd.
Have you backed up today's work? http://www.lingnu.com/backup.html