[Csync2] Regression in 2.0-rc1?

[Dennis Schafroth]

On 17/05/2011, at 13.24, Lars Ellenberg wrote:

> On Tue, May 17, 2011 at 10:16:21AM +0200, Dennis Schafroth wrote:
>> 
>> On 16/05/2011, at 21.55, Lars Ellenberg wrote:
>> 
>>> On Sun, May 15, 2011 at 03:59:47PM +0200, Dennis Schafroth wrote:
>>>> Hi 
>>>> 
>>>> I started using this yesterday, but is seeing a lot of errors: the log from the sending side looks like this:
>>> 
>>> rc1, or current git?
>> 
>> Git. Did a pull. Didnt see any comming down: 
>> 
>> commit b7490b4d2f9069e4e364c9f4824956ade4bad691
> 
>>> standalone csync2, or via inetd?
>> 
>> inetd. 
>> 
>> Since then I have tried running in standalone to debug it, but haven't been able to reproduce. So yeah, been thinking in the direction of inetd issue. 
> 
> If you can give me a short reproducer,
> I'll see to it that it gets fixed.

It's related to the new open_temp_file which tries to open temporary files in the same directory as the file it wants to check. However that idea breaks when the whole directory, where local file was located, has been deleted. So when csync_rs_check is called from csync_update_file_del sig_file will be null and the method call jump to io_error and never read the signature from the peer. 

I reverted to using a slightly modified paranoid_tmpfile but also rewrote the csync_rs_check to handle if the temp files fails. I am not sure the later is a good idea, if we wanna be paranoid.
This patch also contains the fix for backup when using prefixes.
 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: patch.diff
Type: application/octet-stream
Size: 6890 bytes
Desc: not available
URL: <http://lists.linbit.com/pipermail/csync2/attachments/20110522/d26bfdfb/attachment.obj>
-------------- next part --------------


It works for me, but please review and do your own testing :-) 

cheers, 
:-Dennis Schafroth