Re: [rdiff-backup-users] Re: [librsync-users] more info on 25gig files

[Clint Silvester]

Ben Escoto wrote:

>>>>>>Clint Silvester <address@hidden>
>>>>>>wrote the following on Thu, 12 May 2005 14:29:27 -0600
>>>>>>            
>>>>>>
>
>  
>
>>I tried putting this into Rdiff.py to replace the previous divide by
>>2000 section and I am still getting:
>> python: ERROR: (rs_job_iter) internal error: job made no progress
>>[orig_in=120794, orig_out=65536, final_in=120794, final_out=65536]
>>UpdateError TWDOCS2002.adm librsync error 107 while in patch cycle
>>    
>>
>
>Yeah, although that may help performance, the old code should not have
>been the source of any errors.  (As I understand it, at worst deltas
>may have been larger than necessary.)
>
>  
>
>>I also notice in _librsyncmodule.c there is #define RS_JOB_BLOCKSIZE
>>65536 and 65536 is the the orig_out and final_out that it's saying in
>>that message.  Is it something to do with that define perhaps?  By the
>>way, has anyone been able to test files >=25gig and had it work?
>>    
>>
>
>Possibly, looking at librsync 0.9.7 it seems that their
>RS_DEFAULT_BLOCK_LEN (which corresponds to my RS_JOB_BLOCKSIZE I
>think) is 2048.  But offhand I don't see why that would cause any
>problems.
>
>Just to clarify, you say that the rdiff program works, but Rdiff.py
>fails even when it uses the same version of librsync?  Then it would
>seem the problem is not librsync.  Also you said Rdiff.py works on
>files greater than 4GB (but less than 25GB)?
>
>
>  
>
Well, that's what I thought I tested, but now I don't know what I was
running with that rdiff command.  I'm really confused now looking at the
minimal help what to do with rdiff as a test.  Any pointers?

Clint Silvester