duplicity-team team mailing list archive

[Kenneth Loafman <kenneth@xxxxxxxxxxx>]

Kenneth Loafman wrote:
> Peter Schuller wrote:
>>>> is a correct transcribation of the previous version whichi called
>>>> copy_raw() if the file was a manifest.
>>> In the normal course of processing the difference between manifest and
>>> sigtar are necessary in the local side.  The manifest is not compressed
>>> and the sigtar is, so the way I had it was correct.
>>>
>>> On the remote side, it's either encrypted or gzipped.  I'll review your
>>>  changes and see if there are any problems.
>> Hopefully it's still correct but expressed another way; I'm just not
>> sure about it.
>>
>> However I also realized sometime yesterday, though I didn't bother
>> posting until LP's mail queue got flushed, that the "spurious" files
>> likely are there sort of by design given the checkpoint/restart
>> logic. So I'm not sure whether the fix to remove them makes sense. I
>> probably won't have much time to look into it again until next weekend
>> though.
>>
>> In any case, I do think that the recent improvements in archive
>> handling in 6.x are going to be very nice, and I particularly value
>> the ability to 'synchronize' the archive. In my case also in removal
>> of files, to avoid old "cruft" accumulating. But of course I don't
>> want to disable the checkpoint/restart feature either.
> 
> One reason I did not try to remove spurious files on the local side is
> that there just might be a missing sigtar file on the remote even though
> we don't detect it at the moment.  Also, the spurious files might just
> be partial backup files and we would not want to delete those until
> later.  I'd leave any pr.partial files you find.
> 
> So, you'll still need most of the code I had, but refactored to have the
> resolve_basename() function split out.  On the remote side the correct
> suffix is give by get_suffix(globals.encryption, not globals.encryption)
> and mine was wrong, so you need to fix that.
> 
> I've got some time and would like to finish your work and get this
> released soon so that we have a working version out there.

I merged your changes in to the trunk and fixed a few things.

On the local side we still need to look at pr.manifest to decide whether
to compress or not.  I made sure not to include any pr.partial files in
the original list of files, nor any non-duplicity files, just in case.
As far as I can tell from a few test cases, its all working as it
should.  If you've got time, take a look and test it out.  I'll be
adding some unittest cases.


...Ken