subversion-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Julian Foad <julianf...@btopenworld.com>
Subject Re: Testing equality between svnrdump and svnadmin dump
Date Mon, 12 Jan 2015 15:25:44 GMT
Branko Čibej wrote:
> On 12.01.2015 13:54, Julian Foad wrote:
>> svnmucc_tests.py 2 'basic svnmucc tests' generates a no-op modification
>> to an empty file, in revision 16. 'svnadmin dump' outputs this:
>>  [[[
>>  I: Node-path: boozle/buz/svnmucc-test.py
>>  I: Node-kind: file
>>  I: Node-action: change
>>  I: 
>>  I: 
>>  ]]]
>> 
>>  whereas 'svnrdump dump' outputs this:
>>  [[[
>>  I: Node-path: boozle/buz/svnmucc-test.py
>>  I: Node-kind: file
>>  I: Node-action: change
>>  I: Text-delta: true
>>  I: Text-delta-base-md5: d41d8cd98f00b204e9800998ecf8427e
>>  I: Text-content-length: 4
>>  I: Text-content-md5: d41d8cd98f00b204e9800998ecf8427e
>>  I: Content-length: 4
>>  I: 
>>  I: SVN
>>  I: 
>>  ]]]
> 
> The latter looks like an empty delta. This could be an artefact of how a
> particular RA layer represents a no-op change.

Yep, it's an empty delta, but it's not a difference between RA layers, rather a difference
between how the repository sends this revision over the RA layers vs. how the repository dumps
this.

We definitely don't want these outputs to be different. That can't help anybody.

I'm not sure either of these outputs is correct. As I discussed in a previous thread, I think
a no-op 'change' record is an artefact of the FS layer that is not wanted (nor fully supported)
above the FS layer, and so should be hidden above the FS layer. So neither svnadmin nor svnrdump
should output a record here at all.

- Julian


Mime
View raw message