repository uuids

[Stefan Küng <to...@gmail.com>]

Hi,

I always assumed that the repository uuids must be unique for each 
repository. Because of that assumption, the log cache in TSVN relies on 
that fact.

But now, due to users reporting problems with our log cache we detected 
that in the svn book some examples are so that the repository uuid is 
not unique:
http://svnbook.red-bean.com/nightly/en/svn.reposadmin.maint.html#svn.reposadmin.maint.filtering
The example to split one repository into several ones uses the following 
sequence to load the splitted dump files into new repositories:

svnadmin create calc; svnadmin load calc < calc-dumpfile
svnadmin create calendar; svnadmin load calendar < cal-dumpfile
svnadmin create spreadsheet; svnadmin load spreadsheet < ss-dumpfile

but with that sequence, all three repositories will have the same uuid 
as the original repository.

Am I missing something? Or should the sequence better be:

svnadmin create calc; svnadmin load calc --ignore-uuid < calc-dumpfile
svnadmin create calendar; svnadmin load calendar --ignore-uuid < 
cal-dumpfile
svnadmin create spreadsheet; svnadmin load spreadsheet --ignore-uuid < 
ss-dumpfile

?

Stefan

-- 
        ___
   oo  // \\      "De Chelonian Mobile"
  (_,\/ \_/ \     TortoiseSVN
    \ \_/_\_/>    The coolest Interface to (Sub)Version Control
    /_/   \_\     http://tortoisesvn.net

Re: repository uuids

[Daniel Shahaf <da...@tigris.org>]

Stefan Küng wrote on Sat, 17 May 2008 at 09:03 +0200:
> Hi,
> 
> I always assumed that the repository uuids must be unique for each repository.
> Because of that assumption, the log cache in TSVN relies on that fact.
> 

"Unique for each repository", yes; but with svnsync mirroring, 
repositories can have more than one instance (the master instance and
the mirror instance) and all instances have the same UUID -- but not
necessarily the same youngest revision or svn:log revprop values.

In other words, the UUID is unique to a repository, but with svnsync 
repositories can have multiple mutually indistinguishable instances
that might not be fully synced with each other.

There was a dev@ thread about this recently, which concluded that 
inventing another "instance UUID" was unnecessary.

> But now, due to users reporting problems with our log cache we detected that
> in the svn book some examples are so that the repository uuid is not unique:
> http://svnbook.red-bean.com/nightly/en/svn.reposadmin.maint.html#svn.reposadmin.maint.filtering
> The example to split one repository into several ones uses the following
> sequence to load the splitted dump files into new repositories:
> 
> svnadmin create calc; svnadmin load calc < calc-dumpfile
> svnadmin create calendar; svnadmin load calendar < cal-dumpfile
> svnadmin create spreadsheet; svnadmin load spreadsheet < ss-dumpfile
> 
> but with that sequence, all three repositories will have the same uuid as the
> original repository.
> 
> Am I missing something? Or should the sequence better be:
> 
> svnadmin create calc; svnadmin load calc --ignore-uuid < calc-dumpfile
> svnadmin create calendar; svnadmin load calendar --ignore-uuid < cal-dumpfile
> svnadmin create spreadsheet; svnadmin load spreadsheet --ignore-uuid <
> ss-dumpfile
> 
> ?
> 

+1

> Stefan
> 
>