Skip Menu |
 

This queue is for tickets about the App-SVNBinarySearch CPAN distribution.

Report information
The Basics
Id: 20593
Status: resolved
Priority: 0/
Queue: App-SVNBinarySearch

People
Owner: Nobody in particular
Requestors: duff [...] pobox.com
Cc:
AdminCc:

Bug Information
Severity: (no value)
Broken in: (no value)
Fixed in: (no value)



Subject: WC detritus could skew results
Download (untitled) / with headers
text/plain 930b
This is just an observation/feature request based on my usage of this module. Subversion has this "feature" whereby if you downgrade to an earlier revision where certain files/directories didn't exist, then when you attempt an upgrade to a later revision after they've been created, it will complain with something like this: svn: Failed to add directory 'blah': object of the same name already exists. Thus, when doing a binary search through revisions, some updates may not accurately reflect the state of the repository at a particular revision. Also, the test program that determines which revision affected a change may create spurious files in the working copy that are present during subsequent runs and these files may affect the output of the test program. The only guaranteed fix that I see right now is to always checkout pristine revisions rather than using "svn up" to move a single WC from revision to revision.
On Thu Jul 20 17:06:13 2006, duff wrote: Show quoted text
> This is just an observation/feature request based on my usage of this > module. > > Subversion has this "feature" whereby if you downgrade to an earlier > revision where certain files/directories didn't exist, then when you > attempt an upgrade to a later revision after they've been created, it > will complain with something like this: > > svn: Failed to add directory 'blah': object of the same name already
exists. Show quoted text
> > Thus, when doing a binary search through revisions, some updates may not > accurately reflect the state of the repository at a particular revision. > > Also, the test program that determines which revision affected a change > may create spurious files in the working copy that are present during > subsequent runs and these files may affect the output of the test program. > > The only guaranteed fix that I see right now is to always checkout > pristine revisions rather than using "svn up" to move a single WC from > revision to revision.
Good idea, moved to http://code.google.com/p/vc-binary-search/issues/detail?id=2 for tracking.


This service is sponsored and maintained by Best Practical Solutions and runs on Perl.org infrastructure.

Please report any issues with rt.cpan.org to rt-cpan-admin@bestpractical.com.