Re: [Issue 1751] svn switch may corrupt working copy
From: Julian Foad <julianfoad_at_btopenworld.com>
Date: 2004-03-08 21:44:13 CET
http://subversion.tigris.org/issues/show_bug.cgi?id=1751
Marvin Greenberg wrote:
It says "Note: It is irrelevant for the recipe whether wc2 is checked out recursively or not." But, as I understand it, the "switch" command must be non-recursive for that recipe. I did some experiments to confirm this.
And it says "Changed summary, since a part of the problem occurs with recursive working copies too, and in all cases the working copy is dead." Again, as I understand it, this still referred only to problems caused by "switch --non-recursive".
However, there difinitely are problems also with fully recursive operation:
> I have had this happen doing really simple things like a switch from
If you can make a recipe that reproduces the problem, that would be very helpful.
The very first thing mentioned in that issue is a (failed) plain switch command messing up a non-recursive WC (leaving it with a bad URL). I can reproduce something like that:
# Make a repository
# Make a branch
# Check out a WC from branch1
# Show the correct info
# If "branch2" exists (empty), it is OK. If it does not exist, then it fails
# Try to switch WC to "branch2".
# Check the result
I think we will agree that, while it would be acceptable to fail and leave the WC in an incomplete state (or any state that can easily be recovered from), a failure such as this should not leave the WC broken. There are other, similar failures such as the one in the issue in which the destination URL does exist but the switch fails with "Working copy 'www' not locked".
- Julian
---------------------------------------------------------------------
|
This is an archived mail posted to the Subversion Dev mailing list.
This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.