tldr: a node-rev on svn.us differs in two ways from the same noderev on
svn.eu, with at least one of the differences causing user-visible wrong
answers to history queries.
Any information towards getting to the bottom of this would be
appreciated. In particular, if any of those of you with local mirrors
sees either of the corruption forms detailed below, please speak up.
----- Forwarded message from Daniel Shahaf <d.s_at_daniel.shahaf.name> -----
> Date: Thu, 29 Sep 2011 22:12:16 +0300
> From: Daniel Shahaf <d.s_at_daniel.shahaf.name>
> To: infrastructure-private_at_apache.org
> Subject: Long-standing corruption on eris
> Reply-To: infrastructure-private_at_apache.org
> Message-ID: <20110929191216.GA9677_at_daniel3.local>
>
> Here is a node-revision on eris:
[eris == svn.us.apache.org]
>
> [[[
> id: 0-836401.0.r891679/11455
> type: dir
> pred: 0-836401.0.r891672/13070
> count: 43983
> text: 891679 11072 370 370 b03cc8a5e429084dd4fdf39d6fe46b07
> props: 866503 8246 73 0 d34667037ecfd7bab466410a1c52bda6
> cpath: /subversion
> copyroot: 0 /
> minfo-cnt: 34417865037
> ]]]
>
> Here is the same one on harmonia:
[harmonia == svn.eu.apache.org]
>
> [[[
> id: 0-836401.0.r891679/11456
> type: dir
> pred: 0-836401.0.r891677/3630
> count: 43984
> text: 891679 11073 370 370 1b6d020fb55eb7b0c0fca870b93437c4
> props: 866503 8246 73 0 d34667037ecfd7bab466410a1c52bda6
> cpath: /subversion
> copyroot: 0 /
> minfo-cnt: 293
> ]]]
>
> Differences:
>
> - eris's minfo-cnt is bogus.
>
> *Should* be harmless, as long as it's > the correct value and not
> less than it.
>
> - eris's pred is bogus, too.
>
> Ugly. This results in wrong outputs:
>
> $svn log -qvl2 http://svn.eu.apache.org/repos/asf/subversion@r891679
> $svn log -qvl2 http://svn.us.apache.org/repos/asf/subversion@r891679
>
> ---
>
> Possibly-relevant information: min-unpacked-rev is 893000 on both eris
> and harmonia.
>
----- End forwarded message -----
Received on 2011-09-30 19:19:49 CEST