[
https://issues.apache.org/jira/browse/SVN-4858?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17120937#comment-17120937
]
Julian Foad commented on SVN-4858:
----------------------------------
The attached [^test-svn-4858-6.tgz] contains a reproduction recipe that
artificially corrupts a revision and reproduces the observed failure mode
during a commit. (Run the 'test.sh' file. It uses Bash, Python2, and svn 1.10
swig-py bindings. The file named 'output' contains the console output.)
> FSFS corruption - Found malformed header - Corrupt node-revision
> ----------------------------------------------------------------
>
> Key: SVN-4858
> URL: https://issues.apache.org/jira/browse/SVN-4858
> Project: Subversion
> Issue Type: Bug
> Reporter: Julian Foad
> Priority: Major
> Attachments: test-svn-4858-6.tgz
>
>
> This issue exists to document a form of repository corruption that has been
> seen "in the wild".
> The following error was seen by a WANdisco customer at the time of calling
> the FS layer commit-txn API via WANdisco's own replicator product:
> {code:none}
> 160004 Found malformed header '2-8080.0-58760.r87127/41271' in revision file
> 160004 Corrupt node-revision '2-8080.0-58760.r87127/41275'
> {code}
> I can artificially create a repository corruption that produces that exact
> form of error, in a 1.6-compatible repository served by svn 1.10. By my
> analysis the error almost certainly resulted from this kind of corruption.
> The essence of the corruption is: within a directory representation, in
> either a 'dir' entry or a 'file' entry, add 4 to the byte-offset part of the
> entry's node-rev reference.
> The observed failure was in a FSFS repo format 5, filesystem format 4, shard
> size 1000, 60 shards packed; the error reported references r87127 so is in a
> non-packed revision.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
