[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: svn commit: r1186144 - in /subversion/branches/1.7.x: ./ STATUS subversion/libsvn_fs_fs/fs_fs.c

From: Daniel Shahaf <d.s_at_daniel.shahaf.name>
Date: Wed, 19 Oct 2011 15:07:52 +0200

People who are bitten by this will not be able to commit anything with
this revision applied. Should we add a way to turn off the new
validation?

For svn.a.o I'll just patch it locally, but I suppose not everyone who
runs servers also compiles them from source...

hwright_at_apache.org wrote on Wed, Oct 19, 2011 at 12:14:42 -0000:
> Author: hwright
> Date: Wed Oct 19 12:14:42 2011
> New Revision: 1186144
>
> URL: http://svn.apache.org/viewvc?rev=1186144&view=rev
> Log:
> Merge r1178280, r1178282 from trunk:
>
> * r1178280, r1178282
> Add a sanity check to FSFS.
> Justification:
> Detects a user-visible corruption seen on svn-master.apache.org.
> Votes:
> +1: danielsh, rhuijben, stsp
>
> Modified:
> subversion/branches/1.7.x/ (props changed)
> subversion/branches/1.7.x/STATUS
> subversion/branches/1.7.x/subversion/libsvn_fs_fs/fs_fs.c
>
> Modified: subversion/branches/1.7.x/STATUS
> URL: http://svn.apache.org/viewvc/subversion/branches/1.7.x/STATUS?rev=1186144&r1=1186143&r2=1186144&view=diff
> ==============================================================================
> --- subversion/branches/1.7.x/STATUS (original)
> +++ subversion/branches/1.7.x/STATUS Wed Oct 19 12:14:42 2011
> @@ -124,10 +124,3 @@ Approved changes:
> fix here than having to fix all clients.
> Votes:
> +1: rhuijben, pburba, stsp
> -
> - * r1178280, r1178282
> - Add a sanity check to FSFS.
> - Justification:
> - Detects a user-visible corruption seen on svn-master.apache.org.
> - Votes:
> - +1: danielsh, rhuijben, stsp
>
> Modified: subversion/branches/1.7.x/subversion/libsvn_fs_fs/fs_fs.c
> URL: http://svn.apache.org/viewvc/subversion/branches/1.7.x/subversion/libsvn_fs_fs/fs_fs.c?rev=1186144&r1=1186143&r2=1186144&view=diff
> ==============================================================================
> --- subversion/branches/1.7.x/subversion/libsvn_fs_fs/fs_fs.c (original)
> +++ subversion/branches/1.7.x/subversion/libsvn_fs_fs/fs_fs.c Wed Oct 19 12:14:42 2011
> @@ -5873,6 +5873,29 @@ write_hash_rep(svn_filesize_t *size,
> return svn_stream_printf(whb->stream, pool, "ENDREP\n");
> }
>
> +/* Sanity check ROOT_NODEREV, a candidate for being the root node-revision
> + of (not yet committed) revision REV. Use OCEAN for temporary allocations.
> + */
> +static APR_INLINE svn_error_t *
> +validate_root_noderev(node_revision_t *root_noderev,
> + svn_revnum_t rev)
> +{
> + /* Bogosity seen on svn.apache.org; see
> + http://mid.gmane.org/20111002202833.GA12373@daniel3.local
> + */
> + if (root_noderev->predecessor_count != -1
> + && root_noderev->predecessor_count != rev)
> + {
> + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL,
> + _("predecessor count for "
> + "the root node-revision is wrong: "
> + "found %d, committing r%ld"),
> + root_noderev->predecessor_count, rev);
> + }
> +
> + return SVN_NO_ERROR;
> +}
> +
> /* Copy a node-revision specified by id ID in fileystem FS from a
> transaction into the proto-rev-file FILE. Set *NEW_ID_P to a
> pointer to the new node-id which will be allocated in POOL.
> @@ -5890,6 +5913,10 @@ write_hash_rep(svn_filesize_t *size,
> If REPS_TO_CACHE is not NULL, append to it a copy (allocated in
> REPS_POOL) of each data rep that is new in this revision.
>
> + AT_ROOT is true if the node revision being written is the root
> + node-revision. It is only controls additional sanity checking
> + logic.
> +
> Temporary allocations are also from POOL. */
> static svn_error_t *
> write_final_rev(const svn_fs_id_t **new_id_p,
> @@ -5902,6 +5929,7 @@ write_final_rev(const svn_fs_id_t **new_
> apr_off_t initial_offset,
> apr_array_header_t *reps_to_cache,
> apr_pool_t *reps_pool,
> + svn_boolean_t at_root,
> apr_pool_t *pool)
> {
> node_revision_t *noderev;
> @@ -5938,7 +5966,7 @@ write_final_rev(const svn_fs_id_t **new_
> svn_pool_clear(subpool);
> SVN_ERR(write_final_rev(&new_id, file, rev, fs, dirent->id,
> start_node_id, start_copy_id, initial_offset,
> - reps_to_cache, reps_pool,
> + reps_to_cache, reps_pool, FALSE,
> subpool));
> if (new_id && (svn_fs_fs__id_rev(new_id) == rev))
> dirent->id = svn_fs_fs__id_copy(new_id, pool);
> @@ -6036,6 +6064,8 @@ write_final_rev(const svn_fs_id_t **new_
> noderev->id = new_id;
>
> /* Write out our new node-revision. */
> + if (at_root)
> + SVN_ERR(validate_root_noderev(noderev, rev));
> SVN_ERR(svn_fs_fs__write_noderev(svn_stream_from_aprfile2(file, TRUE, pool),
> noderev, ffd->format,
> svn_fs_fs__fs_supports_mergeinfo(fs),
> @@ -6315,7 +6345,7 @@ commit_body(void *baton, apr_pool_t *poo
> root_id = svn_fs_fs__id_txn_create("0", "0", cb->txn->id, pool);
> SVN_ERR(write_final_rev(&new_root_id, proto_file, new_rev, cb->fs, root_id,
> start_node_id, start_copy_id, initial_offset,
> - cb->reps_to_cache, cb->reps_pool,
> + cb->reps_to_cache, cb->reps_pool, TRUE,
> pool));
>
> /* Write the changed-path information. */
>
>
Received on 2011-10-19 15:09:33 CEST

This is an archived mail posted to the Subversion Dev mailing list.