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

Re: Possible incompatibility of svn_repos_verify_fs2() in 1.9.0-rc1

From: Branko Čibej <brane_at_wandisco.com>
Date: Tue, 09 Jun 2015 08:47:51 +0200

On 08.06.2015 20:22, Branko Čibej wrote:
> On 08.06.2015 19:06, Evgeny Kotkov wrote:
>> Branko Čibej <brane_at_wandisco.com> writes:
>>
>>> I completed your patch and committed the fix in r1683311. Please review!
>>>
>>> -- Brane
>> Sorry, I was on vacation last week and couldn't look at this fix earlier.
>>
>> I reviewed the committed patch and tested how Subversion 1.9.x behaves with it,
>> as the change itself got merged into branches/1.9.x in r1683658. From what I
>> witness, 'svnadmin verify' now erroneously reports the same error twice for a
>> corrupted repository:
>>
>> * Verifying repository metadata ...
>> * Verified revision 0.
>> * Verified revision 1.
>> * Error verifying revision 2.
>> svnadmin: E160062: Malformed node revision ID string
>> svnadmin: E160062: Malformed node revision ID string
>>
>> Relevant hunks of r1683311 are located in subversion/libsvn_repos/dump.c:2433
>> and :2466, where we send the error notification and immediately return the same
>> error to the caller:
>> [[[
>> ...
>> else if (err)
>> {
>> notify_verification_error(rev, err, notify_func, notify_baton,
>> iterpool);
>>
>> if (!keep_going)
>> {
>> /* Return the error, the caller doesn't want us to continue. */
>> return svn_error_trace(err);
>> }
>> ...
>> ]]]
>>
>> The caller (subversion/svnadmin/svnadmin.c:1908) writes both of these errors
>> to stderr, and that results in the erroneous output.
> Good catch. I suppose that means we shouldn't send the notification when
> !keep_going, right?

Should be fixed in r1684325.

-- Brane
Received on 2015-06-09 08:48:16 CEST

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