On 10.03.2014 12:11, Julian Foad wrote:
> [switching back to plain text]
> Branko Čibej wrote:
>>> + /* An error in verifying the mergeinfo must not prevent dumping
>>> + the data. Ignore any such error. */
>>> + svn_error_clear(verify_mergeinfo_revisions(
>>> + eb->found_old_mergeinfo,
>>> + mergeinfo_str->data, eb->oldest_dumped_rev,
>>> + eb->notify_func, eb->notify_baton,
>>> + pool));
>> Shouldn't we at least emit a warning here?
> Yes and no. I still have the patch to do so, but I wanted to commit this simple version first in any case, as it is considerably easier to back-port.
> The interesting thing is that this code is only executed on a partial dump, where start revision is greater than 1, and only in order to be able to emit other warnings. It's not even executed at all during 'verify'. There is still merit in just adding a warning here, but it could almost be seen as misleading to warn in some cases yet fail to warn during full dumps and during verify. It seems to me it would be much better to parse the mergeinfo and warn if it's bad in all dumps and especially also during 'verify'.
> What do you think?
AIUI, the warning really does apply only to partial dumps; since it's
not only checking for r0 buit for any mergeinfo that refers to revisions
outside the dump range.
So we're really talking about two different cases, right?
Branko Čibej | Director of Subversion
WANdisco // Non-Stop Data
Received on 2014-03-11 15:34:58 CET