Re: svn commit: r893179 - in /subversion/trunk/subversion/libsvn_fs_base: dag.c revs-txns.c

Julian Foad <julian.foad_at_wandisco.com> writes:

> Philip Martin wrote:
>> julianfoad_at_apache.org writes:
>>
>> > + if (txn_obj->copies)
>> > + {
>> > + int i;
>> > +
>> > + for (i = 0; i < txn_obj->copies->nelts; i++)
>> > + {
>>
>> Is this loop big enough to warrent an iteration pool?
>
> It could be large in some revisions, so yes.

node_origins_update() also has a loop that doesn't use an iteration
pool:

  SVN_ERR(svn_fs_bdb__changes_fetch_raw(&changes, trail->fs, old_txn_id, trail,
                                        scratch_pool));
  for (i = 0; i < changes->nelts; i++)
    {

I suppose fetch_raw is already allocating memory that is proportional
to the number of changes, so perhaps that's sufficient justification
not to use an iteration pool in this case.

-- 
Philip