Re: svn commit: r1205726 - /subversion/trunk/subversion/libsvn_fs_base/reps-strings.c

On 11/24/2011 05:04 PM, Daniel Shahaf wrote:
> cmpilato_at_apache.org wrote on Thu, Nov 24, 2011 at 05:45:25 -0000:
>> Author: cmpilato
>> Date: Thu Nov 24 05:45:24 2011
>> New Revision: 1205726

[...]

>> NOTE: This patch appears to pass all the BDB tests on my laptop, but
>> of course those aren't known to cover large datasets. I would
>> really, really appreciate some extra eyes on this change!
>>
>
> Reviewed the scratch pool usage in these functions and their callees,
> looks good. I'll go ahead and add my +1 to backport.

Thanks so much for the extra eyes!

>> Wondering aloud... if this approach turns out to be correct, should
>> the corresponding stream write function in this same file
>> (rep_write_contents) use a similarly initialized scratch pool and
>> clear it at the start of each invocation, too?
>>
>
> I suppose this might be a good idea; datasets that are problematic for
> read() should also be problematic for write(), correct?

I'm not sure that the same problems exist for write(), but after the weekend
I may take a look around this part of the codebase.

-- 
C. Michael Pilato <cmpilato_at_collab.net>
CollabNet   <>   www.collab.net   <>   Distributed Development On Demand