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

Re: svn commit: r33528 - in trunk/subversion: include libsvn_delta libsvn_wc

From: Greg Stein <gstein_at_gmail.com>
Date: Mon, 13 Oct 2008 05:02:16 -0700

On Sun, Oct 12, 2008 at 4:10 PM, Joe Swatosh <joe.swatosh_at_gmail.com> wrote:
> line 47 is the marked assertion. So it appears that
> svn_txdelta_md5_digest is returning a different value than it used to.
> Not sure if its significant, but I was interested in your opinion
> about if there is a better way to test this or if the problem is the
> changed return value.

I just added a new test: subversion/tests/libsvn_delta/window-test.c.
That appears to show the digest is being returned properly.

Without further pointers, I'm going to assume that there is some
memory corruption in how Ruby transfers that digest into a Ruby value.
Note that the first half of the digest is the same as the expected
value, but the last half has 0x44 ('D'), then a combo of 0x33 ('3')
and 0x34 ('4') values. Given the way that MD5 works, I am *extremely*
suspicious that a hash was able to match the first eight bytes.

Oh. And I think that I see the problem. I added a "result_pool" into
the txstream baton for allocation of the checksum, but didn't actually
use it! I've just fixed that and committed (r33613). Please try your
test again.

This may also fix the memory corruption that Jeremy was seeing.


To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-10-13 14:02:32 CEST

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