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

Re: broken window checksums?

From: <kfogel_at_collab.net>
Date: 2003-11-03 05:20:15 CET

David Kimdon <david@kimdon.org> writes:
> I see what appears to be an inconsistency between the code and the
> documentation.
>
> In subversion/libsvn_fs/structure:288 the svndiff window checksum is
> described as the checksum on just the portion of the fulltext
> represented by this window. In subversion/libsvn_fs/reps-strings.c
> (svn_fs__rep_deltify) around line 1493 the chunk is set to the
> checksum of the complete full-text.
>
> Am I missing something?

If you are, I'm missing it too.

> Do we really want per-window checksums? We already have end-to-end
> checksums so maybe per-window checksums are not needed. (update to
> issue #1096, the md5sums are not zero anymore, I searched back deep
> into revision history and it appears as though they have no been zero
> for a while)

I think you're right, that per-window checksums are just a waste of
time now.

Hmmm. I wonder when the md5 digests started getting stored
(considering that #1096 is still open)?... I don't see the change
that did it. 'svn blame -r1949' doesn't behave as I expected -- it
doesn't show the annotations for the file as it was in r1949, instead
seems to be doing something else I can't quite figure out...

Urk.

I've annotated issue #1096 to point to this thread.

> Should svn_fs__rep_deltify be changed to calculate and save per window
> (chunk) checksums?
>
> Currently there is no code that verifies the per-window checksums.
> Should rep_undeltify_range() be verifying per-window checksums?

No and no, I think. End-to-end and whole-rep checksums are enough
protection.

I sure wish I knew how the code got like that, though. (Probably
something I wrote.) Fixing it is clearly not a matter of a quick dash
through the code tonight, however!

Nice catch, sir.

-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Nov 3 06:00:22 2003

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.