Re: possibly useful info re: FSFS corruption

On Mon, Jul 10, 2006 at 05:02:01PM +0100, Thomas Beale wrote:
> We have just had it again, and it occurred to me that a common factor
> both times was that largish binary files were being uploaded on the
> faulty commit (first time - PDFs, this time, .Net DLLs). Devs here might
> already know that, but just in case....
>

Thanks for the report. I've no evidence to prove it, but I suspect
that the reason that binary files are usually involved is simply because
they tend to generate larger revisions, and so more data to be written,
leading to a higher chance of failure.

> I am presuming that at the moment the only fix was the one I got from
> Michael Brouwer, which was to manually delete the revisions from the
> server back to the broken one, and then re-apply the changes...
>

If you can take that approach, it's certainly the safest option. In case
of emergency, there is actaully a python script that John Szakmeister
posted that can repair a corrupted (svndiff0) repository, but that's
very much unsupported.

Regards,
Malcolm

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Jul 10 18:22:54 2006