Ramkumar Ramachandra wrote on Sat, Jul 24, 2010 at 02:26:47 +0530:
> Hi Daniel,
>
> Daniel Shahaf writes:
> > Good question.
> >
> > I'd probably say "last revision successfully dumped was rN" and then exit with
> > a non-zero code, leaving it to the admin to clean up the mess (i.e., to truncate
> > the dumpfile after the end of rN's data).
> >
> > Unfortunately, since you dump to stdout, you can't do that seek+truncate yourself :-(
>
> Yeah, this makes sense.
>
> > > can we just look for
> > > "^Revision-number: \d+$" from the bottom of the file, clean it up and
> > > then resume?
> > >
> >
> > What if a file contains that line as part of its content?
>
> Hm. Maybe it'll make sense to keep track of the total content length
> successfully written so that we can truncate. But as Stefan pointed,
> we can think about this again in a few days- it's not urgent.
>
You could write a quick tools/ utility to figure out (by parsing from
the start to the end) and cut the dump at the right place.
Except, of course, that figuring the 'right' place might be ambiguous,
since there is no way to tell whether the last revision in the dump was
dumped in whole or in part (since there is no Revision-content-length:
header).
> -- Ram
Received on 2010-07-23 23:08:55 CEST