Jonathan Wakely <jwakely....@gmail.com>:
> My concern is that there is no conversion done using reposurgeon that
> *can* be used to do correctness checks.
We can in fact verify revisions of a GCC conversion in place using
repotool compare. Joseph Myers has been using this with reposurgeon's
readlimit to run tests.
Unfortunately, on a repository this large, it's not practical to run a
verification on every single revision. The blocker is the slowness of
svn checkout. In practice, you have to sample key revisions, with
particular attention to those at and just after known metadata
defects.
The conversion crew - which now includes Joseph Myers and Richard
Earnshaw, in addition to my co-developers Daniel Brooks and Julien
Rivaud - is diligently testing as it refines the last bits of the
conversion.
I believe everybody on the crew is now satisfied that we're converging
on a good result. It helps that we now have a detailed characterization
of the pathological trunk deletion at r184996; most of the conversion
problems radiated from that.
--
<a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>
