Joseph Myers <jos...@codesourcery.com>:
> I did a comparison of git and SVN checkouts to look at missing file
> problems. I've now filed reposurgeon issues 171 and 172 for the problems
> I noted. Issue 171 relates to handling of trunk deletion / recreation.
> Issue 172 relates to the first point where missing file problems appear
> (unless some appeared and then disappeared in the history before then).
> As it's at a very early point in the GCC history (r14877), hopefully it
> shouldn't be too hard to track down if your rewrite doesn't fix it, since
> it shouldn't require loading much of the history to reproduce. (Roughly,
> it's at the start of EGCS, i.e. around the point where we spliced together
> the gcc2 and EGCS CVS histories when converting from CVS to SVN. So some
> bits of the history around then may well look weird, but I don't see
> anything particularly odd about that particular SVN commit.)
Thank you, that is very valuable information to have.
There is probably some odd artifact at the merge point that confuses
my old code. If we are fortunate, the new code won't be confused.
The old code was brittle and had failures in weird places because I
started on branch analysis and handling of mixed-branch commits too
early. The new code essentially replays the dump operations into
commits without trying to do branch analysis or mixed-branch
resolution, then does those latter things in separate passes.
We'll know in a day or two, I think. The rewrite is done; I'm
troubleshooting some problems that I *think* are minor but which are
blocking merging to HEAD.
Once I get the new analyzer passing regressions I'll do a read-limited
conversion up to r14900 and see what's up.
--
<a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>
