abort: malformed response from CVS

[Matt Mackall]

On Fri, 2014-05-09 at 14:26 -0700, Markus Mayer wrote:
> On 8 May 2014 12:19, Sean Farley <sean.michael.farley at gmail.com> wrote:
> >
> > Markus Mayer <code at mmayer.net> writes:
> >
> >> On 8 May 2014 05:16, Matt Mackall <mpm at selenic.com> wrote:
> >>> On Wed, 2014-05-07 at 17:08 -0700, Markus Mayer wrote:
> >>>>
> >>>> Any tips how I could track this down further?
> >>>
> >>> What versions of CVS are you using?
> >>
> >> I had to upgrade the CVS on the Macs. It was 1.11.23 before, and that
> >> caused other problems much earlier on. With the upgrade I got around
> >> those issues. Now it's 1.12.13 everywhere. That's what I used when I
> >> got the results I reported here.
> >>
> >> On both Macs (i.e. the same version works on one and not the other, so
> >> it's likely not CVS causing the problems):
> >
> > Drive-by comment:
> >
> > I've seen similar errors on Ubuntu with the same CVS version while I was
> > playing with converting the NetBSD repo. I can get into the debugger and
> > see that it aborts around cvs.py:250 where the code is testing the
> > beginning of the line.
> >
> > But ... that does little to help me debug because I have almost zero
> > knowledge of how CVS works and what it should be spitting out.
> 
> FWIW, it looks like the issue was caused by a stale ~/.hg.cvsps/
> directory (or rather the cache inside). I found it, removed it -- and
> everything started working as expected. No issues with case
> insensitive file systems or anything like that.
> 
> Is "hg convert" supposed to delete the cache after it's done? If a
> stale cache can cause issues like this it seems like the cache
> shouldn't be left behind.

The cache is intended to be used from run to run for incremental
conversion. Did you change the history of the CVS repo after the cache
was created and thereby make the cache wrong?

-- 
Mathematics is the supreme nostalgia of our time.