Re: 2.13.40 regtests

[Graham Percival]

On Fri, Nov 26, 2010 at 02:15:22AM -0500, Han-Wen Nienhuys wrote:
> The reason I did not do it originally is that it moves the comparison
> farther away from lilypond itself and pixel-per-pixel changes are not
> calibrated for the size of the symbols: a large symbol moving place
> will generate a much higher difference score than a small score.  It
> might have been a quicker way to get started with running comparison
> tests.

True, but at the moment there's no way to catch problems due to
ghostscript (or international fonts being installed / not being
installed in GUB).  Of course, hopefully GUB won't be changing all
that often, but it _does_ happen.
Would it be possible to have both kinds of regtest comparisons?
Maybe not enabled in the usual "make check", but adding a separate
"make pixelcheck" or something like that?

Cheers,
- Graham