Re: Hardcoded LP version in *2ly scripts?

[Graham Percival]

On Wed, Dec 22, 2010 at 08:52:44PM -0700, Colin Campbell wrote:
> I had occasion to try and run convert-ly (from Frescobaldi, but
> LIlyPondTool was the same) over a piece I'd opened in Denemo then saved
> as a .ly file. Convert-ly promptly threw an error about not finding
> lilylib.  After a bit of a look around, I found that convert-ly had
> version 2.13.30 hardwired into a piece of code which looks as though it
> is appending to a PATH.

I'm guessing that the last GUB you installed was 2.13.30?

> Changing the version by hand ( to 2.13.44)
> allowed convert-ly to run.  I also manually deleted all trace of
> lilypond, did a fresh git clone and rebuilt; the python scripts now have
> the current version. I haven't tested the case of uninstalling the
> outdated version, then installing the new build.

Hmm.  This sounds weird, but at the moment I don't think it's
unreasonable to suspect a discrepancy in the GUB install vs. make
install.  I'm not saying that you didn't see some kind of a
problem, but if convert-ly as compiled in GUB was broken, I would
expect to have seen tons of reports about it.

If you're particularly curious about this, then go ahead and look
into precisely what path the convert-ly you're running is in, and
precisely what PYTHONPATH or PYTHONHOME it's using, and precisely
what's in those directories... but I don't see this being a good
effort-vs-payoff exchange.


PS virtually nobody runs "make install", and I certainly wouldn't
expect a new contributor to do this.  The only people running
"make install" should be linux package creators, and extremely
advanced users who are both very familiar with unix software and
too impatient to wait a week or two until the next devel release.

I will clarify this in the next revision of the "quick start"
instructions, for which I will hopefully feel healthy enough to
write tomorrow.

Cheers,
- Graham