Re: GUB lilypond build fails

[David Kastrup]

Andrew Bernard <address@hidden> writes:

> Hi All,
>
> I am new to lilypond dev. work. My intention may be naive, but I am wanting
> to do the work to uplift gub and lilypond to python 3. Am I premature, or
> foolish, or misguided? I did have some encouraging email about this
> previously, but I just wanted to check before I dive in and spend large
> amounts of time, especially re GUB.

I have yesterday asked Masamichi-san to take a look at our current GUB
problems but it would appear that Werner found and fixed the source of
the current roadblock.  Now it sounded like Masamichi-san was currently
rather short of time for this, so it would make sense that now, since
the immediate problem appears to be solved, you take a look at this
instead with a perspective of a Python 3 migration eventually.

So I did some "git blame" searching and it would appear that the
Python 2.5 "ism" was introduced with commit

commit 7b07440da921d979ab492fd284b6198152a8020c
Author:     Alexander Myltsev <address@hidden>
AuthorDate: Thu Jun 2 11:19:07 2016 +0300
Commit:     James Lowe <address@hidden>
CommitDate: Sat Jun 16 10:59:08 2018 +0100

    musicxml2ly: handle hidden time signatures.

Now I don't really feel that we can indeed pass any blame for not being
aware of having to use syntax for an ancient rather than an antique
version of Python.  And we apparently don't have the manpower in place
to figure out what went wrong: this put our release process to a halt
for almost half a year.

So it would be really _urgent_ to get GUB advanced to a recent version
of Python 2 in all supported platforms (and I think we can delist the
PPC platform support by now): the current Python 2 requirement is
obvious, the Python 2.4 requirement not.

Moving to a current Python 2 version should be doable without changing
the LilyPond source code: that's a pure GUB job, though one that might
be bothersome on some platforms.  Decommissioning the PPC platforms
would likely make this easier, and I don't think we really have active
users there.  Masamichi-san also suggested stopping FreeBSD support (I
don't know if there is much of an indication the binary installation
from us is used rather than compiling themselves, and supposedly they
can run Linux binaries though the runtime environment may be
troublesome) and Windows 32-bit support.  I am not sure whether there
aren't people running 32-bit Windows binaries in VMs though.

But at least PPC seems reasonably safe to retire, and possibly FreeBSD
if it helps.

-- 
David Kastrup