Re: GSoC 2020: make hangs when making fonts

[Owen Lamb]

All right, whatever has broken 2.21.X for me broke 2.18.0, too, so I don't
think bisecting will be much of a help. However, make's output was
apparently more verbose back then, so I got a little more info before and
after the copyright notice this time. Some sort of "overlap" errors in
Fontforge are causing it to hang:

Converting PostScript graphics to Type 1 font programs...

Invoking "t1asm feta20.pt1 feta20.pfb"...

Invoking "fontforge -script feta20.pe feta20.pfb"...
[The copyright notice we've already seen]
Internal Error (overlap) in scripts.reverseturn: Fixup error 2 in
MergeIntersections.
Internal Error (overlap) in scripts.reverseturn: Could not find
finalmInternal Error (overlap) in scripts.turn: Fixup error 2 in
MergeIntersections.
Internal Error (overlap) in scripts.turn: Could not find finalm

Ctrl-C'ing yields the same output as before, plus a success message:
*** Successfully generated feta20.pfb! ***
(The file feta20.pfb does indeed exist, and can be viewed in the
build/mf/out, in a temporary folder. It contains all the glyphs included in
feta20.mf, no more, no less. So, no noteheads or flags, etc. The glyphs are
not assigned Unicode codepoints, but that's not surprising. This subfont is
supposed to be temporary and would ordinarily be merged with the other
subfonts with gen-emmentaler-fontforge.py, after which they would receive
Unicode values.)

I also ran make on current master with VERBOSE=1, and got the same stuff as
2.18.0 *before* the copyright notice, but none of 2.18.0's stuff *after* it.

At any rate, that *probably* pins down Fontforge as the source of the
issues. It's odd, though--an update couldn't have introduced the bug, since
a version newer than 2017 isn't available on LilyDev's apt repository. And
the bug couldn't have just been there all along, since it's worked
perfectly for me for over a month.

I'd really like to see if someone can reproduce any of this on a different
machine and a fresh LilyDev VM, so please holler if you're willing to try!

(Also, Jean, I did try git clean -xdf. Forgot to mention that in my last
message, sorry.)

On Tue, Jun 30, 2020 at 3:36 PM Owen Lamb <owendlamb@gmail.com> wrote:

> On Tue, Jun 30, 2020 at 3:04 PM Jean Abou Samra <jean@abou-samra.fr>
> wrote:
>
>> So git clean didn't work? At this point, unfortunately I can't help
>> much--my
>> previous message was merely a shot in the dark but I understand almost
>> nothing
>> about fonts. Also, I'll be going on vacation pretty soon, so I won't have
>> time
>> to set up LilyDev, sorry…
>>
>
> No worries. Have a great vacation!
>
>> That said, if CI itself started to fail recently, there must be something
>> broken
>> in the sources, possibly in relation with a specific environement. Could
>> you try
>> to compile after a git checkout release/2.18.0-1? If that succeeds, you
>> want to
>> go through a git bisect to find the culprit.
>>
> That's a good idea--I'll try that next.
>
>> PS: I guess https://gitlab.com/lilypond/lilypond/-/merge_requests/205 is
>> going
>> to help in debugging.
>>
> This page is all Greek to me at the moment, but I'll look into it. Thanks!
>