Re: Lilypond taking forever to typeset

[Mojca Miklavec]

On 21 July 2016 at 09:14, David Kastrup wrote:
> Werner LEMBERG writes:
>
>>>> It would be good if we could make use of some sort of callback or
>>>> anything in Freetype2 (?) when it knows it is likely to spend a lot
>>>> of time like this.
>>>
>>> I think this should be rather done in fontconfig...
>>>
>>>> Or if we can predict the "cache does not exist" situation reliably,
>>>> we could trigger the cache rebuild explicitly with a suitable pair
>>>> of messages around it?
>>>
>>> I will ask on the fontconfig list how to detect that.
>>
>> Here's my question to the fontconfig developers.
>>
>>   [...] how can an application find out that the font cache creation
>>   (or update) is very time consuming?  Is it at least possible to
>>   catch the situation of a first run?
>>
>> And here's the answer.
>>
>>   current implementation assumes the users will runs fc-cache to
>>   create caches before running. as such APIs isn't available at this
>>   point, so no, it isn't.  though we discussed a bit these days, if it
>>   is wrong assumption, we may need to think about another way IMHO.
>
> They obviously _know_ themselves that this is a wrong assumption or they
> wouldn't have code (re-)creating the cache on the fly in the first
> place.

I agree that one should insist (I wouldn't mind if there was "another
way" if that way made more sense).

>> Maybe the Windows installer of lilypond can execute fc-cache?
>
> With an appropriate warning, This would likely help against the "this
> doesn't work, let's delete it again" first time effect.  It won't help
> much against future occurences even though the initial warning may point
> out that this might reoccur even after installation.

And there are Mac users as well. But I admit that I would feel at
least slightly uncomfortable having to run an installer (.pkg that
usually asks for admin password and can basically do anything on the
system) as opposed to just extracting the tarball. That said (and
getting off-topic, sorry), I believe that for Mac one should:
- provide 64-bit binaries
- provide a .dmg file instead of or in addition to .tar.bz2
- (provide some user-friendly way to run midi2ly :)

By shipping .tar.bz2 files instead of .dmg you might be excluding some
less experienced users. Not that this is you main target audience, but
I know many TeX users who are afraid of the terminal and cannot work
without the GUI editor that calls TeX for them. The situation might be
similar for LilyPond.

(Weird enough I also experienced some cache (re?)generation even with
the lilypond installed from MacPorts, but it was in fact a much
quicker one. So it was probably regeneration rather than first run.)

Mojca