Re: lybook-db etc etc.

[David Kastrup]

Han-Wen Nienhuys <address@hidden> writes:

> On Wed, Oct 26, 2011 at 1:13 PM, David Kastrup <address@hidden> wrote:
>> Well, I am currently in the process of running make info (similar to
>> make doc), and this is totally silly.
>>
>> In my opinion, the whole lybook-db stuff needs to go.  Instead, Lilypond
>> is run _once_ for all snippets of a lybook source, generating _one_
>> PostScript file.  Then GhostScript is run _once_ to generate a bunch of
>> eps files, or a multi-page PDF file with all graphics in them which get
>> referenced as needed.
>>
>> Bleedover of variables/fonts/whatever is not all that crucial since we
>> are talking about a single document source rather than an immaculate
>> database.
>
> on the contrary: the regtest is a tely document, so bleed will cause
> spurious diffs in the regtest, at least in the pixel comparison.

Bleed would be a bug in the snippets.  Global settings would likely need
to be forked out into an initialization file for the whole document.

Anyway, I am more concerned about "make doc" rather than "make test".

To put a perspective to it, I'd figure I could justify working on this
if we managed a pledge of €500 for every factor of 2 I can shave off
"make doc" (starting after "make all" from a fresh checkout), getting
free hand for the interaction between lilypond-book, Lilypond, and
Ghostscript _using_ GNU/Linux (I don't have the means to port, say,
socket communication to Windows, though at first I would stick with
batch processing that should "just work"TM).  That would be for a
uniprocessing setup.  A separate agreement may be possible for
multiprocessing, but then the results would not be deterministic for
bleedover problems since the set of snippets processed by a single
Lilypond/Ghostscript pipeline would vary from run to run.

-- 
David Kastrup