Re: [PATCH] Update & announcement

[John Mandereau]

Le jeudi 10 décembre 2009 à 03:44 +0000, Graham Percival a écrit :
> On Wed, Dec 09, 2009 at 11:11:20AM +0100, John Mandereau wrote:
> > 4 lines of Texinfo multiplied by the number of languages, which
> > completely changes the deal.
> 
> It's still only a few minutes.

This can hardly be done properly in a few minutes for all languages
without error, I'm sure this takes more than 20 minutes.


> If the change is done badly, then the docs stop building.

... as any other doc edit.


> It would be nice if we *didn't* have to require a specific
> versioin of texi2html.

I don't understand this remark: what the mao can we do about texi2html
customization interface being a moving target, after having decided to
require CVS snaphots while at the same time much development has been
done on texi2html to allow it replace makeinfo?


> We'd have to tell writers "use @lilysection{foo} to start a new
> section, unless you want to have a different title instead of the
> node name, in which case do @node{foo} @section{foo}".

As all other existing docs would look like this, this wouldn't be such a
big deal.


> Look, lilypond does *not* have a good record of "clever,
> time-saving" build stuff.  Yesterday I had to spent 15 minutes
> figuring out how to add ja to the tarball for 2.12.3, because
> somebody thought it would be cute to get the language list from
> python/something.py, and they'd commented out the "ja" from line
> 123 in that file.

Sorry, but release 2.12 happened by surprise while I was integrating
Japanese docs.  I've never been able to build GUB, so I didn't and still
don't have an easy way to run an equivalent of dist-check.


> Take Trevor -- by any stretch of the imagination, he's an ideal
> contributor/developer.  Can he build the docs?  No.  Ok, maybe
> this is a one-off error... but could he identify the problem?  No.
> Having asked us about the problem, could well tell me?  Not as far
> as I know; the only way to debug build errors is to skim through
> literally hundreds of lines of warnings and errors, and look for
> some unfamiliar warnings+errors.  That is *incredibly* frustrating
> for people who want to help out.

Han-Wen and/or I already told you that if we want sources that always
compile for doc editors, we must create a dedicated branch.  If you
don't want to listen to this argument, I don't see the point of
complaining on the inherent unstability of master branch.


> You're about to say "yeah, but this is a really simple thing that
> has nothing to do with the doc-building problems".  I call BS.
> It's yet one more complication.  And again, we have a *terrible*
> record of "oh, this will make things easier" garbage that doesn't
> work.  By your own admission, you're going to be busy working on
> your phd soon.

If you really want to know, I'm already busy working on my PhD, but I
just realized that mastering the bureaucracy for setting the PhD joint
supervision between France and Italy and settling down in Italy, take
almost more energy.  As things go, I might finally have a health
insurance in January, allowing me to be Italian residence, which I must
do before February to have the right to stay in Italy.


>   That means that *I'll* be left explaining things
> to new contributors, trying to fix whatever build problems there
> are, trying to make it work in texi2html 1.84 or 2.0 or whatever
> number they give it when they intergrate it into texinfo... etc.

No, I already told that I'll still be available, even after January, to
reply on the mailing lists, but not to the extent of contributing .  I
have 5 to 15 hours to spend for LilyPond a week, and most of this time
has been eaten on the mailing lists (I strive to maximize the accuracy
of my replies, which takes time) and integrating translations, so I had
to left over work on the build system.


> If the waf system was working... mao, if anybody was *working* on
> the waf system... and it had a good logging system for the
> doc-builds, good warnings, etc... then I'd consider it.  If the
> current build system was working, and had been working continually
> for the past 3 months, I'd consider it.  Oh, and what happened to
> the plan of merging the init files?  Has that also been abandoned?

I was just answering a request from Dénes, I never proposed to implement
myself in the coming days, but it seemed important enough to me to bring
this publicly.


> I don't think I can express how unhappy I am with the build
> situation without going off into a huge curse-filled rant that
> wouldn't do anybody any good.  So let's just pretend I did that,
> and know that I am /seriously/ pissed off at the stepmake system,
> the python scripts, the translations, and everybody who worked on
> this stuff in the past.

If you actually went into such a rant, this wouldn't make our docs build
system better or worse in any way.  The translations are solidly
inegrated in LilyPond, and we have the build system we have, there is no
magic wand, nor any fairy that could hear your rant and then solve
issues.

Best,
John

signature.asc
Description: Ceci est une partie de message numériquement signée