Reinhold Kainhofer <reinh...@kainhofer.com> writes:
> On Fr., 23. Sep. 2011 09:55:26 CEST, David Kastrup <d...@gnu.org> wrote:
>> I am somewhat surprised that the patch passed testing by someone else,
>> though. It would seem that doc builds are not part of that.
>
> Nope, because it's the regtetsts' task, so only the regtests are
> checked. It's also a practical issue: On my machine a regtest run
> takes ~20 minutes, while a docs build takes about 3 hours, so it's
> impossible to check the docs build for every patch.
>
> If a regtest run succeeds, but a docs build fails, that's basically a
> bug in our regtests (and we have many missing regtests still).
In the current situation with the optional music function patch, there
is clearly buggy code in the documentation itself (or rather: it calls
functions in a way that is documented but not supported).
So this is something that can't be spotted with regtests, but only with
a documentation build.
> That's also why it's so important to have a proper regtest for each
> new feature.
It would not have helped at all here. For "make all", I already added
commit 31e79fa5f9d1b1af03ca3fc29ce049b39b485529
Author: David Kastrup <d...@gnu.org>
Date: Tue Sep 20 22:28:08 2011 +0200
define-markup-commands.scm: Fix bad parameter type for \on-the-fly
some time ago, again since it had been exposed by the better
type-checking.
I have been considering changing \with-link to use string-or-symbol? and
thus get the argument accepted. However, \label is clearly documented
as taking a symbol, so this is more likely a problem with the caller,
autogenerated code in the documentation.
--
David Kastrup
_______________________________________________
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel
