Comment #7 on issue 2719 by [email protected]: Patch: Brain surgery on
the build system, first stage
http://code.google.com/p/lilypond/issues/detail?id=2719
Reply to Comment #4
Patchy doesn't clean the tree enough to remove generated files ignored by
Git, and aclocal.m4 is one of them without my patch; hence Patchy is not
able to test this patch, so this label needs-work has a limited value. What
about throwing in a "git clean -f -d -x" (yes, a lowercase "x") in Patchy
between baseline build and the application of the patch?
Reply to Comment #5
I'm sure Graham remembers the repeated pain of messing with discrepancies
between files added to Git and files taken into account by the bunch of
make variables that set which source files go rolled in the tarball, often
because translators and the translation meister (it was me some time ago)
didn't take care of this well. So I'm sure Graham wants such changes go
into master, and in the future stable/2.16 as well in case he rolls
releases from this branch as well; here are ten occurrences of such
breakage from the archives (not all related to translations):
http://lists.gnu.org/archive/html/lilypond-devel/2012-06/msg00122.html
http://lists.gnu.org/archive/html/lilypond-devel/2012-05/msg00015.html
http://lists.gnu.org/archive/html/lilypond-devel/2012-03/msg00606.html
http://lists.gnu.org/archive/html/lilypond-devel/2012-02/msg00480.html
http://lists.gnu.org/archive/html/lilypond-devel/2011-11/msg00360.html
http://lists.gnu.org/archive/html/lilypond-devel/2010-07/msg00089.html
http://lists.gnu.org/archive/html/lilypond-devel/2010-07/msg00061.html
http://lists.gnu.org/archive/html/lilypond-devel/2009-12/msg00628.html
http://lists.gnu.org/archive/html/lilypond-devel/2009-11/msg00346.html
http://lists.gnu.org/archive/html/lilypond-devel/2009-10/msg00025.html
http://lists.gnu.org/archive/html/lilypond-devel/2007-05/msg00053.html
http://lists.gnu.org/archive/html/lilypond-devel/2007-03/msg00076.html
In addition, AFAIR it's first time a so detailed testing procedure is
proposed to test a change to the build system; right, its first version
(sent on -devel list) was kind-of broken, but I tried to put a fair effort
in this, so that doubts about this large patch might be more easily cleared.
This is stage 1 of large patches to the build system, indeed, but it can go
to stable/2.16 without other stages, which address issues other than
rolling release tarball.
