On Thu, May 03 2018, Solene Rapenne <sol...@perso.pw> wrote:
> naddy@ founds in his latest bulk build that mailpile was failing due to
> python missing at do-install target.
duh
> sthen@ proposed to set an empty do-build target and remove NO_BUILD so
> the module lang/python will set the BUILD_DEPENDS variable correctly.
Sounds reasonable. It would have been nice to just set explicitely
"MODPY_BUILDDEP=Yes" and have that variable honored, but python.port.mk
doesn't work like that. ok jca@, but please see comments below.
> After this patch, `make print-build-depends` shows the following instead
> of nothing :
>
> This port requires package(s) "bzip2-1.0.6p8 metaauto-1.0p1 xz-5.2.3p0
> help2man-1.47.6 autoconf-2.69p2 libiconv-1.14p3 gettext-0.19.8.1p1
> gdbm-1.14.1 sqlite3-3.23.0 tcl-8.5.19p1 tk-8.5.19p1 db-4.6.21p5v0
> libffi-3.2.1p4 python-2.7.14p1" to build.
>
>
> Index: Makefile
> ===================================================================
> RCS file: /cvs/ports/mail/mailpile/Makefile,v
> retrieving revision 1.2
> diff -u -p -r1.2 Makefile
> --- Makefile 27 Apr 2018 08:31:01 -0000 1.2
> +++ Makefile 3 May 2018 17:08:55 -0000
> @@ -5,7 +5,7 @@ PKGNAME = ${DISTNAME:L}
> GH_ACCOUNT = mailpile
Patch doesn't apply (but I'm not sure it's worth sending another patch).
> GH_PROJECT = Mailpile
> GH_TAGNAME = 1.0.0rc2
> -REVISION = 0
> +REVISION = 1
BUILD_DEPENDS don't end up in the resulting package thus no need
to bump REVISION, right?
> CATEGORIES = mail
>
> @@ -26,11 +26,13 @@ RUN_DEPENDS = devel/py-fasteners \
> TEST_DEPENDS = shells/bash \
> ${RUN_DEPENDS}
>
> -NO_BUILD = Yes
> -
> PKG_ARCH = *
>
> INSTDIR = ${PREFIX}/share/mailpile
> +
> +# Nothing to build but allows MODPY to define BUILD_DEPENDS correctly
> +do-build:
>
> do-install:
> ${INSTALL_DATA_DIR} ${INSTDIR}
>
--
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
