Re: [gentoo-dev] [RFC] New eclass: mate

[Michał Górny]

On Fri, 10 Jun 2016 12:52:40 -0400
NP-Hardass <NP-Hardass@g.o> wrote:

> On 06/10/2016 06:43 AM, Michał Górny wrote:
> > Dnia 9 czerwca 2016 14:19:43 CEST, NP-Hardass <NP-Hardass@g.o> napisał(a):
> >> 	DEPEND="app-arch/xz-utils"
> >> fi
> >>
> >> # @ECLASS-VARIABLE: MATE_TARBALL_SUFFIX
> >> # @INTERNAL
> >> # @DESCRIPTION:
> >> # All projects hosted on mate-desktop.org provide tarballs as tar.xz.
> >> # Undefined in live ebuilds.
> >> [[ ${PV} != 9999 ]] && : ${MATE_TARBALL_SUFFIX:="xz"}
> >
> > You should ask upstream to supply .tar.lz instead, so they don't harm their users.
> >
> I can ask, but I don't think I understand what you mean by "harm their
> users."

It's an internal joke. But I can't find the reference quote right now.

> >> # @ECLASS-VARIABLE: MATE_DESKTOP_ORG_PN
> >> # @DESCRIPTION:
> >> # Name of the package as hosted on mate-desktop.org.
> >> # Leave unset if package name matches PN.
> >> : ${MATE_DESKTOP_ORG_PN:=$PN}
> >>
> >> # @ECLASS-VARIABLE: MATE_DESKTOP_ORG_PV
> >> # @DESCRIPTION:
> >> # Package version string as listed on mate-desktop.org.
> >> # Leave unset if package version string matches PV.
> >> : ${MATE_DESKTOP_ORG_PV:=$PV}
> >>
> >> # @ECLASS-VARIABLE: MATE_BRANCH
> >> # @DESCRIPTION:
> >> # Major and minor numbers of the version number, unless live.
> >> # If live ebuild, will be set to '9999'.
> >> if [[ ${PV} == 9999 ]]; then
> >> 	: ${MATE_BRANCH:=9999}
> >> else
> >> 	: ${MATE_BRANCH:=$(get_version_component_range 1-2)}
> >> fi
> >
> > Unless I'm missing something, the 1-2 range will return 9999 in live version as well. Unless you are trying to avoid the inherit in the live ebuild -- then I don't think it's worth the extra code.
> >
> I will double check and reply again if it is different, but I believe
> the concern had been that if there was a major change in a 9999, and I
> temporarily revbumped the 9999 to 9999-r1, the 1-2 range would return
> 9999-r1 rather then the expected 9999.

Errr, but PV shouldn't contain '-r1'...

> >> # Set SRC_URI or EGIT_REPO_URI based on whether live
> >> if [[ ${PV} == 9999 ]]; then
> >> 	EGIT_REPO_URI="
> >> 		https://github.com/mate-desktop/${MATE_DESKTOP_ORG_PN}.git
> >> 		git://github.com/mate-desktop/${MATE_DESKTOP_ORG_PN}.git
> >> 		http://github.com/mate-desktop/${MATE_DESKTOP_ORG_PN}.git
> >
> > Does github actually support http?
> >
> git clone works, though, I'm unsure if it is redirecting to https.  I'll
> have to double check on this.

Please let me know what you find out.

> >> case "${EAPI:-0}" in
> >> 	5|6)
> >> 		EXPORT_FUNCTIONS src_prepare src_configure src_install pkg_preinst
> >> pkg_postinst pkg_postrm
> >> 		;;
> >> 	*) die "EAPI=${EAPI} is not supported" ;;
> >
> > It is already dead after the first check.
> >
> If I drop EAPI 5 as suggested by perfinion, would you recommend keeping
> this in a case statement for when future EAPIs are added?

No, you can add a case when you need it.

> >> }
> >>
> >> # @FUNCTION: mate_src_prepare
> >> # @DESCRIPTION:
> >> # Call gnome2_src_prepare to handle environment setup and patching,
> >> then
> >> # call eautoreconf if necessary
> >> mate_src_prepare() {
> >> 	debug-print-function ${FUNCNAME} "$@"
> >>
> >> 	local force_autoreconf=${FORCE_AUTORECONF:-false}
> >> 	[[ ${PV} == 9999 ]] && force_autoreconf=true
> >>
> >> 	gen_chksum() {
> >> 		find '(' -name 'Makefile.am' \
> >> 			-o -name 'configure.ac' \
> >> 			-o -name 'configure.in' ')' \
> >> 			-exec cksum {} + | sort -k2
> >> 	}
> >>
> >> 	local chksum=$(gen_chksum)
> >>
> >> 	gnome2_src_prepare "$@"
> >>
> >> 	if ${force_autoreconf}
> >
> > Don't execute random user-provided data.
> >
> Not sure I understand.  force_autoreconf is a local variable that is
> either when the ebuild explicitly says "I want to eautoreconf," when it
> is a 9999, as upstream does not provide configured sources in git, or
> when patching affects the build system, requiring an autoreconf.

But you shouldn't do it the PHP way. Parse the value, not execute it.
If one does FORCE_AUTORECONF=1, it won't work. If one does
FORCE_AUTORECONF='rm -rf /', it won't work!

-- 
Best regards,
Michał Górny
<http://dev.gentoo.org/~mgorny/>