Hi,

Please review tree-sitter-grammar.eclass - a new eclass for building tree-sitter grammars provided by the tree-sitter developers.

The grammars are shipped as one or two source files and with no build
system. Upstream uses Gyp to build the grammars as a NodeJS module,
but they are useful standalone to other applications such as text
editors.

Some new packages utilising the eclass can be viewed on GitHub: https://github.com/gentoo/gentoo/pull/22611

Thanks,
Matthew

---
eclass/tree-sitter-grammar.eclass | 94 +++++++++++++++++++++++++++++++
1 file changed, 94 insertions(+)
create mode 100644 eclass/tree-sitter-grammar.eclass

diff --git a/eclass/tree-sitter-grammar.eclass b/eclass/tree-sitter-grammar.eclass
new file mode 100644
index 00000000000..a107b0b1908
--- /dev/null
+++ b/eclass/tree-sitter-grammar.eclass
@@ -0,0 +1,94 @@
+# Copyright 1999-2021 Gentoo Authors
+# Distributed under the terms of the GNU General Public License v2
+
+# @ECLASS: tree-sitter-grammar.eclass
+# @MAINTAINER:
+# Matthew Smith <matt@offtopica.uk>
+# Nick Sarnie <sarnex@gentoo.org>
+# @AUTHOR:
+# Matthew Smith <matt@offtopica.uk>
+# @SUPPORTED_EAPIS: 8
+# @BLURB: Common functions and variables for Tree Sitter grammars
+
+if [[ -z ${_TREE_SITTER_GRAMMAR_ECLASS} ]]; then +_TREE_SITTER_GRAMMAR_ECLASS=1
+
+case ${EAPI} in
+ 8) ;;
+ *) die "EAPI=${EAPI:-0} is not supported" ;;
+esac
+
+inherit toolchain-funcs
+
+SRC_URI="https://github.com/tree-sitt

On Sun, 2021-10-17 at 11:42 +0100, Matthew Smith wrote:

Hi,

Please review tree-sitter-grammar.eclass - a new eclass for building tree-sitter grammars provided by the tree-sitter developers.

The grammars are shipped as one or two source files and with no build
system. Upstream uses Gyp to build the grammars as a NodeJS module,
but they are useful standalone to other applications such as text
editors.

Some new packages utilising the eclass can be viewed on GitHub: https://github.com/gentoo/gentoo/pull/22611

Thanks,
Matthew

---
eclass/tree-sitter-grammar.eclass | 94 +++++++++++++++++++++++++++++++
1 file changed, 94 insertions(+)
create mode 100644 eclass/tree-sitter-grammar.eclass

diff --git a/eclass/tree-sitter-grammar.eclass b/eclass/tree-sitter-grammar.eclass
new file mode 100644
index 00000000000..a107b0b1908
--- /dev/null
+++ b/eclass/tree-sitter-grammar.eclass
@@ -0,0 +1,94 @@
+# Copyright 1999-2021 Gentoo Authors
+# Distributed under the terms of the GNU General Public License v2
+
+# @ECLASS: tree-sitter-grammar.eclass
+# @MAINTAINER:
+# Matthew Smith <matt@offtopica.uk>
+# Nick Sarnie <sarnex@gentoo.org>
+# @AUTHOR:
+# Matthew Smith <matt@offtopica.uk>
+# @SUPPORTED_EAPIS: 8
+# @BLURB: Common functions and variables for Tree Sitter grammars
+
+if [[ -z ${_TREE_SITTER_GRAMMAR_ECLASS} ]]; then +_TREE_SITTER_GRAMMAR_ECLASS=1
+
+case ${EAPI} in
+ 8) ;;
+ *) die "EAPI=${EAPI:-0} is not supported" ;;
+esac
+
+inherit toolchain-funcs
+ +SRC_URI="https://github.com/tree-sitter/${PN}/archive/${TS_PV:-v${PV}}.tar.gz

${P}.tar.gz"

TS_PV is not documented. Also, this line asks for wrapping.

+S="${WORKDIR}"/${PN}-${TS_PV:-${PV}}/src
+
+# Needed for tree_sitter/parser.h
+DEPEND="dev-libs/tree-sitter"
+
+EXPORT_FUNCTIONS src_compile src_install
+
+# @FUNCTION: get_tsg_abi_ver
+# @DESCRIPTION:
+# This internal function determines the ABI version of a grammar
library based

Then mark it with @INTERNAL? Also prefixing it with '_' would be a good
idea.

+# on the package version.
+get_tsg_abi_ver() {
+ if ver_test -gt 0.21; then
+ die "Grammar too new; unknown ABI version"
+ elif ver_test -ge 0.19.0; then
+ echo 13
+ else
+ die "Grammar too old; unknown ABI version"
+ fi
+}
+
+# @FUNCTION: tree-sitter-grammar_src_compile
+# @DESCRIPTION:
+# Compiles the Tree Sitter parser as a shared library. +tree-sitter-grammar_src_compile() {
+ debug-print-function $FUNCNAME $*

Please keep the prologue consistent with our coding style, i.e.:

debug-print-function ${FUNCNAME} "${@}"

+
+ # Grammars always contain parser.c, and sometimes a scanner.c,
+ # or scanner.cc.
+
+ $(tc-getCC) ${CFLAGS} \
+ ${CPPFLAGS} \
+ -fPIC \

How about exporting CC etc., and using emake to build them?

i.e. something like

tc-export CC
export CFLAGS="${CFLAGS} -fPIC"
emake parser.o

(also I think it would be less confusing to use the same directory for
output file)

+ -c "${S}"/parser.c \
+ -o "${WORKDIR}"/parser.o || die
+
+ local link=$(tc-getCC)
+
+ if [[ -f "${S}/scanner.c" ]]; then
+ $(tc-getCC) ${CFLAGS} \
+ ${CPPFLAGS} \
+ -fPIC \
+ -c "${S}"/scanner.c \
+ -o "${WORKDIR}"/scanner.o || die
+ elif [[ -f "${S}/scanner.cc" ]]; then
+ $(tc-getCXX) ${CXXFLAGS} \
+ ${CPPFLAGS} \
+ -fPIC \
+ -c "${S}"/scanner.cc \
+ -o "${WORKDIR}"/scanner.o || die
+ link=$(tc-getCXX)
+ fi
+
+ local soname=lib${PN}.so.$(get_tsg_abi_ver)

We've got some helpers like get_libname to make this a little bit more portable, or rather to introduce less obstacles when people want to fix portability issues.

+ ${link} ${LDFLAGS} \
+ -shared \
+ "${WORKDIR}"/*.o \
+ -Wl,-soname ${soname} \
+ -o "${WORKDIR}"/${soname} || die
+}
+
+# @FUNCTION: tree-sitter-grammar_src_install
+# @DESCRIPTION:
+# Installs the Tree Sitter parser library. +tree-sitter-grammar_src_install() {
+ debug-print-function $FUNCNAME $*
+
+ dolib.so "${WORKDIR}"/lib${PN}.so.$(get_tsg_abi_ver)
+ dosym lib${PN}.so.$(get_tsg_abi_ver) /usr/$(get_libdir)/lib${PN}.so +}
+fi
--
2.33.0

--
Best regards,
Michał Górny

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi,

Updated with feedback from mgorny: documenting TS_PV, using emake
to build objects, and using get_libname to fix so/dylib
confusion.

Thanks again,
Matthew

---
eclass/tree-sitter-grammar.eclass | 96 +++++++++++++++++++++++++++++++
1 file changed, 96 insertions(+)
create mode 100644 eclass/tree-sitter-grammar.eclass

diff --git a/eclass/tree-sitter-grammar.eclass b/eclass/tree-sitter-grammar.eclass
new file mode 100644
index 00000000000..fc18ac203e1
--- /dev/null
+++ b/eclass/tree-sitter-grammar.eclass
@@ -0,0 +1,96 @@
+# Copyright 1999-2021 Gentoo Authors
+# Distributed under the terms of the GNU General Public License v2
+
+# @ECLASS: tree-sitter-grammar.eclass
+# @MAINTAINER:
+# Matthew Smith <matt@offtopica.uk>
+# Nick Sarnie <sarnex@gentoo.org>
+# @AUTHOR:
+# Matthew Smith <matt@offtopica.uk>
+# @SUPPORTED_EAPIS: 8
+# @BLURB: Common functions and variables for Tree Sitter grammars
+
+if [[ -z ${_TREE_SITTER_GRAMMAR_ECLASS} ]]; then +_TREE_SITTER_GRAMMAR_ECLASS=1
+
+case ${EAPI} in
+ 8) ;;
+ *) die "EAPI=${EAPI:-0} is not supported" ;;
+esac
+
+inherit multilib toolchain-funcs
+
+SRC_URI="https://github.com/

On Sun, 2021-10-17 at 12:31 +0100, Matthew Smith wrote:

Hi,

Updated with feedback from mgorny: documenting TS_PV, using emake
to build objects, and using get_libname to fix so/dylib
confusion.

Thanks again,
Matthew

---
eclass/tree-sitter-grammar.eclass | 96 +++++++++++++++++++++++++++++++
1 file changed, 96 insertions(+)
create mode 100644 eclass/tree-sitter-grammar.eclass

diff --git a/eclass/tree-sitter-grammar.eclass b/eclass/tree-sitter-grammar.eclass
new file mode 100644
index 00000000000..fc18ac203e1
--- /dev/null
+++ b/eclass/tree-sitter-grammar.eclass
@@ -0,0 +1,96 @@
+# Copyright 1999-2021 Gentoo Authors
+# Distributed under the terms of the GNU General Public License v2
+
+# @ECLASS: tree-sitter-grammar.eclass
+# @MAINTAINER:
+# Matthew Smith <matt@offtopica.uk>
+# Nick Sarnie <sarnex@gentoo.org>
+# @AUTHOR:
+# Matthew Smith <matt@offtopica.uk>
+# @SUPPORTED_EAPIS: 8
+# @BLURB: Common functions and variables for Tree Sitter grammars
+
+if [[ -z ${_TREE_SITTER_GRAMMAR_ECLASS} ]]; then +_TREE_SITTER_GRAMMAR_ECLASS=1
+
+case ${EAPI} in
+ 8) ;;
+ *) die "EAPI=${EAPI:-0} is not supported" ;;
+esac
+
+inherit multilib toolchain-funcs
+ +SRC_URI="https://github.com/tree-sitter/${PN}/archive/${TS_PV:-v${PV}}.tar.gz
+ -> ${P}.tar.gz"
+S="${WORKDIR}"/${PN}-${TS_PV:-${PV}}/src
+
+# Needed for tree_sitter/parser.h
+DEPEND="dev-libs/tree-sitter"
+
+EXPORT_FUNCTIONS src_compile src_install
+
+# @ECLASS-VARIABLE: TS_PV
+# @PRE_INHERIT
+# @DEFAULT_UNSET
+# @DESCRIPTION:
+# Used to override upstream tag name if tagged differently, e.g. most releases
+# are v${PV} but some are tagged as rust-${PV}.
+
+# @FUNCTION: _get_tsg_abi_ver
+# @INTERNAL
+# @DESCRIPTION:
+# This internal function determines the ABI version of a grammar
library based
+# on the package version.
+_get_tsg_abi_ver() {
+ if ver_test -gt 0.21; then
+ die "Grammar too new; unknown ABI version"
+ elif ver_test -ge 0.19.0; then
+ echo 13
+ else
+ die "Grammar too old; unknown ABI version"
+ fi
+}
+
+# @FUNCTION: tree-sitter-grammar_src_compile
+# @DESCRIPTION:
+# Compiles the Tree Sitter parser as a shared library. +tree-sitter-grammar_src_compile() {
+ debug-print-function ${FUNCNAME} "${@}"
+
+ # Grammars always contain parser.c, and sometimes a scanner.c,
+ # or scanner.cc.
+
+ tc-export CC CXX
+ export CFLAGS="${CFLAGS} -fPIC"
+ export CXXFLAGS="${CXXFLAGS} -fPIC"
+
+ local objects=( parser.o )
+ if [[ -f "${S}"/scanner.c || -f "${S}"/scanner.cc ]]; then
+ objects+=( scanner.o )
+ fi
+ emake "${objects[@]}"
+
+ local link=$(tc-getCC)
+ if [[ -f "${S}/scanner.cc" ]]; then
+ link=$(tc-getCXX)
+ fi
+
+ local soname=lib${PN}$(get_libname $(_get_tsg_abi_ver))
+ ${link} ${LDFLAGS} \

I'm sorry for missing this before -- when linking, you also need to pass
CFLAGS or CXXFLAGS.

+ -shared \
+ *.o \
+ -Wl,-soname ${soname} \
+ -o "${WORKDIR}"/${soname} || die
+}
+
+# @FUNCTION: tree-sitter-grammar_src_install
+# @DESCRIPTION:
+# Installs the Tree Sitter parser library. +tree-sitter-grammar_src_install() {
+ debug-print-function ${FUNCNAME} "${@}"
+
+ dolib.so "${WORKDIR}"/lib${PN}$(get_libname $(_get_tsg_abi_ver))
+ dosym lib${PN}$(get_libname $(_get_tsg_abi_ver)) \
+ /usr/$(get_libdir)/lib${PN}$(get_libname)
+}
+fi
--
2.33.0

--
Best regards,
Michał Górny

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)