feat: ports - libhubbub, libcss, libdom

Co-Authored-By: Claude Opus 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_018Ae69wJS7sueQMUwNxV3nX

.gitmodules

@@ -13,3 +13,12 @@
 [submodule "dev-libs/libparserutils/upstream"]
 	path = dev-libs/libparserutils/upstream
 	url = https://github.com/netsurf-browser/libparserutils.git
+[submodule "net-libs/libhubbub/upstream"]
+	path = net-libs/libhubbub/upstream
+	url = https://github.com/netsurf-browser/libhubbub.git
+[submodule "net-libs/libcss/upstream"]
+	path = net-libs/libcss/upstream
+	url = https://github.com/netsurf-browser/libcss.git
+[submodule "net-libs/libdom/upstream"]
+	path = net-libs/libdom/upstream
+	url = https://github.com/netsurf-browser/libdom.git

Makefile

@@ -28,6 +28,9 @@ PORTS := \
     dev-libs/zlib \
     dev-libs/libwapcaplet \
     dev-libs/libparserutils \
+    net-libs/libhubbub \
+    net-libs/libcss \
+    net-libs/libdom \
     games-fps/doom
 
 # Bundle directories are keyed by port name, not category path.

dev-libs/libparserutils/Makefile

@@ -17,6 +17,8 @@ UPSTREAM := upstream
 PORT_NAME        := libparserutils
 PORT_LIB         := 1
 PORT_ARCHIVE     := libparserutils.a
+PORT_PC_IN       := $(UPSTREAM)/libparserutils.pc.in
+PORT_PC_VERSION  := 0.2.5
 PORT_HEADER_ROOT := $(UPSTREAM)/include
 PORT_HEADERS     := $(shell find $(UPSTREAM)/include -name '*.h' 2>/dev/null)
 

dev-libs/libwapcaplet/Makefile

@@ -12,6 +12,8 @@ UPSTREAM := upstream
 PORT_NAME        := libwapcaplet
 PORT_LIB         := 1
 PORT_ARCHIVE     := libwapcaplet.a
+PORT_PC_IN       := $(UPSTREAM)/libwapcaplet.pc.in
+PORT_PC_VERSION  := 0.4.3
 PORT_HEADER_ROOT := $(UPSTREAM)/include
 PORT_HEADERS     := $(UPSTREAM)/include/libwapcaplet/libwapcaplet.h
 

mk/port.mk

@@ -165,15 +165,24 @@ NS_MAKE_ARGS := \
 # archive rule tries to *execute* the .a file (exit 127).
 NS_MAKE := MAKEFLAGS= $(MAKE)
 
+# Their makefiles resolve dependencies through pkg-config, so point it at the
+# ports sysroot (prepended, so host packages still resolve for host tools).
+# PYTHON is forced to python3: libcss's codegen invokes $(PYTHON), and this
+# host has no bare `python`.
+NS_ENV := \
+	PKG_CONFIG_PATH="$(PORT_SYSROOT_LIB)/pkgconfig:$$PKG_CONFIG_PATH" \
+	PYTHON=python3
+
 # Build a NetSurf library in $(1) (its checkout dir) and copy the archive to
 # $(PORT_ARCHIVE). Usage in a recipe:  $(call ns_build,upstream,wapcaplet)
 # A port may add its own flags by setting NS_CFLAGS_EXTRA (expanded at call
 # time, so it can be set after this file is included).
 ns_build = \
-	CFLAGS="$(NS_CFLAGS) $(NS_CFLAGS_EXTRA)" $(NS_MAKE) -C $(1) $(NS_MAKE_ARGS) && \
+	CFLAGS="$(NS_CFLAGS) $(NS_CFLAGS_EXTRA)" $(NS_ENV) \
+	$(NS_MAKE) -C $(1) $(NS_MAKE_ARGS) $(NS_MAKE_ARGS_EXTRA) && \
 	cp $(1)/$(NS_BUILDDIR)/lib$(2).a $(PORT_ARCHIVE)
 
-ns_clean = $(NS_MAKE) -C $(1) $(NS_MAKE_ARGS) clean >/dev/null 2>&1 || true
+ns_clean = $(NS_ENV) $(NS_MAKE) -C $(1) $(NS_MAKE_ARGS) clean >/dev/null 2>&1 || true
 
 ifeq ($(PORT_LIB),1)
 
@@ -197,13 +206,29 @@ $(PORT_SYSROOT_INC)/%: $(PORT_HEADER_ROOT)/%
 	@mkdir -p $(dir $@)
 	cp $< $@
 
+# Optional pkg-config file. NetSurf libraries locate their dependencies with
+# pkg-config (REQUIRED_PKGS), so a library that others depend on must publish
+# one into the sysroot or those builds fail to find it. Their .pc.in files use
+# literal PREFIX/LIBDIR/INCLUDEDIR/VERSION placeholders, substituted here the
+# same way their install step would.
+ifneq ($(PORT_PC_IN),)
+INSTALLED_PC := $(PORT_SYSROOT_LIB)/pkgconfig/$(PORT_NAME).pc
+
+$(INSTALLED_PC): $(PORT_PC_IN) Makefile
+	@mkdir -p $(dir $@)
+	sed -e 's|PREFIX|$(PORT_SYSROOT)|' \
+	    -e 's|LIBDIR|lib|' \
+	    -e 's|INCLUDEDIR|include|' \
+	    -e 's|VERSION|$(PORT_PC_VERSION)|' $< > $@
+endif
+
 .PHONY: install-lib
-install-lib: $(INSTALLED_LIB) $(INSTALLED_HEADERS)
+install-lib: $(INSTALLED_LIB) $(INSTALLED_HEADERS) $(INSTALLED_PC)
 	@echo "installed lib: $(PORT_NAME) -> $(PORT_SYSROOT)"
 
 .PHONY: clean-lib
 clean-lib:
-	rm -rf $(OBJDIR) $(PORT_ARCHIVE) $(INSTALLED_LIB) $(INSTALLED_HEADERS)
+	rm -rf $(OBJDIR) $(PORT_ARCHIVE) $(INSTALLED_LIB) $(INSTALLED_HEADERS) $(INSTALLED_PC)
 
 else
 

net-libs/libcss/Makefile

@@ -0,0 +1,34 @@
+# libcss for MontaukOS
+#
+# Upstream: https://github.com/netsurf-browser/libcss (official mirror of
+# git.netsurf-browser.org; submodule at upstream/, pinned, no local patches).
+#
+# CSS parser and selection engine. Depends on libparserutils and
+# libwapcaplet, both of which must already be in the ports sysroot.
+#
+# Needs perl AND python3 on the BUILD host: it generates its property
+# parser/computed-style tables at build time. mk/port.mk forces PYTHON=python3
+# because the codegen invokes $(PYTHON) and Debian has no bare `python`.
+
+UPSTREAM := upstream
+
+PORT_NAME        := libcss
+PORT_LIB         := 1
+PORT_DEPENDS     := libparserutils libwapcaplet
+PORT_ARCHIVE     := libcss.a
+PORT_PC_IN       := $(UPSTREAM)/libcss.pc.in
+PORT_PC_VERSION  := 0.9.2
+PORT_HEADER_ROOT := $(UPSTREAM)/include
+PORT_HEADERS     := $(shell find $(UPSTREAM)/include -name '*.h' 2>/dev/null)
+
+include ../../mk/port.mk
+
+.PHONY: all clean
+all: install-lib
+
+$(PORT_ARCHIVE): $(shell find $(UPSTREAM)/src $(UPSTREAM)/include -name '*.c' -o -name '*.h' 2>/dev/null)
+	$(call ns_build,$(UPSTREAM),css)
+
+clean: clean-lib
+	@$(call ns_clean,$(UPSTREAM))
+	rm -rf $(UPSTREAM)/$(NS_BUILDDIR)

net-libs/libdom/Makefile

@@ -0,0 +1,50 @@
+# libdom for MontaukOS
+#
+# Upstream: https://github.com/netsurf-browser/libdom (official mirror of
+# git.netsurf-browser.org; submodule at upstream/, pinned, no local patches).
+#
+# DOM implementation. Depends on libparserutils, libwapcaplet and libhubbub,
+# all of which must already be in the ports sysroot -- libdom finds the latter
+# two through pkg-config (REQUIRED_PKGS), which is why the library ports
+# publish .pc files into sysroot/lib/pkgconfig.
+#
+# Bindings come from libdom's checked-in Makefile.config, which defaults the
+# expat AND hubbub bindings to yes. Those are plain `:=` assignments, so
+# passing the variables on the sub-make command line overrides them without
+# patching upstream or dropping a Makefile.config.override into the submodule:
+#   WITH_HUBBUB_BINDING=yes   how NetSurf feeds HTML into the DOM; required.
+#   WITH_EXPAT_BINDING=no     defaults to YES upstream and must be turned off
+#                             -- it needs <expat.h>, which MontaukOS has no
+#                             port of, and NetSurf does not use it.
+#   WITH_LIBXML_BINDING       already no in their config; set explicitly so a
+#                             future upstream flip cannot surprise us.
+#
+# Needs perl on the BUILD host.
+
+UPSTREAM := upstream
+
+PORT_NAME        := libdom
+PORT_LIB         := 1
+PORT_DEPENDS     := libparserutils libwapcaplet libhubbub
+PORT_ARCHIVE     := libdom.a
+PORT_PC_IN       := $(UPSTREAM)/libdom.pc.in
+PORT_PC_VERSION  := 0.4.2
+PORT_HEADER_ROOT := $(UPSTREAM)/include
+PORT_HEADERS     := $(shell find $(UPSTREAM)/include -name '*.h' 2>/dev/null)
+
+include ../../mk/port.mk
+
+NS_MAKE_ARGS_EXTRA := \
+    WITH_HUBBUB_BINDING=yes \
+    WITH_EXPAT_BINDING=no \
+    WITH_LIBXML_BINDING=no
+
+.PHONY: all clean
+all: install-lib
+
+$(PORT_ARCHIVE): $(shell find $(UPSTREAM)/src $(UPSTREAM)/bindings $(UPSTREAM)/include -name '*.c' -o -name '*.h' 2>/dev/null)
+	$(call ns_build,$(UPSTREAM),dom)
+
+clean: clean-lib
+	@$(call ns_clean,$(UPSTREAM))
+	rm -rf $(UPSTREAM)/$(NS_BUILDDIR)

net-libs/libhubbub/Makefile

@@ -0,0 +1,34 @@
+# libhubbub for MontaukOS
+#
+# Upstream: https://github.com/netsurf-browser/libhubbub (official mirror of
+# git.netsurf-browser.org; submodule at upstream/, pinned, no local patches).
+#
+# HTML5 tokeniser and tree construction. Depends on libparserutils and
+# libwapcaplet, both of which must already be in the ports sysroot.
+#
+# Needs perl on the BUILD host: src/tokeniser/entities.inc is generated from
+# build/Entities by build/make-entities.pl. That is a host tool, not a target
+# dependency.
+
+UPSTREAM := upstream
+
+PORT_NAME        := libhubbub
+PORT_LIB         := 1
+PORT_DEPENDS     := libparserutils libwapcaplet
+PORT_ARCHIVE     := libhubbub.a
+PORT_PC_IN       := $(UPSTREAM)/libhubbub.pc.in
+PORT_PC_VERSION  := 0.3.8
+PORT_HEADER_ROOT := $(UPSTREAM)/include
+PORT_HEADERS     := $(shell find $(UPSTREAM)/include -name '*.h' 2>/dev/null)
+
+include ../../mk/port.mk
+
+.PHONY: all clean
+all: install-lib
+
+$(PORT_ARCHIVE): $(shell find $(UPSTREAM)/src $(UPSTREAM)/include -name '*.c' -o -name '*.h' 2>/dev/null)
+	$(call ns_build,$(UPSTREAM),hubbub)
+
+clean: clean-lib
+	@$(call ns_clean,$(UPSTREAM))
+	rm -rf $(UPSTREAM)/$(NS_BUILDDIR)

test/Makefile

@@ -18,7 +18,11 @@ PORT_NAME := porttest
 
 include ../mk/port.mk
 
+# Link order matters: dependents before their dependencies.
 SYSROOT_LIBS := \
+    $(PORT_SYSROOT_LIB)/libdom.a \
+    $(PORT_SYSROOT_LIB)/libcss.a \
+    $(PORT_SYSROOT_LIB)/libhubbub.a \
     $(PORT_SYSROOT_LIB)/libparserutils.a \
     $(PORT_SYSROOT_LIB)/libwapcaplet.a \
     $(PORT_SYSROOT_LIB)/libz.a

test/porttest.c

@@ -2,9 +2,23 @@
 #include <parserutils/parserutils.h>
 #include <parserutils/charset/utf8.h>
 #include <parserutils/utils/buffer.h>
+#include <hubbub/hubbub.h>
+#include <hubbub/parser.h>
+#include <libcss/libcss.h>
+#include <dom/dom.h>
 #include <zlib.h>
 #include <stdio.h>
 #include <string.h>
+/* libcss requires a URL resolver; relative resolution is not what is under
+   test here, so just duplicate the reference. */
+static css_error pt_resolve_url(void *pw, const char *base,
+                                lwc_string *rel, lwc_string **abs)
+{
+    (void) pw; (void) base;
+    *abs = lwc_string_ref(rel);
+    return CSS_OK;
+}
+
 int main(void) {
     lwc_string *a = NULL, *b = NULL;
     bool same = false;
@@ -55,6 +69,47 @@ int main(void) {
     printf("zlib %s: gzip roundtrip %lu->%lu %s\n",
            zlibVersion(), (unsigned long)strlen(s), (unsigned long)clen,
            gzok ? "PASS" : "FAIL");
+    /* libhubbub / libcss / libdom: construct the real objects rather than
+       just resolving symbols -- a stub-out or a broken static init would
+       still link but fail here. */
+    {
+        hubbub_parser *hp = NULL;
+        css_stylesheet *sheet = NULL;
+        css_stylesheet_params sp;
+        dom_document *doc = NULL;
+        dom_exception de;
+        hubbub_error he;
+
+        he = hubbub_parser_create("UTF-8", false, &hp);
+        printf("libhubbub: parser_create %s\n",
+               (he == HUBBUB_OK && hp != NULL) ? "PASS" : "FAIL");
+        if (hp != NULL) hubbub_parser_destroy(hp);
+
+        memset(&sp, 0, sizeof(sp));
+        sp.params_version = CSS_STYLESHEET_PARAMS_VERSION_1;
+        sp.level = CSS_LEVEL_21;
+        sp.charset = "UTF-8";
+        sp.url = "http://montaukos.local/";
+        sp.resolve = pt_resolve_url;
+        if (css_stylesheet_create(&sp, &sheet) == CSS_OK && sheet != NULL) {
+            const char data[] = "body { color: #123456; margin: 4px; }";
+            css_error ce = css_stylesheet_append_data(sheet,
+                    (const uint8_t *) data, sizeof(data) - 1);
+            if (ce == CSS_NEEDDATA) ce = CSS_OK;
+            if (ce == CSS_OK) ce = css_stylesheet_data_done(sheet);
+            printf("libcss: parse stylesheet %s\n", ce == CSS_OK ? "PASS" : "FAIL");
+            css_stylesheet_destroy(sheet);
+        } else {
+            printf("libcss: stylesheet_create FAIL\n");
+        }
+
+        de = dom_implementation_create_document(DOM_IMPLEMENTATION_HTML,
+                NULL, NULL, NULL, NULL, NULL, &doc);
+        printf("libdom: create_document %s\n",
+               (de == DOM_NO_ERR && doc != NULL) ? "PASS" : "FAIL");
+        if (doc != NULL) dom_node_unref(doc);
+    }
+
     lwc_string_unref(a); lwc_string_unref(b);
     parserutils_buffer_destroy(buf);
     return 0;