Merge pull request #2866 from sparklemotion/flavorjones-upgrade-libxm…

…l2-2.11.0

dep: update libxml2 to v2.11.3, libxslt to v1.1.38

CHANGELOG.md

@@ -24,6 +24,16 @@ Benchmarks show that this setting will significantly improve performance, but be
 You can read more about this in the decision record at `adr/2023-04-libxml-memory-management.md`.
 
 
+### Dependencies
+
+[CRuby] Vendored libxml2 is updated to v2.11.3 from v2.10.4. Please see the multiple release notes:
+
+- https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.11.0
+- https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.11.1
+- https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.11.2
+- https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.11.3
+
+
 ### Added
 
 * `Encoding` objects may now be passed to serialization methods like `#to_xml`, `#to_html`, `#serialize`, and `#write_to` to specify the output encoding. Previously only encoding names (strings) were accepted. [[#2774](https://github.com/sparklemotion/nokogiri/issues/2774), [#2798](https://github.com/sparklemotion/nokogiri/issues/2798)] (Thanks, [@ellaklara](https://github.com/ellaklara)!)

dependencies.yml

@@ -1,12 +1,12 @@
 libxml2:
-  version: "2.10.4"
-  sha256: "ed0c91c5845008f1936739e4eee2035531c1c94742c6541f44ee66d885948d45"
-  # sha-256 hash provided in https://download.gnome.org/sources/libxml2/2.10/libxml2-2.10.4.sha256sum
+  version: "2.11.3"
+  sha256: "f1acae1664bda006cd81bfc238238217043d586d06659d5c0e3d1bcebe040870"
+  # sha-256 hash provided in https://download.gnome.org/sources/libxml2/2.11/libxml2-2.11.3.sha256sum
 
 libxslt:
-  version: "1.1.37"
-  sha256: "3a4b27dc8027ccd6146725950336f1ec520928f320f144eb5fa7990ae6123ab4"
-  # sha-256 hash provided in https://download.gnome.org/sources/libxslt/1.1/libxslt-1.1.37.sha256sum
+  version: "1.1.38"
+  sha256: "1f32450425819a09acaff2ab7a5a7f8a2ec7956e505d7beeb45e843d0e1ecab1"
+  # sha-256 hash provided in https://download.gnome.org/sources/libxslt/1.1/libxslt-1.1.38.sha256sum
 
 zlib:
   version: "1.2.13"

ext/nokogiri/extconf.rb

@@ -229,7 +229,7 @@ def gnome_source
   if ENV["NOKOGIRI_USE_CANONICAL_GNOME_SOURCE"]
     "https://download.gnome.org"
   else
-    "https://mirror.csclub.uwaterloo.ca/gnome" # old reliable
+    "https://muug.ca/mirror/gnome" # old reliable
   end
 end
 
@@ -882,8 +882,13 @@ def configure
       recipe.patch_files = Dir[File.join(PACKAGE_ROOT_DIR, "patches", "libxml2", "*.patch")].sort
     end
 
+    cppflags = concat_flags(ENV["CPPFLAGS"])
     cflags = concat_flags(ENV["CFLAGS"], "-O2", "-U_FORTIFY_SOURCE", "-g")
 
+    if cross_build_p
+      cppflags = concat_flags(cppflags, "-DNOKOGIRI_PRECOMPILED_LIBRARIES")
+    end
+
     if zlib_recipe
       recipe.configure_options << "--with-zlib=#{zlib_recipe.path}"
     end
@@ -914,6 +919,7 @@ def configure
       "--with-c14n",
       "--with-debug",
       "--with-threads",
+      "CPPFLAGS=#{cppflags}",
       "CFLAGS=#{cflags}",
     ]
   end

ext/nokogiri/html4_sax_parser_context.c

@@ -41,6 +41,12 @@ parse_file(VALUE klass, VALUE filename, VALUE encoding)
                              StringValueCStr(filename),
                              StringValueCStr(encoding)
                            );
+
+  if (ctxt->sax) {
+    xmlFree(ctxt->sax);
+    ctxt->sax = NULL;
+  }
+
   return noko_xml_sax_parser_context_wrap(klass, ctxt);
 }
 
@@ -78,11 +84,6 @@ parse_with(VALUE self, VALUE sax_handler)
   ctxt = noko_xml_sax_parser_context_unwrap(self);
   sax = noko_sax_handler_unwrap(sax_handler);
 
-  /* Free the sax handler since we'll assign our own */
-  if (ctxt->sax && ctxt->sax != (xmlSAXHandlerPtr)&xmlDefaultSAXHandler) {
-    xmlFree(ctxt->sax);
-  }
-
   ctxt->sax = sax;
   ctxt->userData = (void *)NOKOGIRI_SAX_TUPLE_NEW(ctxt, sax_handler);
 

ext/nokogiri/xml_document.c

@@ -74,14 +74,23 @@ dealloc(void *data)
 
   ruby_xfree(doc->_private);
 
-  /* When both Nokogiri and libxml-ruby are loaded, make sure that all nodes
-   * have their _private pointers cleared. This is to avoid libxml-ruby's
-   * xmlDeregisterNode callback from accessing VALUE pointers from ruby's GC
-   * free context, which can result in segfaults.
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wdeprecated-declarations" // xmlDeregisterNodeDefault is deprecated as of libxml2 2.11.0
+  /*
+   * libxml-ruby < 3.0.0 uses xmlDeregisterNodeDefault. If the user is using one of those older
+   * versions, the registered callback from libxml-ruby will access the _private pointers set by
+   * nokogiri, which will result in segfaults.
+   *
+   * To avoid this, we need to clear the _private pointers from all nodes in this document tree
+   * before that callback gets invoked.
+   *
+   * libxml-ruby 3.0.0 was released in 2017-02, so at some point we can probably safely remove this
+   * safeguard (though probably pairing with a runtime check on the libxml-ruby version).
    */
   if (xmlDeregisterNodeDefaultValue) {
     remove_private((xmlNodePtr)doc);
   }
+#pragma GCC diagnostic pop
 
   xmlFreeDoc(doc);
 }

ext/nokogiri/xml_sax_parser_context.c

@@ -77,6 +77,12 @@ static VALUE
 parse_file(VALUE klass, VALUE filename)
 {
   xmlParserCtxtPtr ctxt = xmlCreateFileParserCtxt(StringValueCStr(filename));
+
+  if (ctxt->sax) {
+    xmlFree(ctxt->sax);
+    ctxt->sax = NULL;
+  }
+
   return noko_xml_sax_parser_context_wrap(klass, ctxt);
 }
 
@@ -147,11 +153,6 @@ parse_with(VALUE self, VALUE sax_handler)
   ctxt = noko_xml_sax_parser_context_unwrap(self);
   sax = noko_sax_handler_unwrap(sax_handler);
 
-  /* Free the sax handler since we'll assign our own */
-  if (ctxt->sax && ctxt->sax != (xmlSAXHandlerPtr)&xmlDefaultSAXHandler) {
-    xmlFree(ctxt->sax);
-  }
-
   ctxt->sax = sax;
   ctxt->userData = (void *)NOKOGIRI_SAX_TUPLE_NEW(ctxt, sax_handler);
 

patches/libxml2/0011-rip-out-libxml2-s-libc_single_threaded-support.patch

@@ -0,0 +1,30 @@
+From fb618bd27ac7a9fd32973320016c7ff28dd2f60e Mon Sep 17 00:00:00 2001
+From: Mike Dalessio <mike.dalessio@gmail.com>
+Date: Fri, 5 May 2023 10:10:43 -0400
+Subject: [PATCH] rip out libxml2's libc_single_threaded support
+
+This is a preventative measure because this feature relies on a glibc
+version we can't realistically require of users today. We're not yet
+precompiling on a system with a modern-enough glibc to make this an
+actual problem, but I'm doing it while all of this context is fresh.
+---
+ threads.c | 3 ++-
+ 1 file changed, 2 insertions(+), 1 deletion(-)
+
+diff --git a/threads.c b/threads.c
+index 60dbce4c..eb1c12e5 100644
+--- a/threads.c
++++ b/threads.c
+@@ -27,7 +27,8 @@
+
+ #if defined(HAVE_POSIX_THREADS) && \
+     defined(__GLIBC__) && \
+-    __GLIBC__ * 100 + __GLIBC_MINOR__ >= 234
++    __GLIBC__ * 100 + __GLIBC_MINOR__ >= 234 && \
++    !defined(NOKOGIRI_PRECOMPILED_LIBRARIES)
+
+ /*
+  * The modern way available since glibc 2.32.
+-- 
+2.40.1
+