On 2025-07-24 05:12 +0200, Michael Paquier wrote:
> Switching back to the previous code, where we rely on
> xmlParseBalancedChunkMemory() fixes the issue. A quick POC is
> attached. It fails one case in check-world with SERIALIZE because I
> am not sure it is possible to pass down some options through
> xmlParseBalancedChunkMemory(), still the regression is gone, and I am
> wondering if there is not a better solution to be able to dodge the
> original problem and still accept this case.
The whitespace can be preserved by setting xmlKeepBlanksDefault before
parsing. See attached v2. That function is deprecated, though. But
libxml2 uses thread-local globals, so it should be safe. Other than
that, I see no other way to set XML_PARSE_NOBLANKS with
xmlParseBalancedChunkMemory.
[1]
https://gitlab.gnome.org/GNOME/libxml2/-/blob/408bd0e18e6ddba5d18e51d52da0f7b3ca1b4421/parserInternals.c#L2833
--
Erik Wienhold
>From ef277b72586c9df575c78df1eddf6b481fcfaad2 Mon Sep 17 00:00:00 2001
From: Michael Paquier <mich...@paquier.xyz>
Date: Thu, 24 Jul 2025 12:01:52 +0900
Subject: [PATCH v2] Fix xml2 regression
---
src/backend/utils/adt/xml.c | 32 ++++++++++----------------------
1 file changed, 10 insertions(+), 22 deletions(-)
diff --git a/src/backend/utils/adt/xml.c b/src/backend/utils/adt/xml.c
index f7b731825fc..e405c042c4f 100644
--- a/src/backend/utils/adt/xml.c
+++ b/src/backend/utils/adt/xml.c
@@ -1795,6 +1795,7 @@ xml_parse(text *data, XmlOptionType xmloption_arg,
PgXmlErrorContext *xmlerrcxt;
volatile xmlParserCtxtPtr ctxt = NULL;
volatile xmlDocPtr doc = NULL;
+ int old_keep_blanks = -1;
/*
* This step looks annoyingly redundant, but we must do it to have a
@@ -1900,9 +1901,6 @@ xml_parse(text *data, XmlOptionType xmloption_arg,
}
else
{
- xmlNodePtr root;
- xmlNodePtr oldroot PG_USED_FOR_ASSERTS_ONLY;
-
/* set up document with empty root node to be the
context node */
doc = xmlNewDoc(version);
if (doc == NULL || xmlerrcxt->err_occurred)
@@ -1916,31 +1914,18 @@ xml_parse(text *data, XmlOptionType xmloption_arg,
"could not allocate XML
document");
doc->standalone = standalone;
- root = xmlNewNode(NULL, (const xmlChar *)
"content-root");
- if (root == NULL || xmlerrcxt->err_occurred)
- xml_ereport(xmlerrcxt, ERROR,
ERRCODE_OUT_OF_MEMORY,
- "could not allocate xml
node");
-
- /*
- * This attaches root to doc, so we need not free it
separately;
- * and there can't yet be any old root to free.
- */
- oldroot = xmlDocSetRootElement(doc, root);
- Assert(oldroot == NULL);
-
/* allow empty content */
if (*(utf8string + count))
{
xmlNodePtr node_list = NULL;
- xmlParserErrors res;
- res = xmlParseInNodeContext(root,
-
(char *) utf8string + count,
-
strlen((char *) utf8string + count),
-
options,
-
&node_list);
+ old_keep_blanks =
xmlKeepBlanksDefault(preserve_whitespace ? 1 : 0);
- if (res != XML_ERR_OK ||
xmlerrcxt->err_occurred)
+ res_code = xmlParseBalancedChunkMemory(doc,
NULL, NULL, 0,
+
utf8string + count,
+
&node_list);
+ elog(NOTICE, "res_code = %d err_occurred =
%d", res_code, xmlerrcxt->err_occurred);
+ if (res_code != 0 || xmlerrcxt->err_occurred)
{
xmlFreeNodeList(node_list);
xml_errsave(escontext, xmlerrcxt,
@@ -1975,6 +1960,9 @@ fail:
if (ctxt != NULL)
xmlFreeParserCtxt(ctxt);
+ if (old_keep_blanks != -1)
+ xmlKeepBlanksDefault(old_keep_blanks);
+
pg_xml_done(xmlerrcxt, false);
return doc;
--
2.50.1
