IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
When decoding input data, check whether the "raw" buffer is empty after
parsing the document. Otherwise, the input ends with a truncated
multi-byte sequence which shouldn't be silently ignored.
Covered by: test/VC/ElementValid5
This only affects XML Reader API with LIBXML_REGEXP_ENABLED and
LIBXML_VALID_ENABLED turned on.
* result/VC/ElementValid5.rdr:
- Update result to add missing error message.
* python/tests/reader2.py:
* result/VC/ElementValid6.rdr:
* result/VC/ElementValid7.rdr:
* result/valid/781333.xml.err.rdr:
- Update result to fix grammar issue.
* valid.c:
(xmlValidatePopElement):
- Check return value of xmlRegExecPushString() to handle -1, and
assign 'ret = 0;' to return 0 from xmlValidatePopElement().
This change affects xmlTextReaderValidatePop() from
xmlreader.c.
- Fix grammar of error message by changing 'child' to
'children'.
In commit 21ca8829, we started to ignore namespaces in HTML element
names but we still called xmlSplitQName, effectively stripping the
namespace prefix. This would cause elements like <o:p> being parsed
as <p>. Now we leave the name untouched.
Fixes#508.
Also make text inclusions work with memory buffers, for example when
using a custom entity loader, and fix a memory leak in case of invalid
characters.
Fixes#483.
Avoids resource exhaustion if the maximum recursion depth was exceeded.
Note that the XInclude engine offers no protection against other
"billion laughs"-style amplification attacks as long as they stay below
the maximum depth.
If we try to continue parsing after an error in the internal or external
subset, entity expansion accounting gets more complicated. Simply halt
the parser.
Found with libFuzzer.
Don't set the "checked" flag when checking entities in default attribute
values. These entities could reference other entities which weren't
defined yet, so the check isn't reliable.
This fixes a short-lived regression which could lead to a call stack
overflow later in xmlStringGetNodeList.
Remove inaccurate xmlParseCheckTransition check.
Remove non-incremental xmlParseGetLasts check.
Add functions that check for several boundary constructs more
accurately, keeping track of progress in ctxt->checkIndex.
Fixes#439.
This is another attempt at fixing parser progress checks. Instead of
relying on in->consumed, which could overflow, change some DTD parser
functions to make guaranteed progress on certain byte sequences.
When using xmlreader, XPointer expressions in XIncludes simply cannot
work. Expressions can reference nodes which weren't parsed yet or which
were already deleted.
After fixing nested XIncludes, we reference includes which were parsed
previously. When streaming, these nodes could have been deleted, leading
to use-after-free errors.
Disallow XPointer expressions and truncate the include table in
streaming mode.
Don't create subcontext in xmlXIncludeRecurseDoc. Save and restore 'doc'
and 'incTab' instead.
Make xmlXIncludeLoadFallback call xmlXIncludeCopyNode which seems safer
than xmlXIncludeDoProcess since the latter may modify the document.
This should also be more performant since we need to copy the whole
fallback subtree anyway. Also make sure to avoid replacements in
fallback elements in xmlXIncludeDoProcess.
The reader interface with XIncludes is somewhat broken and can generate
different error messages. Start to move tests which are sketchy with
reader to a separate directory.
This build config resulted in segfaults in 'runtest' because a special
xmlElementContentPtr showed up in a few places. I'm not sure if this is
the right fix.
An error message was changed to conform to the --with-regexps build.
There are still a few missing validity errors, so the tests don't pass.
Commit 4fd69f3e fixed handling of '<' characters not followed by an
ASCII letter. But a '<!' sequence followed by invalid characters should
be treated as bogus comment and skipped.
Fixes#380.
Commit 7618a3b1 didn't account for coalesced text nodes.
I think it would be better if xmlStaticCopyNode didn't try to coalesce
text nodes at all. This code path can only be triggered if some other
code doesn't coalesce text nodes properly. In this case, OSS-Fuzz found
such behavior in xinclude.c.
Remove all the parts of the old test suite which are covered by
runtest.c for quite some time.
The following test programs are removed:
- testC14N
- testHTML
- testReader
- testRelax
- testSAX
- testSchemas
- testURI
- testXPath
This also removes a few results of unimportant tests only run by the old
test suite.
Downgrade the error message to a warning since the error was ignored,
anyway. Also print the name of redeclared entity. For a proper fix that
also shows filename and line number of the invalid redeclaration, we'd
have to
- pass the parser context to the entity functions somehow, or
- make these functions return distinct error codes.
Partial fix for #308.
If a document is parsed with XML_PARSE_DTDVALID and without
XML_PARSE_NOENT, the value of ID attributes has to be normalized after
potentially expanding entities in xmlRemoveID. Otherwise, later calls
to xmlGetID can return a pointer to previously freed memory.
ID attributes which are empty or contain only whitespace after
entity expansion are affected in a similar way. This is fixed by
not storing such attributes in the ID table.
The test to detect streaming mode when validating against a DTD was
broken. In connection with the defects above, this could result in a
use-after-free when using the xmlReader interface with validation.
Fix detection of streaming mode to avoid similar issues. (This changes
the expected result of a test case. But as far as I can tell, using the
XML reader with XIncludes referencing the root document never worked
properly, anyway.)
All of these issues can result in denial of service. Using xmlReader
with validation could result in disclosure of memory via the error
channel, typically stderr. The security impact of xmlGetID returning
a pointer to freed memory depends on the application. The typical use
case of calling xmlGetID on an unmodified document is not affected.
This makes the logic in UTF16BEToUTF8() match UTF16LEToUTF8().
* encoding.c:
(UTF16LEToUTF8):
- Fix comment to describe what the code does.
(UTF16BEToUTF8):
- Fix undefined behavior which was applied to UTF16LEToUTF8() in
2f9382033e.
- Add bounds check to while() loop which was applied to
UTF16LEToUTF8() in be803967db.
- Do not return -2 when (in >= inend) to fix the bug. This was
applied to UTF16LEToUTF8() in 496a1cf592.
- Inline (<< 8) statements to match UTF16LEToUTF8().
Add the following tests and results:
test/text-4-byte-UTF-16-BE-offset.xml
test/text-4-byte-UTF-16-BE.xml
test/text-4-byte-UTF-16-LE-offset.xml
test/text-4-byte-UTF-16-LE.xml
Readd the XML_ERR_TAG_NOT_FINISHED error on unexpected EOF which was
removed in commit 62150ed2.
This commit also introduced a regression for direct users of
xmlParseContent. Unclosed tags weren't checked.
Commit 62150ed2 introduced a small regression in the error messages for
mismatched tags. This typically only affected messages after the first
mismatch, but with custom SAX handlers all line numbers would be off.
This also fixes line numbers in the SAX push parser which were never
handled correctly.
Implement section "4.6 Predefined Entities" of the XML 1.0 spec and
check whether redeclarations of predefined entities match the original
definitions.
Note that some test cases declared
<!ENTITY lt "<">
But the XML spec clearly states that this is illegal:
> If the entities lt or amp are declared, they MUST be declared as
> internal entities whose replacement text is a character reference to
> the respective character (less-than sign or ampersand) being escaped;
> the double escaping is REQUIRED for these entities so that references
> to them produce a well-formed result.
Also fixes#217 but the connection is only tangential. The integer
overflow discovered by fuzzing was more related to the fact that various
parts of the parser disagreed on whether to prefer predefined entities
over their redeclarations. The whole situation is a mess and even
depends on legacy parser options. But now that redeclarations are
validated, it shouldn't make a difference.
As noted in the added comment, this is also one of the cases where
overly defensive checks can hide interesting logic bugs from fuzzers.
The code wasn't dead after all, but I can see no reason in delaying
the XPointer evaluation. This could lead to nodes included earlier
appearing in XPointer results.
When creating XML_XINCLUDE_START nodes, the children of the original
xi:include node must be freed, otherwise fallback content is copied
twice, doubling runtime and memory consumption for each nested
xi:fallback/xi:include pair.
Found with libFuzzer.
Fix a few remaining cases where the HTML push parser would scan more
content during lookahead than being parsed later.
Make sure that htmlParseDocTypeDecl consumes all content up to the
final '>' in case of errors. The old comment said "We shouldn't try to
resynchronize", but ignoring invalid content is also what the HTML5
spec mandates.
Likewise, make htmlParseEndTag skip to the final '>' in invalid end
tags even if not in recovery mode. This is probably the most visible
change in practice and leads to different output for some tests but is
also more in line with HTML5.
Make sure that htmlParsePI and htmlParseComment don't abort if invalid
characters are encountered but log an error and ignore the character.
Change some other end-of-buffer checks to test for a zero byte instead
of relying on IS_CHAR.
Fix usage of IS_CHAR macro in htmlParseScript.
- Bug 757711: heap-buffer-overflow in xmlFAParsePosCharGroup
<https://bugzilla.gnome.org/show_bug.cgi?id=757711>
- Bug 783015 - Integer-overflow in xmlFAParseQuantExact
<https://bugzilla.gnome.org/show_bug.cgi?id=783015>
(Regexptests): Add support for checking stderr output when
running regexp tests. This makes it possible to check in test
cases that fail and not see false-positive error output when
running the tests. Unlike other libxml2 test suites, if there
is no stderr output, no *.err file needs to be created.
Commit eeb99329 removed an important optimization avoiding quadratic
runtime when repeatedly scanning the input buffer for terminating
characters in the HTML push parser. The related bug is
https://bugzilla.gnome.org/show_bug.cgi?id=444994
Make sure that ctxt->checkIndex is always written and store additional
parser state in ctxt->inSubset which is unused in the HTML parser.
Found by OSS-Fuzz.
Despite the comment, I can't see a reason why external entities must be
loaded in the SAX handler. For external entities, the handler is
typically first invoked via xmlParseReference which will later load the
entity on its own if it wasn't loaded yet.
The old code also lead to duplicated SAX events which makes it
basically impossible to reuse xmlSAX2GetEntity for a custom SAX parser.
See the change to the expected test output.
Note that xmlSAX2GetEntity was loading the entity via
xmlParseCtxtExternalEntity while xmlParseReference uses
xmlParseExternalEntityPrivate. In the previous commit, the two
functions were merged, trying to compensate for some slight differences
between the two mostly identical implementations.
But the more urgent reason for this change is that xmlParseReference
has the facility to abort early when recursive entities are detected,
avoiding what could practically amount to an infinite loop.
If you want to backport this change, note that the previous three
commits are required as well:
f9ea1a24 Fix copying of entities in xmlParseReference
5c7e0a9a Copy some XMLReader option flags to parser context
1a3e584a Merge code paths loading external entities
Found by OSS-Fuzz.
Before, reader mode would end up in a branch that didn't handle
entities with multiple children and failed to update ent->last, so the
hack copying the "extra" reader data wouldn't trigger. Consequently,
some empty nodes in entities are correctly detected now in the test
suite. (The detection of empty nodes in entities is still buggy,
though.)
libxml2 correctly rejects any4_0.xsd as invalid schema. I can't figure
out what the intent behind this test case was. Simply adjust the
expected output to match the current behavior.
Closes#92.
Previously, test/relaxng/ambig_name-class2.xml would fail to validate
against test/relaxng/ambig_name-class2.rng:
> test/relaxng/ambig_name-class2.rng:4:
> element attribute: Relax-NG parser error :
> Found anyName attribute without oneOrMore ancestor
> Relax-NG schema test/relaxng/ambig_name-class2.rng failed to compile
Signed-off-by: Jan Pokorný <jpokorny@redhat.com>
Previously, test/relaxng/ambig_name-class.xml would fail to validate
for a simple reason -- interleave within "open-name-class" context
is supposed to be fine with whatever else is pending the consumption,
since effectively, it's unrelated from a higher parsing perspective.
Signed-off-by: Jan Pokorný <jpokorny@redhat.com>
Consolidate code paths evaluating XPath predicates and filters.
Don't push context node on stack when evaluating predicates. I have no
idea why this was done. It seems completely useless and trying to pop
the context node from a corrupted stack has already caused security
issues.
Filter nodesets in-place and don't create node sets with NULL gaps which
allows to simplify merging a great deal. Simply move matched nodes
backward and create a compact node set.
Merge xmlXPathCompOpEvalPositionalPredicate into
xmlXPathCompOpEvalPredicate.
Rewrite conversion of double to int in xmlXPathSubstringFunction, adding
range checks to avoid undefined behavior. Make sure to add start and
length as floating-point numbers before converting to int. Fix a bug
when rounding negative start indices.
Remove unneeded calls to xmlXPathIs{Inf,NaN} and rely on IEEE math
instead. Avoid computing the string length. xmlUTF8Strsub works as
expected if the length of the requested substring exceeds the input.
Found with libFuzzer and UBSan.
When including a grammar from another grammar, we need to make sure that any
redefines of starts and includes that that grammar does inside any of its
include elements are also removed.
The pattern nameClass allows for nested choice elements, for example
<name>
<choice>
<choice>
<name>a</name>
<name>b</name>
</choice>
<name>c</name>
</choice>
</name>
which is semantically equivalent to
<name>
<choice>
<name>a</name>
<name>b</name>
<name>c</name>
</choice>
</name>
The old code didn’t handle this correctly, as it never expected a choice inside
another choice. This patch fixes this by flattening any nested choices.
This pattern of nested choice elements comes up in RELAX NG simplification,
where all choice elements are rewritten in this nested manner, see section 4.12
of the RELAX NG specification.
RELAX NG allows for div elements inside of include elements. We need to look
inside those div elements for start and define elements that may be redefining
start and define elements in the included grammar.
Fix two bugs in xmlXPathNodeValHash which could lead to errors when
comparing nodesets to strings:
- Only use contents of text nodes to compute the hash for element nodes.
Comments, PIs, and other node types don't affect the string-value and
must be ignored.
- Reset `string` to NULL for node types other than text.
Reported by Aleksei on the mailing list:
https://mail.gnome.org/archives/xml/2017-September/msg00016.html
This reverts commit 79c8a6b which caused a serious regression in
streaming mode.
Also reverts part of commit 52ceced "Fix infinite loops with push
parser in recovery mode".
Fixes bug 786554.
When expanding a parameter entity in a DTD, infinite recursion could
lead to an infinite loop or memory exhaustion.
Thanks to Wei Lei for the first of many reports.
Fixes bug 759579.
Now that replacement of parameter entities goes exclusively through
xmlSkipBlankChars, we can account for the surrounding space characters
there and remove the "blanks wrapper" hack.
Pop all extra input streams before resetting the input. Otherwise,
a call to xmlPopInput could make input available again.
Also set input->end to input->cur.
Changes the test output for some error tests. Unfortunately, some
fuzzed test cases were added to the test suite without manual cleanup.
This makes it almost impossible to review the impact of later changes
on the test output.
Make sure to finish all entities in the internal subset. Nevertheless,
readd a sanity check in xmlParseStartTag2 that was lost in my previous
commit. Also add a sanity check in xmlPopInput. Popping an input
unexpectedly was the source of many recent memory bugs. The check
doesn't mitigate such issues but helps with diagnosis.
Always base entity boundary checks on the input ID, not the input
pointer. The pointer could have been reallocated to the old address.
Always throw a well-formedness error if a boundary check fails. In a
few places, a validity error was thrown.
Fix a few error codes and improve indentation.
