Conversation
SQLite3::close() called from within a userland function, aggregate, collation or authorizer callback freed the registered statements and functions while sqlite3 was still executing, corrupting the active statement and crashing the request. Track callback re-entry with a per-database counter shared by all four callback kinds and throw an Error from close() while it is non-zero; the database stays usable and can be closed after the query completes. Closes phpGH-23650
* PHP-8.4: ext/sqlite3: reject close() from inside a callback
The bridge registered an id as XML_ATTRIBUTE_ID only when lexbor reported the HTML namespace, so ids on SVG and MathML elements never reached getElementById(). Separately, lexbor initializes every attribute node with its element's namespace and overrides that only through the foreign-attribute adjust table, which the parser wires up for SVG and MathML alone, so a fragment parsed with an xlink, xml or xmlns context element came back as <z xlink:id="q" xlink:foo="1">. Restrict the three namespace branches to attributes lexbor adjusted, and key the id registration off the namespace the bridge assigned. Closes phpGH-23598
* PHP-8.4: dom: fix attribute namespaces from foreign content in the HTML parser
php_sock_array_to_fd_set() returned a flag rather than the socket count, so PHP_SAFE_MAX_FD() got a zero and never fired. Sets larger than FD_SETSIZE were quietly truncated and select() reported on the survivors. Close phpGH-23632
* PHP-8.4: ext/sockets: socket_select() silent set truncation on Windows.
…ansformation. Fix php#23730 Importing a stylesheet from a php:function callback freed the stylesheet libxslt was still applying, and the transformation methods kept using the stale pointer to save the result. The transformation depth is now tracked on the object and importStylesheet() throws while it is non-zero. Close phpGH-23737
* PHP-8.4: ext/xsl: XSLTProcessor::importStylesheet() use-after-free during a transformation.
… callback. Fix php#23747 ZipArchive::close() called from a progress or cancel callback ran a nested zip_close() that failed, then zip_discard() freed the archive while the outer zip_close() from close() or open() was still using it. Track the close in progress and throw an Error from close() and open() meanwhile. Close phpGH-23749
* PHP-8.4: ext/zip: ZipArchive::close() use-after-free from a progress or cancel callback.
…luation Reconstructing the object from a php:function callback freed the context libxml2 was still evaluating, and php_xpath_eval() wrote back into it once the evaluation returned. The evaluation depth is now tracked on the object and __construct() throws while it is non-zero. Close phpGH-23735
* PHP-8.4: Fix phpGH-23729: DOMXPath::__construct() use-after-free during an evaluation
And thus should never be allowed to be qualified, as this lead to other bugs such as it not considered to be a built-in type by Reflection if it had a namespace component. We *may* want to extend the unqualified restriction to self and parent in the future. Closes phpGH-23768
… during request activation) Closes phpGH-22568
The prompt parser ran as a do-while, so an empty cli.prompt executed the body on the terminator and scanned past it. Use a while loop and smart_str_extract(), which returns the interned empty string when nothing was appended. No regression test: extra unicode warnings from the over-read depend on heap contents, so a .phpt cannot pin the bug red-before. Closes phpGH-23415
* PHP-8.4: Fix heap over-read in cli_get_prompt() for empty cli.prompt
init_process_info() memset the pointer parameter instead of the PROCESS_INFORMATION structure it points at, leaving the struct uninitialized before CreateProcessW(). Zero it through the pointer. find_comspec_nt() dereferences *comspec in its cleanup while the caller only assigns it on success, so a failed SearchPathW() read an indeterminate value. Initialize the caller's variable to NULL. set_proc_descriptor_to_blackhole() tested CreateFileA() against NULL, but CreateFileA() signals failure with INVALID_HANDLE_VALUE, so a failed open went undetected and an invalid handle was inherited by the child. Test against INVALID_HANDLE_VALUE. Closes phpGH-23412
* PHP-8.4: Fix three Win32-only defects in proc_open descriptor handling
…tructions with callable conversion Call level counter was broken: ZEND_CALLABLE_CONVERT was forgotten. Closes phpGH-23777.
* PHP-8.4: Fix OSS-Fuzz #546798343: Heap-buffer-overflow in zend_delete_call_instructions with callable conversion
…ndow inflate_init() applies the preset dictionary eagerly for raw streams via inflateSetDictionary(), gated on encoding == PHP_ZLIB_ENCODING_RAW. But encoding is first adjusted by the window size (encoding += 15 - window), so a raw stream with a non-default window no longer equals PHP_ZLIB_ENCODING_RAW and the dictionary is silently dropped; raw streams carry no header and never emit Z_NEED_DICT, so inflate_add()'s deferred path never applies it either. Gate on the pre-adjustment encoding. The deflate side already applies the dictionary unconditionally, so the roundtrip was broken for this case. Closes phpGH-22381
The x86 matcher folds v = BINOP(a, b); c = CMP(v, 0); GUARD(c) into IR_GUARD_JCC_INT, emitting the BINOP, dropping the CMP and branching on the flags the BINOP left, but it only required the BINOP to precede the CMP in the IR. Once GCM hoists a loop-invariant BINOP into a dominating block, the jcc reads flags the intervening code has clobbered and the guard fires on whatever is in EFLAGS. Require the BINOP to sit in the guard's block and allow only snapshots between the comparison and the guard, the way ir_match_fuse_load() pairs ir_in_same_block() with ir_match_has_mem_deps(). The sibling MEM_BINOP fold already checks the block, the IF side folds are pinned by full ref adjacency, and ir_aarch64.dasc has no guard fold. Mirrors the upstream fix dstogov/ir@51107a3. Fixes phpGH-23693 Closes phpGH-23711
* PHP-8.4: Fix phpGH-23693: JIT guard branches on stale flags across basic blocks
An attribute that fails to set from the PDO constructor options array named argument #3 ($password) instead of #4 ($options), because php_pdo_internal_construct_driver() passed a hardcoded value_arg_num of 3 to pdo_dbh_attribute_set(). PDO::setAttribute(), the only other caller, already passes the correct number.
iliaal
force-pushed
the
fix/aph-hw9-85
branch
from
September 20, 2026 16:11
a9b7883 to
1b4d9f1
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
When an attribute passed through the PDO constructor options array failed to set, the resulting error message named argument #3 ($password) instead of argument #4 ($options), because pdo_dbh_attribute_set() was called from php_pdo_internal_construct_driver() with a hardcoded value_arg_num of 3. This passes 4 so errors like an invalid PDO::ATTR_ERRMODE now correctly report Argument #4 ($options). The only other caller, PDO::setAttribute(), already passes the correct number and was audited; a test asserting the corrected message is included and fails without the patch.