Skip to content

Enable 64bit integer support on 32bit arch --enable-zend-int64 - #19079

Draft
marc-mabe wants to merge 8 commits into
php:masterfrom
marc-mabe:int64
Draft

Enable 64bit integer support on 32bit arch --enable-zend-int64#19079
marc-mabe wants to merge 8 commits into
php:masterfrom
marc-mabe:int64

Conversation

@marc-mabe

@marc-mabe marc-mabe commented Jul 9, 2025 •
edited
Loading

Copy link
Copy Markdown
Contributor

This PR allows to use 64bit integers on 32bit platforms if compiled with --enable-zend-int64

  • allow to enable 64bit integers on 32bit platforms (--enable-zend-int64)
  • add const for system architecture (currently PHP_SYS_SIZE = [4|8])
  • fix all errors in Zend/tests/ and tests/
  • fix all errors in ext/
  • Support 64 bit timestamps
  • Support large files

The main part is that ZEND_ENABLE_ZVAL_LONG64 1 gets defined.

@iluuu1994

iluuu1994 commented Jul 9, 2025 •
edited
Loading

Copy link
Copy Markdown
Member

I think this should go through the RFC process. There have been suggestions to drop 32-bit support, so I'm sure at least some people would object to expanding support further. I could see this as an alternative to dropping 32-bit support iff we also dropped 4-byte zend_long support, as that would align 32 and 64-bit.

Girgias
Girgias reviewed Jul 9, 2025
View reviewed changes
Comment thread ext/gmp/gmp.c Outdated
@marc-mabe

marc-mabe commented Jul 9, 2025 •
edited
Loading

Copy link
Copy Markdown
Contributor Author

I have fixed some cases which failed due to SIZEOF_SIZE_T != SIZEOF_ZEND_LONG, but there are probably more.
But there is something off with closure handling and static variables I'm not able to find :(

Currently, the test state is as follows:

$ sapi/cli/php run-tests.php Zend/tests/ tests/
<snip>
Number of tests :  5929              5791
Tests skipped   :   138 (  2.3%) --------
Tests warned    :     0 (  0.0%) (  0.0%)
Tests failed    :    44 (  0.7%) (  0.8%)
Expected fail   :     2 (  0.0%) (  0.0%)
Tests passed    :  5745 ( 96.9%) ( 99.2%)
---------------------------------------------------------------------
Time taken      : 158.811 seconds
=====================================================================

=====================================================================
FAILED TEST SUMMARY
---------------------------------------------------------------------
php://input is empty when enable_post_data_reading=Off [tests/basic/bug67198.phpt]
Static variables in functions [tests/func/002.phpt]
Bug #23384 (use of class constants in statics) [tests/lang/bug23384.phpt]
Static keyword - basic tests [tests/lang/static_basic_001.phpt]
Bug #78502: Incorrect stack size calculation for indirectly recursive function call [Zend/tests/bug78502.phpt]
Bug #54358 (Closure, use and reference) [Zend/tests/closures/bug54358.phpt]
Closure 009: Using static vars inside lambda [Zend/tests/closures/closure_009.phpt]
Closure 047: Use in preg_replace_callback() using variables by reference [Zend/tests/closures/closure_047.phpt]
Closure 048: Use in preg_replace_callback() using variables by reference [Zend/tests/closures/closure_048.phpt]
Closures in const expressions support static variables. [Zend/tests/closures/closure_const_expr/static_variable.phpt]
Closure use list can have trailing commas [Zend/tests/closures/closure_use_trailing_comma.phpt]
Bug #79836 (Segfault in concat_function) [Zend/tests/concat/bug79836.phpt]
new in constant expressions [Zend/tests/constexpr/new.phpt]
Named params in new in const expr [Zend/tests/constexpr/new_named_params.phpt]
GH-13446: Exception handler attempting to free itself [Zend/tests/exceptions/gh13446_2.phpt]
Fiber::getReturn() after bailout [Zend/tests/fibers/get-return-after-bailout.phpt]
GH-15108 003: Segfault with delegated generator in suspended fiber [Zend/tests/fibers/gh15108-003.phpt]
GH-15108 004: Segfault with delegated generator in suspended fiber [Zend/tests/fibers/gh15108-004.phpt]
GH-15108 007: Segfault with delegated generator in suspended fiber [Zend/tests/fibers/gh15108-007.phpt]
Bug GH-9916 001 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-001.phpt]
Bug GH-9916 002 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-002.phpt]
Bug GH-9916 003 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-003.phpt]
Bug GH-9916 004 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-004.phpt]
Bug GH-9916 005 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-005.phpt]
Bug GH-9916 006 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-006.phpt]
Bug GH-9916 007 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-007.phpt]
Bug GH-9916 008 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-008.phpt]
Bug GH-9916 009 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-009.phpt]
Bug GH-9916 010 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-010.phpt]
Bug GH-9916 011 (Entering shutdown sequence with a fiber suspended in a Generator emits an unavoidable fatal error or crashes) [Zend/tests/fibers/gh9916-011.phpt]
GC 049: Objects created during GC do not participate in the same collection [Zend/tests/gc/gc_049.phpt]
GH-15330 003: Do not scan generator frames more than once [Zend/tests/generators/gh15330-003.phpt]
GH-15330 004: Do not scan generator frames more than once [Zend/tests/generators/gh15330-004.phpt]
GH-15330 005: Do not scan generator frames more than once [Zend/tests/generators/gh15330-005.phpt]
GH-15330 006: Do not scan generator frames more than once [Zend/tests/generators/gh15330-006.phpt]
GH-16408: Array to string conversion warning emitted in optimizer [Zend/tests/gh16408.phpt]
Lazy objects: Initializer effects are reverted after exception (nested) [Zend/tests/lazy_objects/init_exception_reverts_initializer_changes_nested.phpt]
Lazy objects: RFC example 011 [Zend/tests/lazy_objects/rfc/rfc_example_011.phpt]
Test nullsafe in binary op [Zend/tests/nullsafe_operator/014.phpt]
Allow fetching properties in static initializer [Zend/tests/prop_const_expr/static_initalizer.phpt]
Hooks containing static variables [Zend/tests/property_hooks/static_variables.phpt]
Test binary operands exposing the same behavior at compile as at run time [Zend/tests/runtime_compile_time_binary_operands.phpt]
Static Variable Expressions [Zend/tests/static_variables/static_variable.phpt]
Zend signed multiply 64-bit, variation 2 [Zend/tests/zend_signed_multiply-64bit-2.phpt]

@marc-mabe

marc-mabe commented Jul 9, 2025

Copy link
Copy Markdown
Contributor Author

I think this should go through the RFC process. There have been suggestions to drop 32-bit support, so I'm sure at least some people would object to expanding support further. I could see this as an alternative to dropping 32-bit support iff we also dropped 4-byte zend_long support, as that would align 32 and 64-bit.

Yes, of course. At the current state I want to get to know the work needed to be done and do some benchmarks.

For dropping 32-bit support entirely it seems to be to early at least for WebAssembly + there might be people hoping for supporting older (in some cases not that old) systems - so I don't thing it would go through currently.

@iluuu1994 iluuu1994 added the RFC label Jul 9, 2025
@mvorisek mvorisek mentioned this pull request Jul 9, 2025
Closed
@marc-mabe
marc-mabe force-pushed the int64 branch 6 times, most recently from 4fb8912 to 6b552c1 Compare July 9, 2025 21:43
@marc-mabe
marc-mabe force-pushed the int64 branch 2 times, most recently from e28b737 to f41823a Compare July 13, 2025 20:22
@marc-mabe
marc-mabe force-pushed the int64 branch 2 times, most recently from 2c3b9d7 to 9cdbff2 Compare July 14, 2025 06:56
@marc-mabe marc-mabe mentioned this pull request Jul 16, 2025
Open
@marc-mabe marc-mabe mentioned this pull request Jul 26, 2025
Merged
@marc-mabe
marc-mabe force-pushed the int64 branch 3 times, most recently from 327a244 to 6baa1ec Compare November 16, 2025 14:41
@Sjord

Sjord commented Jul 8, 2026

Copy link
Copy Markdown
Contributor

The lower three bits of Bucket byte offset are used to store flags. If Bucket is aligned to 8 bytes, the lower three bits are always zero and this gives no problems. With the change of zend_long from 32 to 64 bits, the size of Bucket changed (to 28 bytes) and it is not aligned to 8 bytes anymore. The memory address and the flags overlap.

Changing Bucket to be 32 bytes fixes things:

typedef struct _Bucket {
        zval              val;
        zend_ulong        h;                /* hash value (or numeric index)   */
        zend_string      *key;              /* string key or NULL for numerics */
        uint32_t padding;
} Bucket;

The flags are ZEND_BIND_REF, ZEND_BIND_IMPLICIT, ZEND_BIND_EXPLICIT. They are stored here:

opline->extended_value = (uint32_t)((char*)value - (char*)CG(active_op_array)->static_variables->arData) | mode;

and extracted here:

value = (zval*)((char*)ht->arData + (opline->extended_value & ~(ZEND_BIND_REF|ZEND_BIND_IMPLICIT|ZEND_BIND_EXPLICIT)));

@Sjord Sjord mentioned this pull request Aug 6, 2026
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@Girgias Girgias Girgias left review comments
@dstogov dstogov Awaiting requested review from dstogov dstogov will be requested when the pull request is marked ready for review dstogov is a code owner
@bukka bukka Awaiting requested review from bukka bukka will be requested when the pull request is marked ready for review bukka is a code owner
@devnexen devnexen Awaiting requested review from devnexen devnexen will be requested when the pull request is marked ready for review devnexen is a code owner
@derickr derickr Awaiting requested review from derickr derickr will be requested when the pull request is marked ready for review derickr is a code owner

Assignees

No one assigned

Labels

ABI break Category: Build System Category: Engine Extension: calendar Extension: date Extension: ftp Extension: gmp Extension: mysqlnd Extension: opcache Extension: pcntl Extension: shmop Extension: soap Extension: sockets Extension: standard Extension: sysvmsg Extension: tidy Extension: xml Extension: zend_test RFC SAPI: cli SAPI: phpdbg

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@marc-mabe @iluuu1994 @Sjord @Girgias