We are experiencing an issue when building PHP with DTrace enabled with
SystemTap (see GH-11847).† The issue is caused by inappropriate use C
preprocessor detected by GNU Autoconf in our “configure” script. C
preprocessor configuration found by AC_PROG_CPP macro is portable only
to run on files with “.c” extension.‡ However, statically-defined tracing
is described by D programs with “.d” extension which causes the issue.
We experience this even on typical Linux distribution with GNU Compiler
Collection (GCC) unless we override the defaults detected by our
“configure” script.
Many major Linux distributions use SystemTap to provide “dtrace”
utility. It relies on both external C preprocessor and external C
compiler. C preprocessor can be customized via CPP environment variable.
Similarly, C compiler can be customized via CC environment variable. It
also allows customization of C compiler flags via CFLAGS environment
variable. We have recently aligned both CPP and CC environment variable
with C preprocessor and C compiler we use to build regular C source code
as provided by our “configure” script (see GH-11643).* We wanted to
allow cross-compilation on Linux for which this was the only blocker. C
compiler flags from CFLAGS_CLEAN macro have already been in place since
versions 5.4.20 and 5.5.4 from 2013-09-18.
We had modified all “dtrace” invocations in the same way to make it look
consistent. However, only the C compiler (CC environment variable) is
necessary to for cross-compilation. There have never been any reported
issue with the C preprocessor. We acknowledge it would be great to allow
C preprocessor customization as well. However, the implementation would
require a lot of effort to do correctly given the limitations of
AC_PROG_CPP macro from GNU Autoconf. This would be further complicated
by the fact that all DTrace implementations, not just SystemTap, allow C
preprocessor customization but Oracle DTrace, Open DTrace, and their
forks do it differently. Nevertheless, they all default to “cpp” utility
and they all have or had been working fine. Therefore, we believe simply
removing CPP stabilizes “dtrace” invocation on Linux systems with
SystemTap and aligns it with other system configurations on other
platforms, until someone comes with complete solution with custom “m4”
and “make” macros, while our build system on Linux with SystemTap
supports cross-compilation.
Fixes GH-11847
Closes GH-12083
† https://github.com/php/php-src/issues/11847
‡ https://www.gnu.org/savannah-checkouts/gnu/autoconf/manual/autoconf-2.71/autoconf.html#index-AC_005fPROG_005fCPP-1
* https://github.com/php/php-src/issues/11643
With DTrace support enabled during ./configure, our custom Autoconf
macro PHP_INIT_DTRACE creates make rules to generate header and object
files using dtrace utility. SystemTap† implementation of dtrace relies
on other utilities to provide header preprocessing and final object file
compilation. These utilities are configured by common environment
variables with common defaults:‡
* preprocessor from CPP defaults to “cpp”
* compiler from CC defaults to “gcc”
* compiler arguments can be expanded with CFLAGS
This has been in SystemTap since version 1.5 released on 2011-05-23. We
have been setting CFLAGS for dtrace since 717b367 released in versions
5.4.20 and 5.5.4 on 2013-09-18. This change fixed build against
SystemTap. It fixes majority of cases since practically all free Linux
distributions use SystemTap for DTrace-like dynamic tracing and
practically all of them use GCC or compatible compiler suite. However,
this becomes an issue when cross-compiling using GCC because utility
names contain target triplets. Autoconf already handles
cross-compilation well —setting correct CC and CPP make macros
(variables).
Therefore, we simply set CC and CPP environment variables using
respective macros when executing dtrace. Although SystemTap dtrace does
not always use CC nor CPP, we set it every time. SystemTap documentation
does not talk about this at all¶, so it is safer to always set it. We
also follow how we set CFLAGS every time in the past.
Original (or ported) DTrace mainly used on Oracle Linux, Solaris and
macOS ignores these and does not support cross compilation.§
† Well-known dynamic tracing infrastructure for Linux compatible with
statically-defined tracing from DTrace.
‡ https://sourceware.org/git/?p=systemtap.git;a=blob;f=dtrace.in;h=73a6f22e2de072773c692e3fea05c4b8cf814e43;hb=ebb424eee5599fcc131901c0d82d0bfc0d2f57ab
¶ https://sourceware.org/systemtap/man/dtrace.1.html
§ https://docs.oracle.com/cd/E88353_01/html/E72487/dtrace-8.html
Closes GH-11643
The next generation of C compilers is going to enforce the C standard
more strictly:
https://wiki.gentoo.org/wiki/Modern_C_porting
One warning that will soon become an error is -Wstrict-prototypes.
This is relatively easy to catch in most code (it will fail to
compile), but inside of autoconf tests it can go unnoticed because
many feature-test compilations fail by design. For example,
$ export CFLAGS="$CFLAGS -Werror=strict-prototypes"
$ ./configure
...
checking if iconv supports errno... no
configure: error: iconv does not support errno
(this is on a system where iconv *does* support errno). If errno
support were optional, that test would have "silently" disabled
it. The underlying issue here, from config.log, is
conftest.c:211:5: error: function declaration isn't a prototype
[-Werror=strict-prototypes]
211 | int main() {
This commit goes through all of our autoconf tests, replacing main()
with main(void). Up to equivalent types and variable renamings, that's
one of the two valid signatures, and satisfies the compiler (gcc-12 in
this case).
Fixes GH-10751
(I agree that null should come last, but then it should rather throw with a proper message than emit an undefined key warning.)
Signed-off-by: Bob Weinand <bobwei9@hotmail.com>
Extended docblock types, according to psalm or phpstan conventions may include array shapes, callable signatures etc.. These are now ignored by ignoring any nested parenthesized expression (followed by optional :type) at the end of the type.
1. Implementation based on https://github.com/WojciechMula/base64simd
2. Only runtime path is added to reduce the complexity of SIMD variants.
3. Expand test case to cover SIMD implementation.
Signed-off-by: Frank Du <frank.du@intel.com>
The release VMs already enforced this, but PHP's configure script did
not.
re2c 0.13.5, which timelib's date/time parser requires is no longer
compatible with the current version of Zend/zend_language_scanner.l, as
it starts spinning in a loop.
As with other SIMD-accelerated functions in php-src, the new UTF-16
encoding and decoding routines can be compiled either with AVX2
acceleration "always on", "always off", or else with runtime detection
of AVX2 support.
With the new UTF-16 decoder/encoder, conversion of extremely short
strings (as in several bytes) has the same performance as before,
and conversion of medium-length (~100 character) strings is about 65%
faster, but conversion of long (~10,000 character) strings is around
6 times faster.
Many other mbstring functions will also be faster now when handling
UTF-16; for example, mb_strlen is almost 3 times faster on medium
strings, and almost 9 times faster on long strings. (Why does mb_strlen
benefit more from AVX2 acceleration than mb_convert_encoding? It's
because mb_strlen only needs to decode, but not re-encode, the input
string, and the UTF-16 decoder benefits much more from SIMD
acceleration than the UTF-16 encoder.)
As an example:
should be translated to:
ZVAL_LONG(&attribute_Attribute_class_test_arg0, ZEND_ATTRIBUTE_TARGET_FUNCTION | ZEND_ATTRIBUTE_TARGET_METHOD);
These are mandatory in C99, so it's a pointless waste of time to check
for them.
(Actually, the fixed-size integer types are not mandatory, but if they
are really not available on some theoretical system, PHP's fallbacks
won't work either, so nothing is gained from this check.)
Until https://github.com/php/php-src/pull/10098 default constructors were sometimes documented, sometimes omitted. The above PR made adding documentation for constructor of all non-abstract classes required. However, this is not desired when such a class inherits a constructor from its parent. The current PR fixes the behavior the following way:
- documents inherited constructors (along with destructors)
- makes it possible to generate/replace the methodsynopsis of implicit default constructors which don't have a stub counterpart
* PHP-8.2:
Replace another root XML element format to the "canonical" one
Remove the superfluous closing parentheses from class synopsis page includes
Always include the constructor on the class manual pages
Backport methodsynopsis role attributes changes from master
