The non-standard major(), minor(), and makedev() can be defined as
macros. These are usually used together with the Autoconf macro
AC_HEADER_MAJOR, which defines the MAJOR_IN_MKDEV if sys/mkdev.h is
available, or MAJOR_IN_SYSMACROS if sys/sysmacros.h is available.
On Solaris/illumos they are in the sys/mkdev.h header (macro defined to
libc implementation) and in sys/sysmacros.h (macro defined with binary
operators and bits shifting). On systems with musl and glibc 2.28 or
later they are defined in sys/sysmacros.h, in glibc 2.27 and earlier
they were in sys/types.h. On BSD-based systems and macOS they are in the
sys/types.h.
Autoconf 2.70 has fixed the AC_HEADER_MAJOR macro, so it detects the
headers properly due to glibc 2.25 throwing deprecation warnings when
using the macros from sys/types.h. With Autoconf 2.69 and earlier the
ac_cv_header_sys_types_h_makedev cache variable can skip the
improper sys/types.h check in the macro.
This change syncs the usage within the ext/fileinfo/libmagic bundled
library and ext/posix.
When sys/mkdev.h header is available, code includes that, otherwise
it conditionally includes the sys/sysmacros.h. The ext/posix has
additional check whether linker sees the makedev, otherwise it checks
if makedev is declared within the given set of headers accoring to the
AC_HEADER_MAJOR logic. Previously the AC_CHECK_FUNCS didn't detect it.
The ext/pdo_mysql symbol has been once used together with the removed
ext/mysql extension and isn't defined on Windows neither used in the
code anymore.
The '__asm__ goto' support is properly recognized by a simpler linking
check instead of a run check for easier cross-compilation. The
compile only check (AC_COMPILE_IFELSE) might produce false positives
results with certain compiler options.
In the same time, let's not verify implementation aliases since they may now legitimately differ from their aliased function/method counterparts (think about the ext/dom refactoring where e.g. many return type declarations have changed). Additionally, unnecessary `@no-verify` tags are cleaned up.
Strict error checking is always true for classes in "new DOM".
This means that we always throw an error when calling
`php_dom_throw_error`, and therefore the false return value is not
actually possible.
Also change the stub to reflect this.
This is not just an issue due to missing initialization since moving the state
struct directly into the module globals. In earlier versions changing the mode
to `MT_RAND_PHP` within a single request would also affect the mode for
subsequent requests.
Original commit message follows:
This is a follow-up fix for GH-13579. The issue was detected in the nightly
MSAN build.
(cherry picked from commit bf0abd1629)
It's illegal to return from a bailout because that doesn't restore the
original bailout data. Return outside of it.
Test by YuanchengJiang
Closes GH-13689.
- Remove a check for an always-true curl version check
- Remove a `TODO` comment for curl_version_info check that should be always available since libcurl 7.10
This fixes two warnings when building litespeed SAPI:
```
.../php-src/sapi/litespeed/lscriu.c:312:14: warning:
'LSCRIU_Error_File_Name' defined but not used
[-Wunused-function]
312 | static char *LSCRIU_Error_File_Name(char *pchFile, int max_len)
| ^~~~~~~~~~~~~~~~~~~~~~
.../php-src/sapi/litespeed/lscriu.c:102:12: warning:
's_criu_debug' defined but not used [-Wunused-variable]
102 | static int s_criu_debug = 0;
| ^~~~~~~~~~~~
```
This was tested on OpenBSD sparc64 and all fiber related tests pass.
On OpenBSD stackghost prevents the modification of the return address
and therefor an extra trampoline is needed in make_fcontext(). This
should not matter on other OS implementing sysv ABI and the trampoline
should work there as well.
Close GH-13382.
The embed SAPI has been around for quite a while now, and many apps
already use it in production. It can be marked as non experimental to
avoid confusion.
[skip ci]
