php-src/UPGRADING.INTERNALS
2016-06-20 18:03:19 +02:00

83 lines
2.9 KiB
Plaintext

PHP 7.1 INTERNALS UPGRADE NOTES
0. Wiki Examples
1. Internal API changes
e. Codepage handling on Windows
f. Path handling on Windows
2. Build system changes
a. Unix build system changes
b. Windows build system changes
3. Module changes
================
0. Wiki Examples
================
The wiki contains multiple examples and further explanations of the internal
changes. See: https://wiki.php.net/phpng-upgrading
========================
1. Internal API changes
========================
e. Codepage handling on Windows
A set of new APIs was introduced, which allows to handle codepage
conversions. The corresponding prototypes and macros are contained
in win32/codepage.h.
Functions with php_win32_cp_* signatures provide handling for various
codepage aspects. Primarily they are in use at various places in the
I/O utils code and directly in the core where necessary, providing
conversions to/from UTF-16. Arbitrary conversions between codepages
are possible as well, whereby UTF-16 will be always an intermediate
state in this case.
For input length arguments, the macro PHP_WIN32_CP_IGNORE_LEN can be
passed, then the API will calculate the length. For output length
arguments, the macro PHP_WIN32_CP_IGNORE_LEN_P can be passed, then
the API won't set the output length.
The mapping between encodings and codepages is provided by the predefined
array of data contained in win32/cp_enc_map.c. To change the data,
a generator win32/cp_enc_map_gen.c needs to be edited and run.
f. Path handling on Windows
A set of new APIs was introduced, which allows to handle UTF-8 paths. The
corresponding prototypes and macros are contained in win32/ioutil.h.
Functions with php_win32_ioutil_* signatures provide POSIX I/O analogues.
These functions are integrated in various places across the code base to
support Unicode filenames. While accepting char * arguments, internally
the conversion to wchar_t * happens. Internally almost no ANSI APIs are
used, but directly their wide equivalents. The string conversion rules
correspond to those already present in the core and depend on the current
encoding settings. Doing so allows to move away from the ANSI Windows API
with its dependency on the system OEM/ANSI codepage.
Thanks to the wide API usage, the long paths are now supported as well. The
PHP_WIN32_IOUTIL_MAXPATHLEN macro is defined to 2048 bytes and will override
the MAXPATHLEN in files where the header is included.
The most optimal use case for scripts is utilizing UTF-8 for any I/O
related functions. UTF-8 filenames are supported on any system disregarding
the system OEM/ANSI codepage.
========================
2. Build system changes
========================
a. Unix build system changes
b. Windows build system changes
========================
3. Module changes
========================