| Age | Commit message (Collapse) | Author |
|---|
|
PathDownloader builds an ArchivableFilesFinder and hands it to
Filesystem::mirror as the iterator that selects what to copy, so a path
package installed by mirroring drops what .gitattributes marks
export-ignore, what .gitignore excludes and the VCS directories. The
port dropped the argument, and mirrored the source tree whole.
Every method mirror() calls on the entries it walks is derived from the
pathname, so the iterator is modeled as a list of pathnames rather than
as a Traversable of SplFileInfo.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
doRemove() renames a directory to a random hidden name before emptying
it, and undoes that rename when the final rmdir fails, so a concurrent
process cannot recreate the path mid-removal. It also walks one level at
a time through FilesystemIterator instead of flattening the whole tree,
and lets an inner rmdir failure pass, both as upstream does.
copy() keeps the mode fopen($targetFile, 'w') would have left rather
than the origin's, symlink() and mirror() call readlink() and
getLinkTarget() where upstream does, and a directory iterator that
cannot be opened propagates its UnexpectedValueException instead of
being swallowed or flattened into an IOException.
Error message text is out of scope per docs/known-incompatibilities.md,
so the three TODO(phase-c) markers that only tracked wording are gone,
along with linkException()'s Windows-only branch. So are the arguments
no caller varies -- symlink()'s copyOnWindows, mirror()'s iterator and
options, copy()'s overwriteNewerFiles -- which removes the last
TODO(phase-c) in the file.
New shim functions: readlink, filesystem_iterator, stream_is_local,
strrev and SplFileInfo::getLinkTarget. base64_encode takes bytes so
random_bytes() can feed it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
The shim raises no PHP-level errors, so error_get_last() always returned
None and every message built from it lost its trailing reason. Now that
the fs mutators and php_strip_whitespace carry an io::Error, take the
reason from the call that actually failed instead of a global last-error
slot.
Filesystem::unlinkImplementation returns that error rather than a bool so
ensureDirectoryExists, unlink and rmdir can report it, and PhpFileParser
appends it to the "following message may be helpful" hint. The wording
is Rust's io::Error text, not PHP's warning text, for the same reason
noted in symfony/filesystem.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
mkdir, rmdir, unlink and symlink each had a bool version and a _result
twin returning the io::Error, which left two names for one call. Keep
only the Result form and let the callers that want a boolean spell out
.is_ok(). Call sites that discard the outcome, as their PHP originals
do, are unchanged.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Symfony wraps its native filesystem calls in box(), which captures the
warning text of a failed call into self::$lastError and appends it to
the IOException message. The port dropped that text because the shim's
mkdir, rmdir, unlink and symlink only return PHP's bare false.
Give each of those a Result-returning variant carrying the io::Error,
and use it to restore the appended reason. That also makes two
conditions portable: doRemove()'s "Permission denied" test, and
linkException()'s Windows error code 1314 special case.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
InstalledVersions is a runtime API for plugins and project code; Composer
itself never reads it. Its consumers run in the PHP worker against the copy
FilesystemRepository dumps to vendor/composer/InstalledVersions.php, whose
static state is already kept in sync by __shirabe_installed_versions_reload.
Nothing in Rust read the mirrored statics, so reload() and the reflection
setters were no-ops.
The tests covered only the Rust port, not the PHP class the worker loads, so
they assert nothing about compatibility; they are left as todo!() skeletons.
This also removes the shim functions method_exists, php_dir and
require_php_file, whose only caller was the deleted module.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Composer turns the cycle collector off around the dependency solver, but
Rust has no GC, so the gc_collect_cycles/gc_disable/gc_enable shims were
no-ops. Remove the call sites and the now-callerless shim functions.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
The only live caller passed the constant 'array_merge', so the list
concatenation is written out in place, as ClassLoader::get_prefixes
already does. That leaves the shim function without callers.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Its doc-by-reference sibling call_user_func_array now carries the full
reason text.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
|
Problem reported the platform php version from the shim's compile-time
PHP_VERSION constant, and the shim's phpversion() hit todo!() for any
extension. Go through the RPC bridge instead, as PHP's
Composer\Platform\Runtime does, and drop the shim function that now has
no callers.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
shirabe never runs on HHVM, so the HHVM_VERSION_ID branches are dead.
Dropping them leaves the shim's unimplemented `constant()` without a
caller, so remove it as well.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
|
The shim reported setExternalAttributesName as missing, so ZipArchiver took
the branch for libzip below 0.11.2 and every archived entry got the zip
crate's default mode. The attributes now travel as arguments to add_file and
add_empty_dir, because the crate fixes an entry's external attributes when
the entry is started and offers no way to amend one already written.
unix_permissions keeps only the low 9 mode bits, on top of which the crate
restores S_IFREG and S_IFDIR, so setuid/setgid/sticky bits and other file
types are still dropped.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
|
The shim's PATH_SEPARATOR was hardcoded to ":", so every PATH list was
split and joined on the wrong character off Unix. std::env::split_paths
and join_paths use the platform's separator, which also lets the bin-dir
membership test in EventDispatcher compare list entries instead of
regex-matching the raw PATH string.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
The shim's DIRECTORY_SEPARATOR was hardcoded to "/", so every ported
`'\\' === DIRECTORY_SEPARATOR` check compared against a constant that
does not track the target platform. std::path::MAIN_SEPARATOR and
MAIN_SEPARATOR_STR carry the same meaning as PHP's constant and resolve
per platform, so the Windows branches are selected on Windows targets.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
The deadline is a constant baked in at build time, so PHP's define() of
COMPOSER_DEV_WARNING_TIME cannot make Application take that branch.
Application now holds the deadline in a field that __set_dev_warning_time
overrides, letting testDevWarning run instead of staying ignored. The
define() shim, whose only caller was that test, goes away with it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Composer hands the exception it caught to the one it throws in its place,
so `getPrevious()` reaches the cause and Application's renderer prints the
whole chain. Every ported site dropped it, because the flat exception
structs had nowhere to put one. `AnyThrowable::into_previous` turns the
caught error into that argument, and the 13 sites now pass it.
`getCode()` came along for the ride at the four sites that derive the new
exception's code from the caught one (PharArchiver, ArrayLoader x2), and
ComposerRepository's message now names the caught exception's class
instead of the literal "Exception".
GitHubDriver::attemptCloneFallback took the previous exception's message
and appended it to its own, which no `\RuntimeException('Fallback to git
driver disabled')` in Composer ever says; it now chains it instead.
Git::syncMirror restores what PHP's `finally` does to an exception in
flight: the `git remote set-url` that scrubs credentials back out of the
URL runs in a `finally`, and when it fails PHP propagates *its* exception
over the one already leaving, chaining the displaced one as previous. The
port discarded the finally's result, so a failure to scrub the URL was
reported as a successful mirror sync. `AnyThrowable::set_previous`
models the engine-level chaining.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Ported exceptions were flat structs reached with `downcast_ref`, so
Composer's `catch (\RuntimeException $e)` only matched the exact leaf
type and `get_class($e)` had nothing to report. Each exception now
embeds an instance of the class it extends and travels inside an
`AnyThrowable`; `Catch::catch`/`catch_mut` walk that chain, and
`PhpClass::php_class_name` yields the PHP FQCN.
Dropping the `std::error::Error` impls from the exception types leaves
`AnyThrowable` as the only route into an `anyhow::Error`, so the walk
cannot be bypassed. A `no_exception_downcast` linter catches the
`downcast::<X>()` calls that would now silently answer `None`.
Three sites change behavior as a result: the `TransportException`
exit-code override reaches `MaxFileSizeExceededException`, the
`catch (\LogicException)` in findSimilar() reaches its subclasses, and
rendered exception titles carry the real class name rather than a
guess. `get_class_err()` is no longer a `todo!()`, which re-enables
FilesystemRepositoryTest::testCorruptedRepositoryFile.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Composer defines COMPOSER_DEV_WARNING_TIME from its phar stub when the
compiled version is a commit hash rather than a tag, so the value exists
only in the build artifact. It was a todo!() in the PHP shim, leaving the
warning branch in Application::do_run unreachable.
A build script now derives it the way Compiler does, from git describe
and the HEAD commit date, and composer::COMPOSER_DEV_WARNING_TIME holds
the result as a Rust constant instead of a runtime-defined one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Getting a rule's identity meant spl_object_hash(&*rule.borrow()), so the
solver borrowed a RefCell just to read an address, and a second function
spl_object_hash_process existed because one generic fn cannot tell Rc<T>
from &T.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
PHP's version_compare takes its operator as a string, so the shim's port
did too, and Constraint carried two families of operator constants plus
translation tables to convert between the string form and its own int
codes. Five copies of those tables had accumulated across Constraint,
CompilingMatcher and the plugin value bridge.
version_compare now takes a CmpOp, which makes an invalid operator
unrepresentable and removes the tables' reason to exist. Constraint
stores a CmpOp and keeps only the string parsing its constructor needs;
getOperator, compile and CompilingMatcher::match speak CmpOp as well.
PHP's OP_* numbering stays observable: a plugin reads the raw integer off
the Constraint object over RPC, so get_operator_constant and its new
inverse hold that 0..5 mapping.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
|
|
Enable clippy::multiple_inherent_impl and fix the 21 sites it reports.
Types whose inherent methods were spread across two or three impl blocks
now keep them in a single block; only the impl headers move, no method
bodies change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Replace the todo!() bodies with real ports. Four autoload-generator
tests now run for real; the rest stay #[ignore]d, but each ignore reason
now names the concrete missing symbol instead of a vague subsystem.
Production additions the ports need: the deprecated
AuthHelper::addAuthenticationHeader wrapper,
EventDispatcher::__set_dispatch_script_override as the seam for PHPUnit
onlyMethods(['dispatchScript']), and a define() stub in the shim.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
PlatformRepository probes ResourceBundle, IntlChar and Imagick to derive
lib-icu-cldr, lib-icu-unicode and lib-imagick-imagemagick. Those probes ran
against the shim's hard-coded class_exists allowlist, which never names
them, so the packages were silently missing: on a machine with intl,
`show --platform` listed fewer libraries than upstream Composer does.
The runtime seam now asks the real PHP: hasClass over RPC, and construct /
invoke through the worker for the three classes PlatformRepository reaches.
A live PHP object has no PhpMixed counterpart, so the seam answers with the
entries the caller reads off it. The seam's own callers read those entries
instead of returning null and the empty string.
Two addLibrary calls also had replaces and provides swapped, dropping
`lib-libxslt replaces lib-xsl` and `lib-zip-libzip replaces lib-zip`.
`show --platform` now matches upstream Composer byte for byte, and all 59
provideLibraryTestCases datasets pass.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
getDownloaderForPackage reports get_class($downloader) when the resolved
downloader's installation source does not match. Rust has no runtime class
name, so the message was built from a shim stub that panicked instead —
the error could never be returned.
DownloaderInterface now requires PhpClass, the trait already used for the
same purpose on Command, and each downloader states the name PHP reports.
That leaves get_class_obj without callers, so it is gone.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
pathinfo only ever receives a string and only ever returns one for the
single-component options it supports, so the PhpMixed wrapping forced
every call site to pack and unpack the value again.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
|
|
Port the Symfony Filesystem branches this file had left out:
* copy() preserves the origin mtime via touch()
* exists() rejects paths longer than PHP_MAXPATHLEN - 2, so its return
type becomes anyhow::Result<bool>
* doRemove()'s symlink branch keeps the DIRECTORY_SEPARATOR disjunct,
which stops it from throwing on Unix where upstream never does
* symlink() normalizes separators and mirrors instead of linking when
copyOnWindows is set
* mirror() skips entries whose real path is the target directory or was
already created earlier in the same call
PHP_MAXPATHLEN is new in shirabe-php-shim.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
The shim's filesystem entry points took `&str` even though each one resolves
to a local path through `std::fs` or a syscall, so callers holding a `PathBuf`
had to stringify it at the call site. They now take `impl AsRef<Path>`, the
form `file_exists`, `is_dir`, `unlink` and the rest of the already-converted
set use.
`Phar`, `PharData` and `ZipArchive` keep their archive path as a `PathBuf`.
`PharData::compress` names the compressed sibling by appending the suffix to
the file name rather than formatting the path into a `String`.
Arguments PHP resolves through a stream wrapper (`fopen`, `file_put_contents`,
`include`) still take `&str`, as do the byte-string operations (`dirname`,
`basename`, `pathinfo`) and archive-internal entry names, which are `/`-joined
logical names rather than OS paths.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Port Application::getPluginCommands: resolve the local composer with
plugins force-enabled, fall back to Factory::createGlobal, and collect
commands from CommandProvider capability adapters. PhpCommandProxy now
mirrors name/description/aliases/hidden over RPC so `list` output
matches upstream; input definitions remain TODO(plugin).
PhpClass::php_class_name returns an owned String because PHP-backed
proxies only know their class at runtime (the override-skip warning
prints get_class). CommandProvider::getCommands hands out shared
Rc<RefCell> handles since the commands are stored in the application.
Factory::createGlobal now propagates createConfig errors instead of
swallowing them, and Application::getComposer only catches the
exception classes upstream catches, so a ParsingException reaches
doRun's GithubActionError path as in Composer.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
PhpMixed::Object hit a todo!(); the capability-name validation message in
PluginManager (var_export of an invalid getCapabilities value) reaches it
with a stdClass. Match the PHP 8.5.8 output byte for byte — the
"(object) array(...)" shape, properties indented one space deeper than
array elements, keys always quoted strings — verified against the real
PHP as the oracle and pinned by a unit test. Other classes would render
as \Class::__set_state(...), which stays out of reach because
PhpMixed::Object carries no class name.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
mb_check_encoding
http_get_last_response_headers()/http_clear_last_response_headers() now back
a thread-local store with a recording hook for the (still unported) HTTP
stream layer; with no request recorded they return None, matching PHP.
mb_check_encoding gains the trivial ASCII arm; other encodings still need
the mbstring tables and stay todo!().
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
Phar::running has its only Composer call site in SelfUpdateCommand, and
self-update will not go through a phar in Shirabe. The .phar writing API
(and the SHA512 constant it takes) is only used by Composer's Compiler,
which packages composer.phar and has no counterpart in a native binary.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Retag every Shirabe-authored TODO comment to one of the fixed tags:
phase-c, phase-d, plugin, php-runtime, phase-e.
Upstream-authored TODO comments from Composer/Symfony are left
untouched to preserve the ported code shape.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
Rust has no runtime class name, so `Command::get_class` existed purely to
let each command hand back its PHP class name, supplied through the
two-argument variant of `delegate_command_trait_impls_to_inner!` at the
impl site. Replace it with a general `PhpClass` trait plus an
`impl_php_class!` macro, so the name is stated once next to the type
definition and the mechanism is reusable outside commands.
`Command` gains `PhpClass` as a supertrait and drops `get_class`, and
`VcsDriverKind`'s hand-rolled `php_class_name` table moves onto the trait.
Behavior is unchanged: the same class-name strings are reported, and the
base command state still panics when asked for a name it cannot supply.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Several shim functions were left as todo!() because the standard library
exposes no equivalent and no syscall crate was available. Adding nix
unblocks them:
- proc_open now wires descriptors beyond stderr, creating the pipe itself
and installing the child end with dup2(2) from pre_exec
- proc_terminate and posix_kill deliver arbitrary signals via kill(2)
- get_current_user reports the owner of the running executable
- php_uname answers every mode from uname(2) instead of only "s" and "r"
It also closes gaps that were previously approximated:
- fstat stats the stdio streams and pipes rather than reporting failure
- touch stamps mtime/atime on an existing path, including directories
- is_writable/is_executable use access(2) instead of permission bits
- umask falls back to the read-modify-write umask(2) off Linux
The hand-written repr(C) structs and extern "C" declarations for getpwuid,
utime, statvfs, fcntl and select are replaced by their nix wrappers, which
in turn lets disk_free_space drop its Linux-only cfg.
cli_set_process_title, setproctitle and the pcntl_signal pair stay as
todo!(): the first two need access to the process's own argv block, and
the latter two depend on the signal-handling subsystem rather than on
sigaction(2) itself.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
|
|
Rust's &str is always valid UTF-8, so the mbstring/iconv sanitization
chain can never trigger. Reduce it to a no-op with a TODO(phase-c)
marker: once the codebase strictly separates Vec<u8> from String, this
should take &[u8] and convert lossily. This removes the last caller of
the php-shim iconv(), so delete it as well.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
Remove the eight Cursor methods that have no callers, and the sscanf
shim whose only caller was Cursor::get_current_position.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
The only callers were trivial fixed-format uses: reading the first
four hash bytes as a native int, splitting in_addr byte strings, and
building a constant ZIP EOCD record. Each site now does the byte
manipulation directly, so the general-purpose shims are no longer
needed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
%e/%E/%g/%G never appear in Composer's own format strings (verified by
sweeping composer/src and vendor); supporting PHP's exponent formatting
is intentionally out of scope, so fail permanently instead of marking
the specifiers as pending work.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
Adopt the tar, flate2, and bzip2 crates to fill in the phar.rs and
compress.rs todos: PharData tar/zip reading, building, and whole-archive
compression, plus a native .phar reader that follows the php.net
file-format manual and verifies hash-based signatures. Callers now
propagate the constructor/extract errors PHP throws, and fwrite accepts
byte strings so gzread no longer needs lossy UTF-8.
The native .phar writing API stays todo!() (no call sites; Composer's
Compiler is not ported) and OPENSSL phar signatures are accepted
unverified (TODO(phase-c)).
This unblocks Tar::getComposerJson and the tar/phar/gzip downloaders;
tar_test (7), artifact_repository_test (2), and phar_archiver_test zip
(1) are un-ignored. The archive command itself still panics because
ArchiveManager::archive always generates glob excludes whose look-ahead
regexes the regex crate cannot compile; converting those patterns to
regex-compatible ones is a separate, still-undecided work item.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
The audit in .ken/php-shim-copying.md judged 14 functions in
shirabe-php-shim (plus php_wordwrap in shirabe-external-packages) to be
line-by-line transcriptions or structural imitations of php-src. PHP's
relicensing to 3-clause BSD makes keeping them legal, but the boundary
between BSD-derived and MIT code was invisible in the source tree.
Moving them into their own crate puts the license into the build
metadata (so NOTICE generation follows the binary), makes a reverse
dependency a compile error, and encodes the origin in the module path,
which mirrors php-src's ext tree. Each function records its origin in a
fixed-format doc comment, and a new php_src_derivation_boundary linter
fails if `php-src` appears in any Rust source outside the crate.
Public paths under shirabe_php_shim:: are unchanged: functions that are
themselves derived are re-exported with `pub use`, and the wrappers that
only validate arguments stay on the MIT side.
This also resolves the duplicate wordwrap implementation.
shirabe_php_shim::wordwrap was todo!(), so SymfonyStyle::block panicked,
while shirabe-external-packages carried its own copy. Both now go
through the single port, verified against real PHP on 13 cases covering
multi-character breaks and cut.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
diagnose used to read hardcoded shim stubs, so it described a fictional
runtime: OPENSSL_VERSION_NUMBER was always 0 and tripped the TLSv1.1/1.2
check, PHP_BINARY and OPENSSL_VERSION_TEXT were empty, and the extension,
function and ini probes answered from a fixed table.
The PHP worker gained a `diagnose` entry that returns every fact the
command needs as one PHP array, cached in a OnceLock so the several call
sites share a single round trip. Reading it back needed array support in
the serialize() parser, which in turn lets get_loaded_extensions and
get_all_ini_files return real lists instead of comma-joined strings.
Also fixes the openssl_version message, which dropped strstr()'s
before_needle argument during the port, and check_connectivity's
allow_url_fopen test, which did not follow PHP string truthiness.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
|
Problem::getPrettyString sorts same-priority reasons by
getSortableString(), whose RULE_LEARNED key is a '-'-joined literal id
string (e.g. "-95"). PHP's <=> compares two numeric strings
numerically, but the port used plain String::cmp (byte-wise), which
reverses relative order for same-length negative-number keys. Added
shirabe_php_shim::loosely_compare to approximate PHP's <=> for this
pattern (numeric compare when both sides parse as numbers, else byte
compare) and switched the sort comparator to use it.
The diagnosis this replaces (from the commit being amended) blamed
Pool package-id assignment order diverging from PHP under
COMPOSER_POOL_OPTIMIZER=0. That was disproven this session: direct
instrumentation of both PHP and the Rust port confirmed identical
relative package-id order, including on the ~335-package
github-issues-7665 fixture (ids matched up to a constant +2 offset
from a platform-mock package count difference). The sort comparator
was the actual bug, not package loading order.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
|
Two shim gaps made the extension-related branches of solver-problem
messages wrong:
- phpversion($ext) with a non-empty extension can't be known statically
(it's shirabe-php-shim's todo!()); Problem::get_missing_package_reason
now calls shirabe_php_rpc::phpversion, the same RPC bridge
platform::runtime::Runtime::get_extension_version already uses.
- extension_loaded's hardcoded allowlist was missing "pcre", a mandatory
always-compiled-in PHP extension, so ext-pcre was misreported as
"missing from your system" instead of "disabled by your platform
config" whenever a platform override disabled it.
Also fix XdebugHandler::getAllIniFiles() always returning `[""]`
(a php-runtime stub): create_extension_hint()'s early-return guard
(`paths[0] empty && len==1`) fired unconditionally, silently dropping
the entire "To enable extensions..." hint from every solver-problem
message that mentions missing extensions. shirabe-external-packages
can't depend on shirabe-php-rpc (shirabe-php-rpc already depends on
shirabe-external-packages), so IniHelper::get_all() queries a new
get_all_ini_files RPC command directly instead of going through the
stub.
This exposed that XdebugHandler is never constructed with a name
because bin/composer's restart-without-Xdebug bootstrap was never
ported to main.rs, making COMPOSER_ORIGINAL_INIS-driven behavior
unreachable; documented with a TODO(phase-c) and updated
ini_helper_test.rs's ignore reasons (and ignored test_with_no_ini,
which only passed before by coincidence with the old stub's constant
output) to match.
|
|
The b'>' match arm in strip_tags's state machine never pushed the
character to the output buffer when encountered outside a tag (state
0), unlike every other special-character arm (!, ?, -, and the
catch-all) which does push in that state. Every literal '>' not part
of an HTML-like tag was silently dropped.
This corrupted "=>" into "=" in Composer's operation trace strings
(e.g. "Upgrading foo/bar (1.0.0 => 1.1.0)"), which go through
strip_tags to remove the <info>/<comment> markup before comparison,
un-ignoring 82 integration tests that were asserting on that exact
arrow.
|
|
reload to PHP runtime
Rust has no PHP interpreter, so eval() can never be ported faithfully.
safely_load_installed_versions()'s job of priming Composer\InstalledVersions
before plugins run only matters within a single shared PHP process, which
the RPC-based plugin architecture does not have; the PHP runtime process
can call InstalledVersions::reload() itself instead.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
