Remove postmortem of bridge work

Add the macOS/arm64 UNO exception RTTI post-mortem

The comments in solenv/src/component.map and solenv/bin/addsym-macosx.sh point
here rather than restating why keyless-class typeinfo cannot be expo

Add the macOS/arm64 UNO exception RTTI post-mortem

The comments in solenv/src/component.map and solenv/bin/addsym-macosx.sh point
here rather than restating why keyless-class typeinfo cannot be exported.

show more ...

Run the remaining Extension Manager dialogs on the main thread

Run the Extension Manager update dialog on the main thread

Dialog::Execute() was called from the deployment worker thread, where
AquaSalInstance::Yield cannot pump the Cocoa event queue, so on macOS

Run the Extension Manager update dialog on the main thread

Dialog::Execute() was called from the deployment worker thread, where
AquaSalInstance::Yield cannot pump the Cocoa event queue, so on macOS
"Check for Updates" left the dialog unstable.

show more ...

ustring.c: gate the mixed UTF-16/ASCII NULL guards behind OSL_DEBUG_LEVEL

strtmpl.c's aImplGuardEmptyStr, which these macros reference, only exists
under OSL_DEBUG_LEVEL > 0. Without a matching guar

ustring.c: gate the mixed UTF-16/ASCII NULL guards behind OSL_DEBUG_LEVEL

strtmpl.c's aImplGuardEmptyStr, which these macros reference, only exists
under OSL_DEBUG_LEVEL > 0. Without a matching guard here, a product build
fails: "use of undeclared identifier 'aImplGuardEmptyStr'". Mirrors
strtmpl.c's own #if/#else pattern. Found rebuilding AOO42X in product
configuration after picking this same fix (f0853378f1) there.

show more ...

testtools: rebuild uno_types.rdb when bridgetest.idl changes

The registry merges in bridgetest.rdb but did not depend on it, so an
IDL edit left it describing the old interface while the headers and

testtools: rebuild uno_types.rdb when bridgetest.idl changes

The registry merges in bridgetest.rdb but did not depend on it, so an
IDL edit left it describing the old interface while the headers and
libraries described the new one, and the bridge dispatched the added
methods through stale vtable slots. It only ever came out right on a
clean build.

$? can no longer name the copy source now that there are two
prerequisites.

Found while building AOO42X on macOS/arm64, where it broke bridgetest
after the arm64 bridge commits added methods to the IDL; committed
there as 4421c0a19d. Build-tested on that branch only.

show more ...

Fix the --with-system-icu build plumbing

icuversion.mk is only delivered when building the bundled ICU, so guard
the include; with SYSTEM_ICU=YES even -include still lets make try, and
fail, to rema

Fix the --with-system-icu build plumbing

icuversion.mk is only delivered when building the bundled ICU, so guard
the include; with SYSTEM_ICU=YES even -include still lets make try, and
fail, to remake it through the generic Package rule.

Also pass icu-config --cppflags through to compiles. That until now
only worked by accident on Linux, where the ICU headers already sit in
a default search path, and not on macOS.

Found while building AOO42X on macOS/arm64; committed there as
61ceef5ddb. Build-tested on that branch only.

show more ...

libxml2: build --without-iconv, and drop the resulting -liconv linkage

No OpenOffice code calls iconv, and libxml2-configure.patch has always
meant to disable it -- but it patches the generated xmlv

libxml2: build --without-iconv, and drop the resulting -liconv linkage

No OpenOffice code calls iconv, and libxml2-configure.patch has always
meant to disable it -- but it patches the generated xmlversion.h, which
configure regenerates, so it never took effect. Windows already passed
iconv=no.

--with-static-system-libs=libiconv stays, for a system libxml2 whose own
build enabled iconv and whose xml2-config therefore reports -liconv.

show more ...

Fix rtl string qa NULL-deref crashes (test bug + source hardening) (#486)

* Fix rtl string qa NULL-deref crashes (test bug + source hardening)

Error case 1 of the test migration: ~40 *_000 cases in

Fix rtl string qa NULL-deref crashes (test bug + source hardening) (#486)

* Fix rtl string qa NULL-deref crashes (test bug + source hardening)

Error case 1 of the test migration: ~40 *_000 cases in the rtl string
qa suites passed NULL into C string functions that dereference it
(e.g. rtl_str_compare(NULL, NULL)), causing 0xC0000005 AVs. These were
dormant under the old dmake build and only surface now that the tests
actually run. NULL violates the functions' documented contract
("must be null-terminated"), so the defect was in the tests, not the
(correct, unchanged) source. Fixed both sides for defense in depth.

Tests (qa/rtl/ostring/rtl_str.cxx, rtl_string.cxx,
qa/rtl/oustring/rtl_ustr.cxx):
- Rewrote the UB NULL-deref cases as contract-respecting boundary
tests (empty string, prefix/ordering-sign < 0 / > 0), which also
closes a previously-untested coverage gap (result sign was never
asserted).
- Added real assertions to the safe NULL-with-length-0 cases, which
document the length-bounded functions' tolerance as a regression
guard.

Source (rtl/source/strtmpl.c, ustring.c):
- Added entry-point NULL guards: OSL_PRECOND (diagnoses misuse loudly
in non-product builds, compiles out in product builds) plus a
defined empty-string fallback so the library never dereferences
NULL. Guards sit at function entry, outside the per-character loops,
so string-processing throughput is unchanged.
- strtmpl.c: one edit covers both the sal_Char and sal_Unicode
instantiations. getLength is the choke point (guarding it
transitively protects hashCode, lastIndexOf*, indexOfStr, trim);
compare/compareIgnoreAsciiCase/indexOfChar/replaceChar/
toAscii{Lower,Upper}Case/valueOfChar guarded directly.
- ustring.c: guarded the 6 mixed UTF-16/ASCII comparison helpers;
length-bounded args clamp the length to 0 to avoid NULL+0 pointer
arithmetic.

BUILD.bazel: rtl_str/rtl_ustr/rtl_string removed from the
"known upstream failures" notes; they now pass.

Verified: sal3.dll rebuilds (both template instantiations) and
rtl_str / rtl_ustr / rtl_string all pass.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* use guards only in debug mode.

* improved comment by elaborate that it has been the former test and not just The former

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

show more ...

--with-static-system-libs: accept NAME:DIR, and support libiconv

iconv is only ever pulled in as a transitive dep of the XML libs and has no
--with-system-libiconv, so it has no prefix of its own --

--with-static-system-libs: accept NAME:DIR, and support libiconv

iconv is only ever pulled in as a transitive dep of the XML libs and has no
--with-system-libiconv, so it has no prefix of its own -- hence the NAME:DIR
form, e.g. --with-static-system-libs=libiconv:/usr/local. Pinning it rewrites
-liconv in the libxml/libxslt/curl link flags and in the bundled-libxml2 macOS
path in RepositoryExternal.mk and libs.mk.

The SDK ships only a .tbd stub for the iconv dylib, so there is no system
libiconv.a to fall back on; without a prefix the link stays -liconv.

show more ...

Add --with-static-system-libs, and drop -licuuc from the dmake LIBXML2LIB

Takes a comma-separated list (libxml, libxslt, curl) and rewrites each
"-lfoo" to the archive's absolute path, so a dylib si

Add --with-static-system-libs, and drop -licuuc from the dmake LIBXML2LIB

Takes a comma-separated list (libxml, libxslt, curl) and rewrites each
"-lfoo" to the archive's absolute path, so a dylib sitting beside it can't
win. Without it the choice is implicit -- and those dylibs carry absolute
install names that don't exist on an end user's machine.

LIBXML2LIB now filters -licuuc as RepositoryExternal.mk already did, so
dmake and gbuild consumers agree.

show more ...

deliver.pl: fix greedy base-name match in the macOS dylib linklib regex

\S matches dots, so the base name swallowed leading version components:
libxslt.1.1.34.dylib keyed as libxslt.1.1 rather than

deliver.pl: fix greedy base-name match in the macOS dylib linklib regex

\S matches dots, so the base name swallowed leading version components:
libxslt.1.1.34.dylib keyed as libxslt.1.1 rather than libxslt, and the
unversioned symlink was never created. Two-component names were unaffected.

show more ...

Revert "RepositoryExternal.mk, libs.mk: prefer /usr/local/lib static libs on macOS"

This reverts commit 9b92099a7350282ddfb677eafe03e37ee774e545.

RepositoryExternal.mk, libs.mk: prefer /usr/local/lib static libs on macOS

If the user already has a prebuilt libxml2.a/libxslt.a/libexslt.a/
libcurl.a/libiconv.a under /usr/local/lib, use that inst

RepositoryExternal.mk, libs.mk: prefer /usr/local/lib static libs on macOS

If the user already has a prebuilt libxml2.a/libxslt.a/libexslt.a/
libcurl.a/libiconv.a under /usr/local/lib, use that instead of what
this tree just built. Falls back to the tree's own copy when absent.

show more ...

libs.mk: link libxml2's static-lib deps for dmake modules on macOS

deliver.pl: recognize macOS libfoo.N.dylib naming in linklib:

macosx-dylib-link-list.pl: skip .framework dependencies

The @rpath/... entry for a system framework (e.g. Xcode's bundled
Python3.framework) isn't a flat libFOO.dylib, so locate() can never
find it

macosx-dylib-link-list.pl: skip .framework dependencies

The @rpath/... entry for a system framework (e.g. Xcode's bundled
Python3.framework) isn't a flat libFOO.dylib, so locate() can never
find it and the script died with "unknown". These are system
frameworks, not tree-built dylibs needing a -dylib_file fixup.

show more ...

RepositoryExternal.mk: link libxml2's static-lib private deps on macOS

A static archive doesn't record its transitive deps the way a dylib
does. libxml-2.0.pc's own "Libs.private: -lpthread -liconv

RepositoryExternal.mk: link libxml2's static-lib private deps on macOS

A static archive doesn't record its transitive deps the way a dylib
does. libxml-2.0.pc's own "Libs.private: -lpthread -liconv -lm" says
what's actually needed; add it explicitly.

show more ...

libxml2, libxslt: deliver macOS static archives to solver

curl: fix macOS bundled-openssl linker flags and dylib naming

RepositoryExternal.mk: link static libxml2/libxslt on macOS

gbuild's !SYSTEM_LIBXML/!SYSTEM_LIBXSLT branches register a PLAINLIB
expecting a dylib, but the bundled build produces a static .a on macO

RepositoryExternal.mk: link static libxml2/libxslt on macOS

gbuild's !SYSTEM_LIBXML/!SYSTEM_LIBXSLT branches register a PLAINLIB
expecting a dylib, but the bundled build produces a static .a on macOS.
Link the archive by absolute path there instead.

show more ...

libxslt: pin to bundled libxml2 and fix static xsltproc output path

The macOS static-build config (c46d036de4) didn't stop libxslt's own
configure from discovering libxml2 via pkg-config, which reso

libxslt: pin to bundled libxml2 and fix static xsltproc output path

The macOS static-build config (c46d036de4) didn't stop libxslt's own
configure from discovering libxml2 via pkg-config, which resolves to
whatever libxml2 a package manager (MacPorts, Homebrew) has registered
-- not the bundled copy. Modern libxml2 turns xmlParserMaxDepth into a
macro, breaking xsltproc.c's assignment to it. --with-libxml-prefix
pins the lookup to the bundled xml2-config instead.

Separately, --enable-shared=no makes libtool link xsltproc directly
without a .libs wrapper dir (unlike the static .a archives, which still
land in .libs), so OUT2BIN's path was wrong and the binary never made
it into the solver.

show more ...

solenv: skip .pc pkg-config files in macOS install-name fixup

libxslt: add missing stdlib.h include for getenv/strtol

Clang defaults to C23, which errors on implicit function declarations.
libxslt 1.1.34's extensions.c and libexslt/date.c call getenv/strtol
wi

libxslt: add missing stdlib.h include for getenv/strtol

Clang defaults to C23, which errors on implicit function declarations.
libxslt 1.1.34's extensions.c and libexslt/date.c call getenv/strtol
without including stdlib.h, breaking the bundled build on modern
toolchains (e.g. Xcode clang on macOS).

show more ...

vcl: fix CPUNAME token for Win64 entrypoint (X86-64 -> X86_64)