This is an automated email from the ASF dual-hosted git repository. leginee pushed a commit to branch bazel-migration in repository https://gitbox.apache.org/repos/asf/openoffice.git
commit aa95bf4bb4ea59834c8ecc3cec471baaac0b1871 Author: Peter Kovacs <[email protected]> AuthorDate: Sun Aug 2 04:50:15 2026 +0200 docs: record BACKPORT #3 (Win64 type-correctness follow-ups) Four trunk cherry-picks, green on both arches. Also records what was deliberately NOT ported from the same batch: two build commits where the Bazel side is already ahead (version.lib, /ENTRY:LibMain — trunk's @16 looks wrong for MSVC x64), the ADO/ADABAS packcomponents fixes that our label-based component map already gets right, and the gbuild ports that have no Bazel equivalent. Notes the 861c48608f DataAccess *.xcu move as future drift against main/postprocess/BUILD.bazel. Co-Authored-By: Claude Opus 5 <[email protected]> --- CLAUDE.md | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+) diff --git a/CLAUDE.md b/CLAUDE.md index 7b9f7c5d39..ba0552af62 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -259,6 +259,41 @@ comments). Root causes per defect: main/filter/source/graphicfilter/idxf/Readme still-open cases (99892 Phase B encoding-chooser UI, text-dense thumbnail) live on the local branch bugfix-dxf-filter-open, deliberately out of this tree. +BACKPORT #3 LANDED 2026-08-02 (Damjan's Win64 type-correctness follow-ups, 4 commits, +GREEN on BOTH arches per user build). These refine hunks ALREADY in this branch — +two of them (dllmgr, sfx2 Get10ThSec) are trunk rewriting OUR OWN upstreamed edits, so +skipping them would have left those files permanently divergent and re-flagged by every +`git cherry` run. None fixes a live x64 defect (uInt32→uLong is a widening, IMR_* are +small constants, Windows clock_t is 32-bit long on both arches) — the value is drift +removal. Source-only, 5 files: vcl salgdi.cxx ImplPreparePolyDraw nCurrPoint(s) +sal_uLong→sal_uInt32 (48f8cc8f69, completes 30f18d3fb8), vcl salframe.cxx drop the +static_cast<int>(wParam) on IMR_* (1e5d8e72b5), basic dllmgr.cxx guard +!defined(_WIN64)→defined(INTEL) (14a9b8c1d9 — equivalent here, the toolchain injects +INTEL/X86_64 globally, see build/toolchain/windows_cc_toolchain_config.bzl), sfx2 +Get10ThSec() sal_uInt32→clock_t + SfxStatusIndicator::_nStartTime long→clock_t +(448dc86220; progress.cxx already includes <time.h> at line 54, no new include needed). + NOT PORTED from the same trunk batch, because the Bazel side is ALREADY AHEAD: + 29ca9883b9 (link `version` for VerQueryValueW on x64) — vcl/BUILD.bazel already has + version.lib; 25912db745 (Win64 vcl entrypoint) — we already select /ENTRY:LibMain on + x64, and trunk's `/ENTRY:LibMain@16` looks WRONG: MSVC x64 drops the stdcall @N + decoration, so the symbol from vcl/win/source/app/salshl.cxx is undecorated LibMain + (that is what links and boots here). Report upstream. + ALSO NOT PORTED: f116fde0f6 + the packregistry half of 70996efe05 (ADO component path + `drivers/ado`→`source/drivers/ado`, ADABAS platform condition) — both already correct + in main/postprocess/BUILD.bazel, which resolves components by TARGET LABEL so the + malformed path cannot exist in our form. 70996efe05's BFunctions.cxx half is a + UNX/MACOSX-only refactor (WNT branch untouched) = no-op here. The gbuild ports in + that batch (connectivity 07bf101f83/5db65c3a88, chart2 89e8f0e82d/f1c404e081/ + 93fbba6999 + 0a9d806e68 version-map workaround, officecfg ef56053619/323ac1e9a6/ + 9e90d1d104, solenv Configuration.mk 431aafefb5/4dc45a5a3e, Ant jar sealing 861c48608f) + have NO Bazel equivalent — Bazel replaces that layer. + DRIFT TO WATCH (not actionable yet): 861c48608f also MOVES the DataAccess *.xcu out of + connectivity into officecfg registry/data/org/openoffice/Office/DataAccess/ and adds + *.xcs schemas for them; main/postprocess/BUILD.bazel still references the old + //main/connectivity:source/drivers/*/*.xcu paths and would break on a future sync. + d50e6300ed (drop -Xbootclasspath, unsupported on Java > 8) is inert now but is a real + constraint for scripting/java when the Java bucket lands. + DO NOT PORT (VC9 island landmines — see the trunk drift audit above): boost 1.84 (2e60be5056), pyuno→Python 3 (27dee00ef8), C++11 floor (7ce5b5df31), autoconf 2.72, the macOS/Apple-Silicon batch, GTK/iODBC/lld/icc-Clang16.
