Hi Redhuan — we reproduced the BIMCompiler + bim-ootb baseline from clean checkouts
(master 55900e2d, bim-ootb main 18959f23, Ubuntu 24.04, Java 17.0.19 / Maven 3.8.7 /
Node 20) and collected evidence. The browser viewer and the Op-Log kernel POCs work
beautifully from a fresh clone — SampleHouse renders, a 2.3 MB IFC imports end-to-end
(218 elements → validated SQLite), poc_showstopper and poc_volume both PASS.
A few things did not reproduce from a clean checkout. Reporting them in case they're
useful — we know master lags your working branches and that fast CI defers the Java gate.
1. git-lfs is a hard prerequisite but isn't in the README Quick Start. All *.db are
LFS-tracked; without git-lfs the pipeline fails with
sqlite3.DatabaseError: file is not a database on the pointer files. The error doesn't
point at LFS, so it costs a while to diagnose.
2. run_RosettaStones.sh classify_sh.yaml fails at populate on a clean master clone:
[SQLITE_ERROR] no such table: M_Product. The shipped library/component_library.db
(LFS oid 7e5df6e1…, identical on master and fable/meshdb-livewire) contains only
ad_* tables, while library/schema_snapshot_component.sql at the same commit defines
M_Product. It looks like the migration / self-heal loader for the compiler-side complib
isn't wired on master. Downstream effect: no *_BOM.db fixtures are produced, so
run_tests.sh gave us 157 PASS / 85 RED / 3 SKIP against the documented 382 / 94 / 1
(DAGCompiler ran 1 test).
3. Module builds assume a seeded ~/.m2. The script's mvn -pl … calls need
mvn install -N plus one full mvn install -DskipTests first, or
com.bim:bim-compiler-parent / bim-eyes don't resolve. One README line would cover it.
4. ERP oracle witnesses aren't runnable from a clean checkout. diff_oracle.js needs
build/erp/ad_full.db and $HOME/bim-ootb/viewer/manifest.json (that path no longer
exists in current bim-ootb); poc_fold_complete.js needs build/erp/glassbowl_data.db.
Both DBs appear to exist only in your local dockerized iDempiere environment — we swept
the public OCI buckets and found no hits.
5. Two asset paths in the deploy tree don't resolve.
- The default building in
deploy/dev references buildings/Duplex_meta.db, which is
absent → empty first paint. An explicit ?db= works fine.
deploy/landing2.html loads sandbox/locale_loader.js, sandbox/YT.png and
sandbox/submit.png. None of the three exist in deploy/sandbox/ — all three exist
only in deploy/dev/. (sandbox/import_db_builder.js and sandbox/Sysnova.png do
resolve, the latter from SYSNOVA/.) Looks like the landing page moved between
directories and the relative paths weren't updated.
Happy to share full logs. Thanks for publishing this — the browser-native direction is
genuinely impressive to test against.
Hi Redhuan — we reproduced the BIMCompiler + bim-ootb baseline from clean checkouts
(master
55900e2d, bim-ootb main18959f23, Ubuntu 24.04, Java 17.0.19 / Maven 3.8.7 /Node 20) and collected evidence. The browser viewer and the Op-Log kernel POCs work
beautifully from a fresh clone — SampleHouse renders, a 2.3 MB IFC imports end-to-end
(218 elements → validated SQLite),
poc_showstopperandpoc_volumeboth PASS.A few things did not reproduce from a clean checkout. Reporting them in case they're
useful — we know master lags your working branches and that fast CI defers the Java gate.
1. git-lfs is a hard prerequisite but isn't in the README Quick Start. All
*.dbareLFS-tracked; without git-lfs the pipeline fails with
sqlite3.DatabaseError: file is not a databaseon the pointer files. The error doesn'tpoint at LFS, so it costs a while to diagnose.
2.
run_RosettaStones.sh classify_sh.yamlfails at populate on a clean master clone:[SQLITE_ERROR] no such table: M_Product. The shippedlibrary/component_library.db(LFS oid
7e5df6e1…, identical on master andfable/meshdb-livewire) contains onlyad_*tables, whilelibrary/schema_snapshot_component.sqlat the same commit definesM_Product. It looks like the migration / self-heal loader for the compiler-side complibisn't wired on master. Downstream effect: no
*_BOM.dbfixtures are produced, sorun_tests.shgave us 157 PASS / 85 RED / 3 SKIP against the documented 382 / 94 / 1(DAGCompiler ran 1 test).
3. Module builds assume a seeded
~/.m2. The script'smvn -pl …calls needmvn install -Nplus one fullmvn install -DskipTestsfirst, orcom.bim:bim-compiler-parent/bim-eyesdon't resolve. One README line would cover it.4. ERP oracle witnesses aren't runnable from a clean checkout.
diff_oracle.jsneedsbuild/erp/ad_full.dband$HOME/bim-ootb/viewer/manifest.json(that path no longerexists in current bim-ootb);
poc_fold_complete.jsneedsbuild/erp/glassbowl_data.db.Both DBs appear to exist only in your local dockerized iDempiere environment — we swept
the public OCI buckets and found no hits.
5. Two asset paths in the deploy tree don't resolve.
deploy/devreferencesbuildings/Duplex_meta.db, which isabsent → empty first paint. An explicit
?db=works fine.deploy/landing2.htmlloadssandbox/locale_loader.js,sandbox/YT.pngandsandbox/submit.png. None of the three exist indeploy/sandbox/— all three existonly in
deploy/dev/. (sandbox/import_db_builder.jsandsandbox/Sysnova.pngdoresolve, the latter from
SYSNOVA/.) Looks like the landing page moved betweendirectories and the relative paths weren't updated.
Happy to share full logs. Thanks for publishing this — the browser-native direction is
genuinely impressive to test against.