Build WASM with the latest stable ViennaRNA 2.x

.github/workflows/build-wasm.yml

@@ -31,9 +31,18 @@ jobs:
         with:
           python-version: '3.12'
       - name: Test dependency download integrity and cache recovery
-        run: python scripts/wasm/test-downloads.py
+        run: |
+          python scripts/wasm/test-downloads.py
+          python scripts/wasm/test-vienna-release.py
       - name: Test input, worker lifecycle and CSV handling
         run: npm test --prefix webinterface
+      - name: Resolve newest stable ViennaRNA 2.x release and checksum
+        env:
+          GITHUB_TOKEN: ${{ github.token }}
+        run: |
+          python scripts/wasm/vienna-release.py --latest > "$RUNNER_TEMP/vienna-release.json"
+          cat "$RUNNER_TEMP/vienna-release.json"
+          echo "WASM_VIENNA_RELEASE=$RUNNER_TEMP/vienna-release.json" >> "$GITHUB_ENV"
       - name: Cross-compile dependencies and IntaRNA
         run: bash scripts/wasm/build.sh
       - name: Check real WASM predictions against the native reference

ChangeLog

@@ -84,6 +84,10 @@
 
 ## Technical changes and Optimizations
 
+- build WASM with ViennaRNA 2.7.2 locally and the newest stable 2.x in CI;
+  verify published archive checksums, isolate versioned caches and retain
+  selected dependency provenance and notices (PR #292 review)
+
 - BUGFIX : restore WASM dependency downloads using the pinned official ViennaRNA
   release asset; verify downloads before caching and recover invalid caches
   (PR #291)
@@ -249,6 +253,27 @@ energies, restricted partition sums, and trackers.
 ################################################################################  
 ################################################################################
 
+261022 Alexander Mitrofanov
+ * scripts/wasm/{vienna-release.py,vienna-release.json}, build.sh
+   + Resolve the newest stable ViennaRNA >=2.7.2 and <3 from all official GitHub
+     release pages in CI; require its release tarball and published SHA-256.
+   * Pin 2.7.2 for reproducible local builds and accept saved release manifests.
+     Isolate RNAlib sources/installations and IntaRNA objects by archive identity
+     and keep Boost/zlib in a separate prefix to prevent reuse or include-path
+     shadowing of libraries/headers from another release (PR #292).
+ * scripts/wasm/patch-vienna.py
+   - Remove the dlib 19.24 compatibility patch; ViennaRNA 2.7.2 bundles dlib 20.0.
+ * scripts/wasm/{write-build-info.mjs,stage.sh,smoke-protocol.mjs,test-browser.py}
+   * Record the selected archive URL/checksum, check the compiled ViennaRNA
+     version, stage matching dependency notices and test actual version exports.
+ * scripts/wasm/test-vienna-release.py, .github/workflows/build-wasm.yml
+   + Cover numeric version selection, pagination, major-3/prerelease exclusion,
+     missing/invalid asset metadata, API failures, manifest validation and
+     rejection of compiled engines with a stale ViennaRNA version.
+ * webinterface/docs/step1-wasm-compilation.md
+   * Document latest-2.x CI selection, pinned local reproduction, versioned
+     caches, current bundled dlib and archive provenance.
+
 261022 Alexander Mitrofanov
  * scripts/wasm/{build,download}.sh
    * Use the official GitHub ViennaRNA 2.7.0 release asset and pin its SHA-256;

scripts/wasm/build.sh

@@ -5,9 +5,11 @@ repo=$(cd "$(dirname "$0")/../.." && pwd)
 build=${WASM_BUILD_DIR:-"$repo/build-wasm"}
 mkdir -p "$build"
 build=$(cd "$build" && pwd)
-prefix="$build/deps"
+# Keep legacy RNAlib headers out of Boost's -I path: ordinary -I searches take
+# precedence over the -isystem ViennaRNA paths produced by IntaRNA's configure.
+prefix="$build/deps/common"
 jobs=${JOBS:-2}
-for tool in emcc em++ emar emranlib emconfigure emmake curl pkg-config node; do
+for tool in emcc em++ emar emranlib emconfigure emmake curl pkg-config node python3; do
   command -v "$tool" >/dev/null || { echo "Missing tool: $tool" >&2; exit 1; }
 done
 mkdir -p "$build/downloads" "$build/sources" "$prefix" "$build/site/wasm"
@@ -18,16 +20,27 @@ fetch https://zlib.net/fossils/zlib-1.3.1.tar.gz zlib-1.3.1.tar.gz \
   9a93b2b7dfdac77ceba5a558a580e74667dd6fede4585b91eefb60f03b72df23
 fetch https://archives.boost.io/release/1.87.0/source/boost_1_87_0.tar.bz2 boost_1_87_0.tar.bz2 \
   af57be25cb4c4f4b413ed692fe378affb4352ea50fbe294a11ef548f4d527d89
-fetch https://github.com/ViennaRNA/ViennaRNA/releases/download/v2.7.0/ViennaRNA-2.7.0.tar.gz ViennaRNA-2.7.0.tar.gz \
-  e2addf33e7c0dd4fff31d8e9a46fda8e9efcde0d199f789f20eb0a13b7f7135d
+# CI supplies the newest stable 2.x manifest; local builds use the checked-in pin.
+read -r vienna_version vienna_url vienna_sha256 < <(
+  python3 "$repo/scripts/wasm/vienna-release.py" --fields \
+    --manifest "${WASM_VIENNA_RELEASE:-$repo/scripts/wasm/vienna-release.json}"
+)
+echo "Building ViennaRNA $vienna_version (SHA-256 $vienna_sha256)"
+fetch "$vienna_url" "ViennaRNA-$vienna_version.tar.gz" "$vienna_sha256"
+# Separate source, installed RNAlib and IntaRNA objects by archive identity so
+# changing the release cannot reuse headers, libraries or objects from an old one.
+vienna_id="vienna-$vienna_version-$vienna_sha256"
+vienna_prefix="$build/deps/$vienna_id"
+vienna_sources="$build/sources/$vienna_id"
+intarna_build="$build/intarna-$vienna_id"
 
 # Keep pkg-config searches inside the WASM prefix, including configure probes.
-export PKG_CONFIG_LIBDIR="$prefix/lib/pkgconfig"
+export PKG_CONFIG_LIBDIR="$vienna_prefix/lib/pkgconfig:$prefix/lib/pkgconfig"
 export PKG_CONFIG_PATH="$PKG_CONFIG_LIBDIR"
 export CFLAGS='-O3 -fexceptions'
 export CXXFLAGS='-O3 -fexceptions'
-export CPPFLAGS="-I$prefix/include"
-export LDFLAGS="-L$prefix/lib -fexceptions"
+export CPPFLAGS="-I$vienna_prefix/include -I$prefix/include"
+export LDFLAGS="-L$vienna_prefix/lib -L$prefix/lib -fexceptions"
 if [[ ! -f "$prefix/lib/libz.a" ]]; then
   tar -xzf "$build/downloads/zlib-1.3.1.tar.gz" -C "$build/sources"
   (cd "$build/sources/zlib-1.3.1"
@@ -47,27 +60,26 @@ if [[ ! -f "$prefix/.boost-1.87.0-js-exceptions-complete" ]]; then
      -sZLIB_LIBPATH="$prefix/lib" --prefix="$prefix" install)
   touch "$prefix/.boost-1.87.0-js-exceptions-complete"
 fi
-if [[ ! -f "$prefix/lib/libRNA.a" ]]; then
-  tar -xzf "$build/downloads/ViennaRNA-2.7.0.tar.gz" -C "$build/sources"
-  (cd "$build/sources/ViennaRNA-2.7.0"
-   CPPFLAGS="$CPPFLAGS -I$PWD/src" emconfigure ./configure --host=wasm32-unknown-emscripten --prefix="$prefix" \
+if [[ ! -f "$vienna_prefix/.complete" ]]; then
+  mkdir -p "$vienna_sources"
+  tar -xzf "$build/downloads/ViennaRNA-$vienna_version.tar.gz" -C "$vienna_sources"
+  (cd "$vienna_sources/ViennaRNA-$vienna_version"
+   CPPFLAGS="$CPPFLAGS -I$PWD/src" emconfigure ./configure --host=wasm32-unknown-emscripten --prefix="$vienna_prefix" \
      --disable-shared --enable-static --disable-openmp --disable-pthreads \
      --disable-lto --disable-simd --disable-vectorize --disable-mpfr \
      --without-svm --without-gsl --without-swig --without-perl --without-python \
      --without-doc --without-cla --without-check --without-kinfold \
      --without-forester --without-rnalocmin --without-rnaxplorer
-   # Backport dlib's upstream char32_t typedef for modern libc++ char_traits.
-   # Only the downloaded dependency is patched; RNAlib algorithms are unchanged.
-   python3 "$repo/scripts/wasm/patch-vienna.py" "$PWD"
    emmake make -C src/ViennaRNA -j "$jobs"
    emmake make -C src/ViennaRNA install
    emmake make install-pkgconfigDATA)
+  touch "$vienna_prefix/.complete"
 fi
 
 # Legacy Autoconf probes declare exit()/inflateEnd() with the wrong signatures,
 # which wasm-ld rejects. Verify real dependency APIs before caching those probes.
-em++ -std=c++23 -O2 -fexceptions -I"$prefix/include" \
-  "$repo/scripts/wasm/dependency-probe.cpp" -L"$prefix/lib" \
+em++ -std=c++23 -O2 -fexceptions -I"$vienna_prefix/include" -I"$prefix/include" \
+  "$repo/scripts/wasm/dependency-probe.cpp" -L"$vienna_prefix/lib" -L"$prefix/lib" \
   -lboost_filesystem -lboost_system -lboost_program_options -lboost_regex \
   -lboost_iostreams -lboost_serialization -lRNA -lz -o "$build/dependency-probe.js"
 node "$build/dependency-probe.js"
@@ -78,15 +90,15 @@ export ac_cv_lib_boost_regex_exit=yes
 export ac_cv_lib_boost_iostreams_exit=yes
 export ac_cv_lib_z_inflateEnd=yes
 # RNAlib's .pc embeds a native-only -Wl,-fno-lto; this is a static WASM build.
-export VRNA_LIBS="-L$prefix/lib -lRNA -lm"
+export VRNA_LIBS="-L$vienna_prefix/lib -lRNA -lm"
 
 # Use the native Autotools source lists and CLI entry point, with a JS output.
 (cd "$repo" && bash autotools-init.sh)
-mkdir -p "$build/intarna"
-(cd "$build/intarna"
- CXXFLAGS="$CXXFLAGS -Werror=inconsistent-missing-override" CPPFLAGS="-I$build/intarna/src -I$repo/src $CPPFLAGS" emconfigure "$repo/configure" --host=wasm32-unknown-emscripten \
+mkdir -p "$intarna_build"
+(cd "$intarna_build"
+ CXXFLAGS="$CXXFLAGS -Werror=inconsistent-missing-override" CPPFLAGS="-I$intarna_build/src -I$repo/src $CPPFLAGS" emconfigure "$repo/configure" --host=wasm32-unknown-emscripten \
    --disable-multithreading --disable-log-coloring --with-mdspan=kokkos \
-   --with-vrna="$prefix" --with-boost="$prefix" \
+   --with-vrna="$vienna_prefix" --with-boost="$prefix" \
    --with-boost-libdir="$prefix/lib" --with-zlib="$prefix"
  emmake make -C src -j "$jobs" EXEEXT=.js \
    LDFLAGS="$LDFLAGS -sMODULARIZE=1 -sEXPORT_ES6=1 -sEXPORT_NAME=createIntaRNA -sENVIRONMENT=web,worker,node -sINVOKE_RUN=0 -sEXIT_RUNTIME=1 -sEXPORTED_RUNTIME_METHODS=FS,callMain -sALLOW_MEMORY_GROWTH=1 -sMAXIMUM_MEMORY=1073741824 -sSTACK_SIZE=8388608"
@@ -96,5 +108,6 @@ mkdir -p "$build/intarna"
 printf '{"type":"module"}\n' > "$build/site/wasm/package.json"
 node "$repo/scripts/wasm/write-build-info.mjs" "$repo" "$build/site" \
   "$(emcc -dumpversion)" "$(git -C "$repo" rev-parse HEAD)" \
-  "$(git -C "$repo" status --porcelain --untracked-files=no)"
+  "$(git -C "$repo" status --porcelain --untracked-files=no)" \
+  "$vienna_version" "$vienna_url" "$vienna_sha256"
 echo "WASM assets: $build/site/wasm"

scripts/wasm/smoke-protocol.mjs

@@ -1,5 +1,6 @@
 // Integration coverage of browser argument translation against the real CLI.
 import assert from 'node:assert/strict';
+import { readFileSync } from 'node:fs';
 import { resolve } from 'node:path';
 import { pathToFileURL } from 'node:url';
 import { DEFAULTS, EVALUATION_EXAMPLE, prepareRequest, makeArguments } from '../../webinterface/protocol.js';
@@ -8,6 +9,7 @@ import { parseResults, projectCsv } from '../../webinterface/results.js';
 import { archiveFiles } from '../../webinterface/archive.js';
 const asset = resolve(process.argv[2] || 'build-wasm/site/wasm/intarna.js');
 const { default: createIntaRNA } = await import(pathToFileURL(asset));
+const compiledVersion = readFileSync(new URL('../build-info.txt', pathToFileURL(asset)), 'utf8');
 async function run(input, format = 'C') {
   const request = prepareRequest({ ...DEFAULTS, ...input });
   const diagnostics = [];
@@ -54,7 +56,7 @@ async function replayArchive(input) {
   const outputs = fs.readdir('/output').filter((name) => name !== '.' && name !== '..').map((name) => ({ name, text: fs.readFile(`/output/${name}`, { encoding: 'utf8' }) }));
   const reports = [['normal', 'N'], ['detailed', 'D'], ['ensemble', 'E']].filter(([key]) => request.outputs.includes(key));
   for (const [key, format] of reports) outputs.push({ name: `intarna-${key}.txt`, text: (await run(input, format)).text });
-  const files = archiveFiles(request, { csv: original.text, files: outputs }, 'IntaRNA 3.4.1\nViennaRNA 2.7.0\n');
+  const files = archiveFiles(request, { csv: original.text, files: outputs }, compiledVersion);
   for (const config of ['parameters', ...reports.map(([key]) => key)]) {
     const diagnostics = [];
     const module = await createIntaRNA({ noInitialRun: true, locateFile: () => new URL('./intarna.wasm', pathToFileURL(asset)).href, print: (line) => diagnostics.push(line), printErr: (line) => diagnostics.push(line) });

scripts/wasm/stage.sh

@@ -6,6 +6,24 @@ build=${WASM_BUILD_DIR:-"$repo/build-wasm"}
 site="$build/site"
 test -s "$site/wasm/intarna.js"
 test -s "$site/wasm/intarna.wasm"
+# Select notices from the archive used by these assets, including dlib upgrades.
+read -r vienna_version vienna_sha256 dlib_directory < <(python3 - "$site" "$build" <<'PYSOURCE'
+import json, re, sys
+from pathlib import Path
+site, build = map(Path, sys.argv[1:])
+metadata = json.loads((site / "build-info.json").read_text())
+archive = metadata["viennarnaArchive"]
+version, digest = archive["version"], archive["sha256"]
+if version != metadata["viennarna"] or not re.fullmatch(r"2\.\d+\.\d+", version) or not re.fullmatch(r"[0-9a-f]{64}", digest):
+    raise SystemExit("Invalid compiled ViennaRNA archive identity")
+source = build / "sources" / f"vienna-{version}-{digest}" / f"ViennaRNA-{version}"
+dlib = list((source / "src").glob("dlib-*/dlib/LICENSE.txt"))
+if len(dlib) != 1 or not re.fullmatch(r"dlib-[0-9.]+", dlib[0].parent.parent.name):
+    raise SystemExit("Expected one bundled dlib license in the compiled ViennaRNA source")
+print(version, digest, dlib[0].parent.parent.name)
+PYSOURCE
+)
+vienna_source="$build/sources/vienna-$vienna_version-$vienna_sha256/ViennaRNA-$vienna_version"
 mkdir -p "$site/licenses"
 cp -R "$repo/webinterface/assets" "$site/"
 for file in index.html styles.css app.js mark.svg protocol.js runner.js results.js options.js columns.js archive.js visualization.js; do
@@ -18,20 +36,20 @@ cp "$repo/src/easylogging++.LICENCE.txt" "$site/licenses/easylogging.txt"
 cp "$repo/src/mdspan/LICENSE" "$site/licenses/mdspan.txt"
 cp "$build/sources/boost_1_87_0/LICENSE_1_0.txt" "$site/licenses/Boost.txt"
 cp "$build/sources/zlib-1.3.1/zlib.h" "$site/licenses/zlib.h"
-cp "$build/sources/ViennaRNA-2.7.0/COPYING" "$site/licenses/ViennaRNA.txt"
-cp "$build/sources/ViennaRNA-2.7.0/src/dlib-19.24/dlib/LICENSE.txt" "$site/licenses/dlib.txt"
-cp "$build/sources/ViennaRNA-2.7.0/src/ViennaRNA/plotting/naview/naview.c" "$site/licenses/naview.c"
+cp "$vienna_source/COPYING" "$site/licenses/ViennaRNA.txt"
+cp "$vienna_source/src/$dlib_directory/dlib/LICENSE.txt" "$site/licenses/dlib.txt"
+cp "$vienna_source/src/ViennaRNA/plotting/naview/naview.c" "$site/licenses/naview.c"
 emscripten_dir=$(dirname "$(command -v emcc)")
 cp "$emscripten_dir/LICENSE" "$site/licenses/Emscripten.txt"
 cp "$emscripten_dir/system/lib/libc/musl/COPYRIGHT" "$site/licenses/musl.txt"
 for library in libcxx libcxxabi libunwind compiler-rt; do
   cp "$emscripten_dir/system/lib/$library/LICENSE.TXT" "$site/licenses/$library.txt"
 done
-cat > "$site/licenses/index.html" <<'HTML'
+cat > "$site/licenses/index.html" <<HTML
 <!doctype html><html lang="en"><meta charset="utf-8"><meta name="viewport" content="width=device-width"><title>IntaRNA licenses</title>
 <h1>IntaRNA and third-party notices</h1>
 <p>IntaRNA: Bioinformatics Group of Prof. Backofen, University of Freiburg.<br>ViennaRNA: authors and Institute for Theoretical Chemistry, University of Vienna.</p>
-<p>This distribution uses ViennaRNA 2.7.0, Boost 1.87.0, zlib 1.3.1, bundled dlib 19.24 (with Unicode/template compatibility fixes), Kokkos mdspan, Easylogging++ and the Emscripten 4.0.23 runtime.</p>
+<p>This distribution uses ViennaRNA $vienna_version, Boost 1.87.0, zlib 1.3.1, bundled dlib ${dlib_directory#dlib-}, Kokkos mdspan, Easylogging++ and the Emscripten 4.0.23 runtime.</p>
 <ul>
 <li><a href="IntaRNA.txt">IntaRNA</a></li><li><a href="ViennaRNA.txt">ViennaRNA</a></li>
 <li><a href="Boost.txt">Boost</a></li><li><a href="zlib.h">zlib</a></li>

scripts/wasm/test-browser.py

@@ -2,7 +2,8 @@
 import argparse
 from pathlib import Path
 from io import BytesIO
-from urllib.parse import parse_qs, urlparse
+import re
+from urllib.parse import parse_qs, urljoin, urlparse
 from zipfile import ZipFile
 from playwright.sync_api import sync_playwright, expect
 
@@ -53,7 +54,11 @@ def predict():
         expect(status).to_contain_text('complete:', timeout=120000)
         expect(page.locator('#input-error')).not_to_be_visible()
 
-    expect(page.locator('#version-info')).to_contain_text('ViennaRNA 2.7.0')
+    metadata = page.request.get(urljoin(args.url, 'build-info.json')).json()
+    vienna_version = metadata['viennarna']
+    assert re.fullmatch(r'2\.\d+\.\d+', vienna_version)
+    assert metadata['viennarnaArchive']['version'] == vienna_version
+    expect(page.locator('#version-info')).to_contain_text(f'ViennaRNA {vienna_version}')
     for logo in page.locator('.site-header img').all():
         assert logo.evaluate('node => node.complete && node.naturalWidth > 0')
     for key in ['query-options', 'target-options', 'seed-options', 'helix-options', 'shape-options', 'interaction-options', 'output-options', 'general-options', 'sequence-input']:
@@ -122,7 +127,7 @@ def predict():
     assert archive['intarna-interactions.csv'] == csv
     assert archive['query.fa'] == '>query\nGGGAGGAGGG\n'
     assert '# IntaRNA 3.4.1' in archive['parameters.cfg']
-    assert 'Vienna RNA package 2.7.0' in archive['parameters.cfg']
+    assert f'Vienna RNA package {vienna_version}' in archive['parameters.cfg']
     assert 'outNumber = 10' in archive['parameters.cfg']
     assert 'IntaRNA --parameterFile=parameters.cfg' in archive['parameters.cfg']
     assert 'temperature =' not in archive['parameters.cfg']