The token lookup only read PATs from tea's config — with tea's OAuth
login the config has no token, so the script silently continued with an
empty one and failed 30 minutes later at the upload step. Fall back to
the git credential store (the access token stored by every git push,
accepted by Gitea for API auth), and error out immediately when nothing
is found instead of burning a full build.
Replace the static hero screenshot with a self-contained React mockup of the
full Gridline DB viewer (sidebar, table tree, tab bar, fake SQL editor,
data grid, queries/objects/schema-visualizer views, disabled Tools state).
- Add @astrojs/react and mount the island via client:visible
- 16:9 aspect-ratio mockup, hidden on mobile (static image fallback)
- JSON cell popover: solid surface, anchored within the mockup window,
Escape/outside-click close
- Dependency-free HTML5 drag-to-reorder tabs
- Theme-aware app tokens (canvas, surface-raised, popover bg)
- Site-wide max-w-7xl (was max-w-[1200px]) so sections center on big screens
- Regenerate standalone www/bun.lock for deterministic Dokploy builds
Hide entry-animated elements at first paint (html.anime-ready guard in
global.css) so the landing page no longer flashes visible content before
anime.js loads. Adds a <noscript> fallback and a JS load-error fallback to
guarantee content is never permanently hidden, and reveals the hero
immediately for prefers-reduced-motion users. Includes #macos-hint in the
hero entrance timeline so the macOS command box animates in correctly.
www/package.json declares vitest ^4.1.10 but the standalone www/bun.lock
(Railpack/Dokploy builds from /www with --frozen-lockfile) was stale and
missing it, failing every deployment with 'lockfile had changes, but
lockfile is frozen'. Regenerated www/bun.lock from www/package.json.
Add 'Copy connection URL' (with password, fetched from the OS keychain on
demand) and 'Copy connection URL (no password)' to the connection card kebab
menu. Builds postgresql://, mysql://, sqlite://, redis:// strings with
percent-encoded credentials and the PG sslmode appended. Bump to v0.7.10.
- Add 2-feature_request.md and 3-improvement.md templates (chooser now
lists Bug report, Feature request, Improvement in order)
- Improve bug report template: cross-template pointer, Linux log hint,
Related issues/PRs section, demo-DB reproduction hint
- Auto-assign every newly opened issue to the repo owner:
assignees in template front matter + a workflows/auto-assign.yml
(covers issues opened without a template)
- Add ISSUE_TEMPLATE/config.yml with Roadmap/README contact links
- Update README Contributing section
The bundle contains read-only files (e.g. the bundled OpenSSL dylibs), so a
sudo-less 'xattr -dr com.apple.quarantine' fails with Permission denied for
some users. Standardize on 'sudo xattr -dr com.apple.quarantine' in the
README, ROADMAP, the release.yml SIGNING STATUS comment, and the release
notes; AGENTS.md now requires release notes to carry the macOS first-launch
instructions until a Developer ID cert is secured.
Users at the top table previously saw no mention of the unsigned-app
warnings; the details lived ~200 lines down under Getting Started. Add a
callout right under the Download table linking to the macOS first-launch
instructions (right-click Open / xattr workaround).
* fix: macOS Finder launch (store path, ad-hoc signing, vendored openssl)
- Store::open was resolving 'gridline.db' relative to the working directory;
Finder/LaunchServices launches run with cwd=/ so the .expect() panicked
(exit 101, silent 'app does nothing') before Tauri ever started. Open the
store under app.path().app_data_dir() in setup instead (same dir as the
demo DB), creating the dir when needed.
- Add bundle.macOS.signingIdentity "-" + hardenedRuntime false so the
bundler ad-hoc signs the whole bundle. Previously only the inner binary
got Xcode 16's linker-signed signature, which macOS treats as unsigned:
quarantined downloads showed 'damaged and can't be opened', and after
quarantine removal LaunchServices silently refused to spawn it.
- ssh2 now builds OpenSSL vendored (feature vendored-openssl): the release
binary previously carried an absolute LC_LOAD_DYLIB to the build machine's
/opt/homebrew/opt/openssl@3/lib, which dyld aborted on (and hardened
runtime library-validation rejected even when present).
- README: document the macOS first-launch paths (right-click Open / xattr
for the damaged-error case, re-run after every upgrade).
- release.yml: update SIGNING STATUS comment to reflect ad-hoc signing.
* chore: bump 0.7.8 -> 0.7.9 (release prep)
- Version sync across package.json, src-tauri/Cargo.toml, src-tauri/tauri.conf.json
- version.test.ts / bundle-config.test.ts / docs-coverage.test.ts expect 0.7.9
- README: both download tables -> v0.7.9 asset names; new v0.7.9 changelog entry;
tag instructions -> v0.7.9; MAINTENANCE comment updated
- AGENTS.md maintenance note example -> v0.7.9
- ROADMAP: Shipped (0.7.9) section; Next up retitled (0.8.0) TBD
The releaseAssetNamePattern input was never honored by tauri-action —
the draft release got default versioned asset names (Gridline_0.7.5_*.dmg,
Gridline-0.7.5-1.x86_64.rpm, ...), so the version-free releases/latest
links in the README would 404. Per user decision, drop the dynamic
scheme: both README download tables now link statically to the v0.7.5
assets, with a MAINTENANCE comment and an AGENTS.md checklist entry
reminding to bump both tables on each release. Remove the dead
releaseAssetNamePattern config from release.yml.
Three independent pg-tools step failures on the 5th release run:
- Linux: 'ar: libpgcommon_srv.a: file format not recognized' — parallel
`make -C` tool builds race each other's recursion into src/common
(rm -f / ar crs on the same archive). Pre-build generated-headers +
libpq (which pulls in common/port) serially so the -j tool builds find
the archives up to date.
- macOS x86_64: strchrnul marked 'introduced in macOS 15.4' by Xcode 16.4
fails -Werror=unguarded-availability-new and would crash at runtime on
macOS < 15.4. Force ac_cv_func_strchrnul=no and rename PG16's inline
fallback (pg_strchrnul) so it compiles on any SDK without touching the
system symbol.
- Windows: EDB pg_dump.exe needs libcrypto/libssl/libiconv/libintl/
libwinpthread/liblz4/libzstd/ICU DLLs; copy every bin/*.dll, not just
libpq.dll.
All three verified locally (macOS: full build + tools run from resource
dir; YAML + bash -n for all platform variants).
The per-platform binary-suffix check used
f="$OUT/${b}$( [ platform = windows-latest ] && echo .exe )".
The [ ] test exits 1 when false (linux/macos), and under 'set -euo
pipefail' the failing command substitution aborts the whole step before
any check output — exit 1 right after checksums with no error message.
Replace with an if/else BIN_EXT variable set before the loops.
Verified locally: full script block (checksums -> file check -> sanity)
passes with real pg_dump/pg_restore/psql + libpq, exit 0.
- patchelf was passed '\$ORIGIN' (backslash-dollar inside single quotes),
so the ELF rpath was a literal \$ORIGIN path — the loader never expanded
it and the tools couldn't find libpq at runtime. Use '$ORIGIN'.
- macOS awk program had \$1, which awk rejects as a syntax error; use $1
(verified locally: otool -> install_name_tool -> all 3 tools run).
cp ran from /tmp/postgresql-16.4 after 'cd', so the relative
src-tauri/resources/pg_tools target didn't exist. Use
${GITHUB_WORKSPACE}/src-tauri/resources/pg_tools so copies land in the repo.
CI (release.yml) — pg-tools bundle step failed on Linux/macOS with
'catalog/pg_*_d.h not found': building src/bin/pg_dump directly raced its
generated-headers prerequisite under -j. Fix: run 'make -C src/backend
generated-headers' serially first. Also drop the unrecognized --disable-shared
flag (PG16 client tools link shared libpq) and bundle libpq alongside the
tools — @loader_path rewrite via install_name_tool on macOS, $ORIGIN rpath
via patchelf on Linux — plus a 'tools run' sanity check so a broken bundle
fails the step before release. Windows (EDB download + libpq.dll) already
passed. Verified locally: all 3 tools run from the bundled dir.
Cmd+K palette — selecting a result did nothing visible:
- non-table results set selectedObjectType but never switched the view (the
Objects page only mounts when DbViewerScreen's local currentView ===
'objects'), so nothing happened. Added a store 'requestedView' field that
DbViewerScreen consumes and clears; the palette now requests the right view
('objects' for functions/triggers/sequences/enums/etc., 'db-viewer' for
tables/views/matviews before openTab) so the tab or Objects list actually
appears.
- Added keyboard navigation: ↑/↓ move the highlight (wrapping), Enter picks,
Esc closes. scrollIntoView guarded with optional call so jsdom tests don't
crash on the ref callback.
- Tests: palette select now asserts requestedView; new ArrowDown/Enter,
ArrowUp-wrap, empty-Enter cases; store test for requestedView setter.
adrianbonpin