The first import through scripts/upstream-import.py (#48). It merges upstream 1.0.12 and 1.0.13.
Commits
97708c1 is the fork's base (486a384), renamed. a02f6bb merges it into main with -s ours, so later imports merge against it.
568ada3 is v1.0.13, renamed.
1d3bb54 is the merge into main.
Taken from upstream
VacationResponse import and export.
--replace-source for the DAV importers.
The executable flag on files.
IMAP \\Noselect parents keep their children in place.
JMAP import memory stays bounded.
iCalendar and vCard with bare CR line breaks are read.
Parse errors quote the line.
The EWS bare --auth-bearer fix, reading INBUXA_MIGRATE_API_KEY first.
Files.Read in the Graph sign-in scopes.
--account-name takes the data account from the principal (RFC 9670).
Kept ours: email export. Upstream's separate batching rewrite (blobpool.rs) and its tests are left out. Ours already batches and uploads in parallel, and it predicts failures in a dry run.
Changed: FileNode. Upstream went to draft-18 only (@type, mediaType). inbuxa refuses that with Invalid property for @type, so file exports into inbuxa would all have failed. Now:
Both shapes are read, and a get asks for both shapes' names.
An export probes the target once (a query sorted by type, as ihasmail does) and writes in the shape it takes.
Upstream scan
No hidden Unicode and no AI-addressed text in the added lines.
All new files are Apache-2.0 OR MIT.
Tests
cargo fmt --check and cargo clippy --all-targets -D warnings are clean.
cargo test: 1572 passed.
New tests: unit tests for both shapes, and a mock export test for each shape.
Live, on the pinned inbuxa image: all 4 pass, including a new test that sends files into inbuxa and imports them back. That test fails as expected when the draft-18 shape is forced.
Live Dovecot, Radicale, Baikal, export-dav, export-imap and export-managesieve: all pass.
After this lands
I push the upstream branch, so the watch reads the import as v1.0.13. Until then the watch fails.
Upstream main has 4 more commits after v1.0.13, including Takeout pt-BR labels and HTTP read timeouts. They come with their next release.
The first import through `scripts/upstream-import.py` (#48). It merges upstream 1.0.12 and 1.0.13.
**Commits**
- `97708c1` is the fork's base (486a384), renamed. `a02f6bb` merges it into main with `-s ours`, so later imports merge against it.
- `568ada3` is v1.0.13, renamed.
- `1d3bb54` is the merge into main.
**Taken from upstream**
- VacationResponse import and export.
- `--replace-source` for the DAV importers.
- The `executable` flag on files.
- IMAP `\\Noselect` parents keep their children in place.
- JMAP import memory stays bounded.
- iCalendar and vCard with bare CR line breaks are read.
- Parse errors quote the line.
- The EWS bare `--auth-bearer` fix, reading `INBUXA_MIGRATE_API_KEY` first.
- `Files.Read` in the Graph sign-in scopes.
- `--account-name` takes the data account from the principal (RFC 9670).
**Kept ours: email export.** Upstream's separate batching rewrite (`blobpool.rs`) and its tests are left out. Ours already batches and uploads in parallel, and it predicts failures in a dry run.
**Changed: FileNode.** Upstream went to draft-18 only (`@type`, `mediaType`). inbuxa refuses that with `Invalid property` for `@type`, so file exports into inbuxa would all have failed. Now:
- Both shapes are read, and a get asks for both shapes' names.
- An export probes the target once (a query sorted by `type`, as ihasmail does) and writes in the shape it takes.
**Upstream scan**
- No hidden Unicode and no AI-addressed text in the added lines.
- All new files are Apache-2.0 OR MIT.
**Tests**
- `cargo fmt --check` and `cargo clippy --all-targets -D warnings` are clean.
- `cargo test`: 1572 passed.
- New tests: unit tests for both shapes, and a mock export test for each shape.
- Live, on the pinned inbuxa image: all 4 pass, including a new test that sends files into inbuxa and imports them back. That test fails as expected when the draft-18 shape is forced.
- Live Dovecot, Radicale, Baikal, export-dav, export-imap and export-managesieve: all pass.
**After this lands**
- I push the `upstream` branch, so the watch reads the import as v1.0.13. Until then the watch fails.
- Upstream main has 4 more commits after v1.0.13, including Takeout pt-BR labels and HTTP read timeouts. They come with their next release.
Merges upstream's 1.0.12 and 1.0.13, renamed by scripts/upstream-import.py.
Taken: VacationResponse import and export, --replace-source for the DAV
importers, executable files, IMAP \Noselect parents, bounded JMAP import
memory, bare-CR iCalendar and vCard, parse errors that quote the line, the
EWS bare --auth-bearer fix (reading INBUXA_MIGRATE_API_KEY first, as the
rest of the tool does), Files.Read in the Graph scopes, and --account-name
taking the data account from the principal's accounts (RFC 9670).
Kept ours: email export. Upstream batched it in its own way (blobpool.rs);
ours already batches and uploads in parallel, and predicts failures in a dry
run. Its batching tests are left out with it.
Changed: upstream moved FileNode to draft-ietf-jmap-filenode-18 only, which
inbuxa refuses ("Invalid property" for @type). FileNodes are now read in
either shape, a get asks for both shapes' names, and an export asks the
target once which shape it takes (a query sorted by `type`) and writes that.
A live test sends files into the pinned inbuxa image and reads them back;
with the draft-18 shape forced it fails as above. The read-only layout check
also wants the new executable column.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
The first import through
scripts/upstream-import.py(#48). It merges upstream 1.0.12 and 1.0.13.Commits
97708c1is the fork's base (486a384), renamed.a02f6bbmerges it into main with-s ours, so later imports merge against it.568ada3is v1.0.13, renamed.1d3bb54is the merge into main.Taken from upstream
--replace-sourcefor the DAV importers.executableflag on files.\\Noselectparents keep their children in place.--auth-bearerfix, readingINBUXA_MIGRATE_API_KEYfirst.Files.Readin the Graph sign-in scopes.--account-nametakes the data account from the principal (RFC 9670).Kept ours: email export. Upstream's separate batching rewrite (
blobpool.rs) and its tests are left out. Ours already batches and uploads in parallel, and it predicts failures in a dry run.Changed: FileNode. Upstream went to draft-18 only (
@type,mediaType). inbuxa refuses that withInvalid propertyfor@type, so file exports into inbuxa would all have failed. Now:type, as ihasmail does) and writes in the shape it takes.Upstream scan
Tests
cargo fmt --checkandcargo clippy --all-targets -D warningsare clean.cargo test: 1572 passed.After this lands
upstreambranch, so the watch reads the import as v1.0.13. Until then the watch fails.486a384(v1.0.11 and later), renamed 97708c111cMerges upstream's 1.0.12 and 1.0.13, renamed by scripts/upstream-import.py. Taken: VacationResponse import and export, --replace-source for the DAV importers, executable files, IMAP \Noselect parents, bounded JMAP import memory, bare-CR iCalendar and vCard, parse errors that quote the line, the EWS bare --auth-bearer fix (reading INBUXA_MIGRATE_API_KEY first, as the rest of the tool does), Files.Read in the Graph scopes, and --account-name taking the data account from the principal's accounts (RFC 9670). Kept ours: email export. Upstream batched it in its own way (blobpool.rs); ours already batches and uploads in parallel, and predicts failures in a dry run. Its batching tests are left out with it. Changed: upstream moved FileNode to draft-ietf-jmap-filenode-18 only, which inbuxa refuses ("Invalid property" for @type). FileNodes are now read in either shape, a get asks for both shapes' names, and an export asks the target once which shape it takes (a query sorted by `type`) and writes that. A live test sends files into the pinned inbuxa image and reads them back; with the draft-18 shape forced it fails as above. The read-only layout check also wants the new executable column.