build: hard-code zrepl v0.7.0 package release 2

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
rpm: Use %config(noreplace) for the config file (#932 )
2026-02-27 08:20:12 +00:00 · 2026-02-27 08:19:16 +00:00
11 changed files with 18 additions and 677 deletions
@@ -7,21 +7,13 @@ import requests
 import time
 import os
 import yaml
 import argparse
 from pathlib import Path
 circle_token = os.environ.get('CIRCLE_TOKEN')
 if not circle_token:
-    cli_yml = Path.home() / ".circleci" / "cli.yml"
+    raise ValueError('CIRCLE_TOKEN environment variable not set')
    if cli_yml.exists():
        with open(cli_yml) as f:
            data = yaml.safe_load(f)
            if data:
                circle_token = data.get("token")
 if not circle_token:
    raise ValueError('CIRCLE_TOKEN not set and no token found in ~/.circleci/cli.yml')
 parser = argparse.ArgumentParser(description='Download artifacts from CircleCI')
 parser.add_argument('build_num', type=str, help='Build number')
@@ -1,95 +0,0 @@
 ---
 description: Generate GitHub release notes from docs/changelog.rst and create a draft release with artifacts
 argument-hint: <version-tag, e.g. v0.7.0>
 model: sonnet
 allowed-tools: Read, Write
 ---
 # Task
 Create a draft GitHub release for zrepl version $ARGUMENTS and upload release artifacts.
 ## Inputs
 Here is the current changelog RST source:
 !`cat docs/changelog.rst`
 ## Instructions
 ### Phase 1: Pre-flight Sanity Checks
 Before creating the release, verify that the previous build steps have been completed.
 Run these checks using bash commands:
 ```bash
 test -d artifacts/release && echo "✓ artifacts/release directory exists" || echo "✗ Missing artifacts/release directory - run: make download-circleci-release JOB_NUM=<num>"
 test -f artifacts/release/sha512sum.txt && echo "✓ sha512sum.txt exists" || echo "✗ Missing sha512sum.txt - run: make download-circleci-release JOB_NUM=<num>"
 test -f artifacts/release/sha512sum.txt.asc && echo "✓ sha512sum.txt.asc exists (signed)" || echo "✗ Missing signature - run: make verify-and-sign"
 ls artifacts/release/zrepl-* >/dev/null 2>&1 && echo "✓ Release artifacts present" || echo "✗ No release artifacts found"
 ```
 If ANY check fails (shows ✗), **STOP** and print the error messages. Do not proceed with release creation.
 ### Phase 2: Generate Release Notes
 2. Extract the changelog section for version $ARGUMENTS from the RST source above.
 3. Extract all GitHub usernames mentioned in that changelog section (look for @username patterns or other contributor attributions).
 4. Write a short "Highlights" section (2-4 bullets) summarizing the most impactful user-facing changes in plain language. Do NOT include commit/issue links.
 5. Write a "Contributors" thank you line for all GitHub users found in step 3. Format as: "Thanks to @user1, @user2, and @user3 for their contributions to this release!"
 6. Write a "Breaking Changes" section. Convert RST formatting to plain Markdown text (no commit or issue links):
   - `:commit:\`abc123\`` → just remove it or describe what it changed
   - `:issue:\`123\`` → just remove it or describe the issue in plain text
   - `:ref:\`display text <anchor>\`` → just use `display text` as plain text
   - `:repomasterlink:\`path\`` → just mention the path without a link
   - RST inline code ` ``code`` ` → markdown `` `code` ``
   - RST links `` `text <url>`_ `` → just use `text` without the link
   If there are no breaking changes, write "No breaking changes. vX.Y.Z-1 is interoperable with vX.Y.Z." (fill in actual versions).
 7. Assemble the full release notes using the template below. **IMPORTANT**: Do NOT include a detailed changelog list. Only include what is specified in the template.
 8. Write the result to `artifacts/release-notes.md`.
 9. Validate all links in the release notes by checking HTTP status codes with curl:
   - Extract all URLs from the markdown file
   - Check each URL with `curl -sI -w "%{http_code}" -o /dev/null <url>`
   - Report any broken links (non-200 status codes)
   - If any links are broken, stop and notify the user before creating the release
 ### Phase 3: Create Draft Release and Upload Artifacts
 10. Run: `gh release create $ARGUMENTS --title "$ARGUMENTS" --notes-file artifacts/release-notes.md --draft`
 11. Upload all artifacts: `gh release upload $ARGUMENTS artifacts/release/*`
 12. Verify upload succeeded: `gh release view $ARGUMENTS --json assets --jq '.assets | length'`
 13. Print the release URL and confirm that artifacts were uploaded successfully.
 ## Template
 ```
 The full changelog is [on the docs site](https://zrepl.github.io/changelog.html).
 ## Highlights
 {2-4 bullet plain-language summary of the most impactful changes}
 {Thank all GitHub users mentioned in the changelog, e.g., "Thanks to @user1, @user2, and @user3 for their contributions to this release!"}
 ## Breaking Changes
 {breaking changes, or "No breaking changes."}
 ## New Users
 We provide [quick-start guides](https://zrepl.github.io/quickstart.html) for different usage scenarios.
 We also recommend studying the [overview section of the configuration chapter](https://zrepl.github.io/configuration/overview.html).
 ## Testing & Upgrading
 * Read the [Changelog](https://zrepl.github.io/changelog.html)
 * [Run the platform tests](https://zrepl.github.io/usage.html#platform-tests) on a test system.
 * Download & deploy the `zrepl` binary / distro package.
 ## Donations
 zrepl is a spare-time project primarily developed by [Christian Schwarz](https://cschwarz.com).
 Express your support through a donation to keep maintenance and feature development going.
 [![Support me on Patreon](https://img.shields.io/badge/dynamic/json?color=yellow&label=Patreon&query=data.attributes.patron_count&suffix=%20patrons&url=https%3A%2F%2Fwww.patreon.com%2Fapi%2Fcampaigns%2F3095079)](https://patreon.com/zrepl) [![Donate via GitHub Sponsors](https://img.shields.io/static/v1?label=Sponsor&message=%E2%9D%A4&logo=GitHub&style=flat&color=yellow)](https://github.com/sponsors/problame) [![Donate via Liberapay](https://img.shields.io/liberapay/patrons/zrepl.svg?logo=liberapay)](https://liberapay.com/zrepl/donate) [![Donate via PayPal](https://img.shields.io/badge/donate-paypal-yellow.svg)](https://www.paypal.com/cgi-bin/webscr?cmd=_s-xclick&hosted_button_id=R5QSXJVYHGX96)
 ```
@@ -1,149 +0,0 @@
 ---
 description: Update supporters list from git history and changelog contributors
 argument-hint: <time-range-or-release>
 model: sonnet
 ---
 # Task
 Automatically update `docs/supporters.rst` by extracting GitHub contributors from:
 1. Git commit history (code contributors)
 2. `docs/changelog.rst` (all mentioned GitHub users)
 ## Arguments
 Specify a time range or release range to filter contributors:
 **Time range examples:**
 - `--since="2024-01-01"` - Contributors since a specific date
 - `--since="1 year ago"` - Contributors in the last year
 - `--since="6 months ago"` - Contributors in the last 6 months
 **Release range examples:**
 - `v0.6.0..HEAD` - Contributors between v0.6.0 and current HEAD
 - `v0.6.0..v0.7.0` - Contributors between two releases
 - `--all` - All contributors (default if no argument provided)
 ## Instructions
 ### Step 1: Extract Contributors from Git History
 Parse the arguments to determine the git log range:
 - If `--since="..."` is provided, use: `git log --since="..." --format='%aN <%aE>' | sort -u`
 - If a commit range like `v0.6.0..HEAD` is provided, use: `git log v0.6.0..HEAD --format='%aN <%aE>' | sort -u`
 - If `--all` or no argument, use: `git log --format='%aN <%aE>' | sort -u`
 Parse the output to identify:
 - GitHub users (look for @users.noreply.github.com emails or known GitHub patterns)
 - Real names of code contributors
 - Filter out the main maintainer (Christian Schwarz / problame) as they're already credited
 ### Step 2: Extract GitHub Users from Changelog
 Read `docs/changelog.rst` and extract all GitHub usernames mentioned:
 - Pattern: `` `@username <https://github.com/username>`_ ``
 - Pattern: `@username` in text
 - Collect all unique usernames
 ### Step 3: Categorize Contributors
 For each contributor, determine their tier:
 **Code contributors** (`|supporter-code|`):
 - Anyone who appears in git history as a commit author
 - Prioritize those with substantial contributions (multiple commits)
 **Other contributors** (keep existing `|supporter-gold|` and `|supporter-std|`):
 - Keep all existing gold/std supporters as-is (monetary supporters)
 - Only update code contributors
 ### Step 4: Read Current Supporters File
 Read `docs/supporters.rst` and parse existing entries to:
 - Preserve all `|supporter-gold|` entries (monetary supporters - don't touch these)
 - Preserve all `|supporter-std|` entries (monetary supporters - don't touch these)
 - Preserve ALL existing `|supporter-code|` entries in their current order
 - Extract list of existing code contributor names/GitHub usernames to avoid duplicates
 ### Step 5: Identify Contributors to Promote/Add
 **IMPORTANT: Find contributors in the range, and either ADD them (if new) or MOVE them up (if recurring).**
 For each GitHub user found in changelog or git history for the specified range:
 1. Check if they're already listed as `|supporter-code|` in the existing file
 2. Categorize them:
   - **NEW**: Not currently in the file → will be added
   - **RECURRING**: Already in the file with new contributions in this range → will be moved to top
   - **UNCHANGED**: Not in this range → stays in original position
 3. For both NEW and RECURRING contributors:
   - Get their latest contribution date from the range: `git log <range> --author="email" --format="%ai" -1`
   - Format as: `* |supporter-code| `Name <https://github.com/username>`_`
   - Use their real name if available from git history, otherwise use GitHub username
 4. Sort NEW + RECURRING contributors together by their contribution date (newest first)
 5. Keep list of UNCHANGED contributors in their original order
 **The final order will be: [NEW + RECURRING from range, sorted newest first] + [UNCHANGED in original order]**
 ### Step 6: Generate Updated File
 Create the updated supporters list:
 - Keep all `|supporter-gold|` entries unchanged and in their exact positions
 - Keep all `|supporter-std|` entries unchanged and in their exact positions
 - For `|supporter-code|` entries, reorganize as follows:
  1. **Top section**: Contributors from the analyzed range (NEW + RECURRING), sorted newest first
  2. **Bottom section**: UNCHANGED contributors (not in this range), in their original order
 - Update the spacer comment showing the range that was just processed:
  ```rst
  ..
     ↓ claude --permission-mode default /update-supporters <actual-range-used>
  ```
  Replace `<actual-range-used>` with the actual argument passed (e.g., `v0.6.1..v0.7.0`)
  This comment should appear immediately before the first code contributor entry
  If there's an existing comment in this format, replace it; otherwise add it
 - Maintain all other file structure and RST formatting, including the "Before /update-supporters" marker comment
 ### Step 7: Apply Changes
 - Use Edit tool to update the `docs/supporters.rst` file
 - Display summary of changes made:
  - Time/release range analyzed
  - Number of contributors found
  - New code contributors added
  - Final counts by tier
 ## Implementation Notes
 - **DO NOT modify, remove, or re-sort** any `|supporter-gold|` or `|supporter-std|` entries
 - **For code contributors**: ADD new ones and MOVE UP recurring ones (those with contributions in the analyzed range)
 - When extracting from git history, focus on substantial contributors (filter out single-commit or trivial changes if needed)
 - **CRITICAL:** Contributors from the analyzed range go to the top
  - Use `git log <range> --author="<email>" --format="%ai" -1` to get latest contribution date in the range
  - Both NEW and RECURRING contributors from this range are sorted together (newest first) and placed at the TOP
  - UNCHANGED contributors (not in this range) stay in their original relative order below
  - This creates a chronological history: most recent release's contributors appear first
 - Add/update the spacer comment showing the processed range
 - Handle edge cases:
  - Contributors mentioned in changelog but not in git history (external contributors, reporters)
  - Contributors in git history but not mentioned in changelog (decide whether to include)
 ## Example Output
 ```
 Analyzing contributors for range: v0.6.0..v0.7.0
 - Found 8 code contributors in git history
 - Found 12 GitHub users mentioned in changelog.rst (v0.7.0 section)
 - Current supporters.rst has 8 gold, 12 std, 10 code contributors
 New contributors:
 + Jane Doe (@janedoe)
 + John Smith (@jsmith)
 Promoted (recurring) contributors:
 ↑ Alice Developer (@alice) - new contributions in v0.6.0..v0.7.0
 Updated docs/supporters.rst:
 - Gold supporters: 8 (unchanged)
 - Std supporters: 12 (unchanged)
 - Code contributors: 10 → 12
 - Promoted to top: 3 contributors (2 new + 1 recurring)
 ```
@@ -4,17 +4,8 @@
 ARTIFACTDIR := artifacts
-ifdef ZREPL_VERSION
+_ZREPL_VERSION := v0.7.0
-    _ZREPL_VERSION := $(ZREPL_VERSION)
+ZREPL_PACKAGE_RELEASE := 2
 endif
 ifndef _ZREPL_VERSION
    _ZREPL_VERSION := $(shell git describe --always --dirty 2>/dev/null || echo "ZREPL_BUILD_INVALID_VERSION" )
    ifeq ($(_ZREPL_VERSION),ZREPL_BUILD_INVALID_VERSION) # can't use .SHELLSTATUS because Debian Stretch is still on gmake 4.1
        $(error cannot infer variable ZREPL_VERSION using git and variable is not overriden by make invocation)
    endif
 endif
 ZREPL_PACKAGE_RELEASE := 1
 GO := go
 GOOS ?= $(shell bash -c 'source <($(GO) env) && echo "$$GOOS"')
@@ -56,7 +47,7 @@ release-docker-mkcachemount:
 endif
 ##################### PRODUCING A RELEASE #############
-.PHONY: release wrapup-and-checksum check-git-clean verify-and-sign clean ensure-release-toolchain
+.PHONY: release wrapup-and-checksum check-git-clean sign clean ensure-release-toolchain
 ensure-release-toolchain:
 	# ensure the toolchain is actually the one we expect
@@ -211,8 +202,7 @@ tag-release:
 	test -n "$(ZREPL_TAG_VERSION)" || exit 1
 	git tag -u '328A6627FA98061D!' -m "$(ZREPL_TAG_VERSION)" "$(ZREPL_TAG_VERSION)"
-verify-and-sign:
+sign:
 	cd "$(ARTIFACTDIR)/release" && sha512sum -c sha512sum.txt
 	gpg -u '328A6627FA98061D!' \
 		--armor \
 		--detach-sign $(ARTIFACTDIR)/release/sha512sum.txt
@@ -223,7 +213,7 @@ clean: docs-clean
 download-circleci-release:
 	rm -rf "$(ARTIFACTDIR)"
 	mkdir -p "$(ARTIFACTDIR)/release"
-	python3 .circleci/download_artifacts.py --prefix 'artifacts/release/' "$(JOB_NUM)" "$(ARTIFACTDIR)/release"
+	python3 .circleci/download_artifacts.py --prefix 'artifacts/release/' "$(BUILD_NUM)" "$(ARTIFACTDIR)/release"
 ##################### MULTI-ARCH HELPERS #####################
@@ -93,50 +93,25 @@ The procedure to issue a release is as follows:
 * Prepare the release (as a PR to `master`):
  * Finalize `docs/changelog.rst` for the release.
  * Update the supporters list based on contributors for this release:
    ```bash
    claude --permission-mode default '/update-supporters v0.6.1..v0.7.0'
    ```
    Replace version tags with the appropriate range for your release.
  * Merge the PR. Docs are auto-published to zrepl.github.io on merge.
 * Tag the release:
  * Git tag the release on the `master` branch (e.g., `vMAJOR.MINOR.0`).
    ```
    make tag-release ZREPL_TAG_VERSION=v0.7.0
    ```
  * Push the tag.
 * Build and publish:
-  * Run the `release` pipeline against the `master` branch (trigger via CircleCI UI).
+  * Run the `release` pipeline (trigger via CircleCI UI).
-    This URL: https://app.circleci.com/pipelines/github/zrepl/zrepl?branch=master.
+  * Download artifacts: `make download-circleci-release BUILD_NUM=<circleci-build-number>`
-    Example pipeline: https://app.circleci.com/pipelines/github/zrepl/zrepl/8547
+  * Create GitHub release and upload artifacts:
  * Download artifacts using this handy makefile target.
    Note: `JOB_NUM` must be the **job number** from the `release-upload` job, **not the pipeline number**.
    Find it via: pipeline → `release` workflow → `release-upload` job number.
    Example URL: https://app.circleci.com/pipelines/github/zrepl/zrepl/8547/workflows/65feb2c9-15d7-46ab-a551-46d62a5769b0/jobs/66079/steps
    ```
    make download-circleci-release JOB_NUM=66079
    ```
  * Verify checksums and sign the checksums file:
    ```
    make verify-and-sign
    ```
  * Create GitHub draft release and upload artifacts:
    ```bash
-    claude --permission-mode default '/draft-release v0.7.0'
+    gh release create vX.Y.Z --title "vX.Y.Z" --notes "See changelog" --draft
    gh release upload vX.Y.Z artifacts/release/*
    ```
-    This command will verify that artifacts are ready, create the draft release, and upload all artifacts.
+  * Review the draft release, edit the changelog, then publish.
  * Review the draft release on GitHub, then publish,
    with the box checked to create a GitHub "Discussion" for the release.
 * Immediately after hitting publish.
  * Update the release on GitHub to link to the discussion.
  * Update `docs/changelog.rst` to link to the GitHub release.
  * Update `docs/_templates/versions.html` to link to the GitHub release.
  * Announce release in the zrepl Matrix channel & other socials as applicable.
    Link to the GitHub release as the entrypoint.
 * After a couple of days
  * Add the .rpm and .deb files to the official zrepl repos.
    * Code for management of these repos: https://github.com/zrepl/package-repo-ops (private repo at this time)
 * Update docs version list:
  * Update `docs/_templates/versions.html` with the new release.
  * Verify the link to `zrepl-noarch.tar` in the GitHub release works.
  * Merge to `master` (docs auto-publish).
 #### Patch releases, Go toolchain updates, APT/RPM Package rebuilds
@@ -7,7 +7,6 @@
  <div class="rst-other-versions">
    <dl>
      <dt>Releases</dt>
      <dd><a href="https://github.com/zrepl/zrepl/releases/tag/v0.7.0">v0.7.0</a></dd>
      <dd><a href="https://github.com/zrepl/zrepl/releases/tag/v0.6.1">v0.6.1</a></dd>
      <dd><a href="https://github.com/zrepl/zrepl/releases/tag/v0.5.0">v0.5.0</a></dd>
      <dd><a href="https://github.com/zrepl/zrepl/releases/tag/v0.4.0">v0.4.0</a></dd>
@@ -16,8 +16,6 @@ Changelog
 0.7.0
 -----
 `GitHub Release <https://github.com/zrepl/zrepl/releases/tag/v0.7.0>`_
 * |feature| Config file inclusion using ``include`` directive.
  This allows distributing zrepl job definitions across multiple YAML files in a ``conf.d`` style directory.
  (:commit:`4d6583e`, thanks, `@ZeyadTamimi <https://github.com/zeyadtamimi>`_).
@@ -28,15 +26,10 @@ Changelog
 * |bugfix| Detect duplicate and internal job names at config parse time.
  zrepl will now refuse to load a config with duplicate or internally reserved job names, instead of failing later at daemon startup.
  (:commit:`860a9be`).
 * |bugfix| Fix ``last_n`` keep rule behavior (:commit:`ebc46cf`, thanks, `@dsh2dsh <https://github.com/dsh2dsh>`_).
 * |bugfix| Allow different jobs to use the same ``root_fs`` for sinks (:commit:`27012e5`, thanks, `@dsh2dsh <https://github.com/dsh2dsh>`_).
 * |docs| Add installation instructions for openSUSE (:commit:`40c4827`, thanks, `@findesgh <https://github.com/findesgh>`_).
 * |docs| Improve documentation on ``send_properties`` configuration and NFS/SMB considerations (:commit:`b5d8538`, thanks, `@Malvineous <https://github.com/Malvineous>`_).
 * |docs| Warn more prominently about the risks of the ``not_replicated`` keep rule (:commit:`affe00a`, thanks, `@wxiaoguang <https://github.com/wxiaoguang>`_).
 * |docs| Improve TLS/EasyRSA setup instructions (:commit:`e524b60`, thanks, `@alorimer <https://github.com/alorimer>`_).
 * |docs| Remove dead Bountysource link (:commit:`440b074`, thanks, `@lpulley <https://github.com/lpulley>`_).
 * |docs| Fix missing newline in compile-from-source documentation (:commit:`e2fcf9f`, thanks, `@Raupinger <https://github.com/Raupinger>`_).
 * |docs| Fix apt repository installation snippet (:commit:`8305367`, thanks, `@fermino <https://github.com/fermino>`_).
 * |docs| ``zrepl.github.io``: auto-publish from ``master`` branch, retire ``stable`` branch (:commit:`4f950bb`).
 * |maint| ``zrepl status``: switch from the unmaintained ``cview`` fork back to the actively maintained ``tview`` library (:commit:`d7ede3f`).
 * |maint| Update to Go 1.25 toolchain with Go 1.24 language level & Go dependencies
@@ -32,30 +32,6 @@ We would like to thank the following people and organizations for supporting zre
   <div class="fa fa-code" style="width: 1em;"></div>
 ..
   The list below is (roughly) sorted by date of latest contribution, newest first.
  ..
     ↓ post v0.7.0
 * |supporter-std| `drbawb <https://git.sr.ht/~hime>`_
  ..
     ↓ claude --permission-mode default /update-supporters v0.6.1..v0.7.0
 * |supporter-code| `Bakhtiyar Neyman <https://github.com/bakhtiyarneyman>`_
 * |supporter-code| `Zeyad Tamimi <https://github.com/ZeyadTamimi>`_
 * |supporter-code| `Orsiris de Jong <https://github.com/deajan>`_
 * |supporter-code| `Martin Glatzle <https://github.com/findesgh>`_
 * |supporter-code| `Andrew Lorimer <https://github.com/alorimer>`_
 * |supporter-code| `Adam Nielsen <https://github.com/Malvineous>`_
 * |supporter-code| `wxiaoguang <https://github.com/wxiaoguang>`_
 * |supporter-code| `Logan Pulley <https://github.com/lpulley>`_
 * |supporter-code| `Florian <https://github.com/Raupinger>`_
 * |supporter-code| `Fermín Olaiz <https://github.com/fermino>`_
 * |supporter-code| `Denis Shaposhnikov <https://github.com/dsh2dsh>`_
   ..
     ↓ Before /update-supporters
 * |supporter-gold| `Hostsharing eG – die Hosting-Genossenschaft <https://www.hostsharing.net/>`_
 * |supporter-std| `Max Christian Pohle <https://coderonline.de>`_
 * |supporter-gold| Prominic.NET, Inc.
@@ -635,25 +635,6 @@ func (f subroot) UserSpecifiedDatasets() zfs.UserSpecifiedDatasetsSet {
 	}
 }
 // senderPrefixDatasetPath returns the configured literal sender-side prefix
 // (e.g. "hdd_9XG1J9J1") to strip from incoming filesystem paths, or nil if
 // ZREPL_STRIP_SENDER_PREFIX is unset/empty. Using a literal string (rather
 // than a component count) lets us losslessly reconstruct the sender's
 // original path later in ListFilesystems, which is required for the
 // sender/receiver filesystem-matching in the replication planner to work.
 func senderPrefixDatasetPath() *zfs.DatasetPath {
 	s := envconst.String("ZREPL_STRIP_SENDER_PREFIX", "")
 	if s == "" {
 		return nil
 	}
 	dp, err := zfs.NewDatasetPath(s)
 	if err != nil || dp.Length() == 0 {
 		return nil
 	}
 	return dp
 }
 func (f subroot) MapToLocal(fs string) (*zfs.DatasetPath, error) {
 	p, err := zfs.NewDatasetPath(fs)
 	if err != nil {
@@ -662,17 +643,11 @@ func (f subroot) MapToLocal(fs string) (*zfs.DatasetPath, error) {
 	if p.Length() == 0 {
 		return nil, errors.Errorf("cannot map empty filesystem")
 	}
 	// PATCH: strip a configured literal prefix (e.g. the source zpool name)
 	// from the *sender's* path before appending it under root_fs.
 	if prefix := senderPrefixDatasetPath(); prefix != nil && p.HasPrefix(prefix) {
 		p.TrimPrefix(prefix)
 	}
 	c := f.localRoot.Copy()
 	c.Extend(p)
 	return c, nil
 }
 func (s *Receiver) ListFilesystems(ctx context.Context, req *pdu.ListFilesystemReq) (*pdu.ListFilesystemRes, error) {
 	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
@@ -713,28 +688,12 @@ func (s *Receiver) ListFilesystems(ctx context.Context, req *pdu.ListFilesystemR
 		a.TrimPrefix(root)
 		// PATCH: re-prepend the literal sender prefix that MapToLocal strips,
 		// so that the Path we report here matches what the sender reports
 		// for the same filesystem (see replication_logic.go's Path-based
 		// sender/receiver matching). Without this, the planner can never
 		// find a matching receiverFS and will treat every cycle as if the
 		// filesystem doesn't exist yet on the receiver.
 		reportPath := a.ToString()
 		if prefix := senderPrefixDatasetPath(); prefix != nil {
 			if reportPath == "" {
 				reportPath = prefix.ToString()
 			} else {
 				reportPath = prefix.ToString() + "/" + reportPath
 			}
 		}
 		fs := &pdu.Filesystem{
-			Path:          reportPath,
+			Path:          a.ToString(),
 			IsPlaceholder: ph.IsPlaceholder,
 			ResumeToken:   token,
 		}
 		fss = append(fss, fs)
 	}
 	if len(fss) == 0 {
 		getLogger(ctx).Debug("no filesystems found")
@@ -1144,4 +1103,4 @@ func doDestroySnapshots(ctx context.Context, lp *zfs.DatasetPath, snaps []*pdu.F
 	return &pdu.DestroySnapshotsRes{
 		Results: ress,
 	}, nil
-}
+}
@@ -1,177 +0,0 @@
 https://claude.ai/chat/f688c411-b11e-4455-918b-fde5e6577430
 Diagnosed replication snapshot duplication caused by version listing failures
 Diagnosed replication snapshot duplication caused by version listing failures
 Отличная зацепка — снапшот chunk_9XG1J8T0/nomad/ha-zah@zrepl_20260724_054520_000 реально существует (0B allocated, 195M refer — то есть данные там есть), но планировщик каждый цикл продолжает считать, что на приёмнике версий нет вообще. Значит проблема не в наличии данных, а в том, что ListFilesystemVersions (то, чем сверяется планировщик) их почему-то не видит. Погружусь в код листинга снапшотов на ресивере.
 Ran 3 commands, viewed 3 files
 Ran 3 commands, viewed 3 files
 Кажется, нашёл настоящий баг — и это очень похоже на правду. Проверю до конца, как формируется список receiverFS/IsPlaceholder, которым пользуется планировщик (не тот же код, что читает placeholder при самом receive).
 Viewed 5 files, ran 4 commands
 Viewed 5 files, ran 4 commands
 Нашёл настоящую причину, и она серьёзнее, чем казалось — это баг именно от моего патча, а не от zrepl самого по себе. Дело в сопоставлении списков sender/receiver в планировщике.
 Проверяю, есть ли envconst.String для строкового значения
 Проверяю, есть ли envconst.String для строкового значения
 Вот и реальный баг — и он от патча
 Планировщик (replication_logic.go:302-306) сопоставляет sender и receiver filesystem'ы строковым сравнением путей:
 go
 for _, rfs := range rfss {
    if rfs.Path == fs.Path {   // fs.Path — путь СЕНДЕРА, напр. "hdd_9XG1J9J1/nomad/ha-zah"
        receiverFS = rfs
    }
 }
 А rfs.Path формируется в Receiver.ListFilesystems (endpoint.go:695) как a.TrimPrefix(root) — локальный путь chunk_9XG1J8T0/nomad/ha-zah минус root_fs = nomad/ha-zah.
 До патча это совпадало с sender-путём один-в-один (root_fs + sender_path → обрезаем root_fs → получаем обратно sender_path). А теперь, когда MapToLocal откусывает ещё и hdd_9XG1J9J1, обратная обрезка даёts nomad/ha-zah — а сендер сообщает hdd_9XG1J9J1/nomad/ha-zah. Строки не совпадают → receiverFS всегда nil → планировщик каждый раз считает, что на приёмнике версий нет → пытается full receive → ловит "destination exists". Именно поэтому цикл повторяется бесконечно.
 Чиню — добавляю обратное восстановление отрезанного префикса именно в отчёт ListFilesystems, чтобы сопоставление путей снова работало:
 Architected shared helper function to consolidate prefix-handling logic
 Architected shared helper function to consolidate prefix-handling logic
 Есть всё нужное. Переписываю патч целиком на литеральный строковый префикс — это устраняет и проблему с расчётом пути, и делает обратное сопоставление в ListFilesystems возможным:
 Reconciled edge case behavior and prepared filesystem prefix restoration
 Reconciled edge case behavior and prepared filesystem prefix restoration
 Смотрю текущее состояние MapToLocal перед переработкой
 Переключаю strip с "N компонентов" на литеральный строковый префикс, чтобы его можно было симметрично восстановить в ListFilesystems для сопоставления путей с sender'ом
 endpoint.go
 +35
 -18
 Reconciled edge case behavior and prepared filesystem prefix restoration.
 Done
 Теперь патчу ListFilesystems, чтобы при формировании отчётного пути (используемого для сверки с sender) обрезанный префикс восстанавливался обратно:
 Восстанавливаю отрезанный sender-префикс в пути, который ListFilesystems возвращает планировщику — иначе receiver Path не совпадает с sender Path и планировщик считает receiver пустым на каждом цикле
 Восстанавливаю отрезанный sender-префикс в пути, который ListFilesystems возвращает планировщику — иначе receiver Path не совпадает с sender Path и планировщик считает receiver пустым на каждом цикле
 Проверяю синтаксис и то, что старая переменная ZREPL_STRIP_SENDER_PREFIX_COMPS больше нигде не используется:
 Проверяю синтаксис и итоговое использование переменных окружения
 Проверяю синтаксис и итоговое использование переменных окружения
 Синтаксис чистый, старой переменной нигде не осталось. Собираю (насколько позволяет toolchain в песочнице):
 Ran 2 commands
 Ran 2 commands
 Отлично, всё корректно, включая граничный случай равных путей. Патч готов. Итог:
 Что было не так
 Планировщик сопоставляет filesystem'ы sender'а и receiver'а строковым равенством путей (replication_logic.go:303: rfs.Path == fs.Path). До патча receiver.Path (локальный путь минус root_fs) всегда совпадал с sender.Path один в один. Мой первый патч ломал этот инвариант: обрезав ещё и hdd_9XG1J9J1 при записи, но не восстанавливая его при отчёте — receiver стал сообщать "nomad/ha-zah", а sender — "hdd_9XG1J9J1/nomad/ha-zah". Строки никогда не совпадали → receiverFS всегда nil → планировщик каждый цикл считал, что на приёмнике ничего нет → пытался полный recv в уже занятый датасет → ошибка. Бесконечный цикл был не багом zrepl, а прямым следствием несимметричности моего же патча.
 Что изменилось
 Заменил обрезку "N компонентов" на обрезку литерального строкового префикса — это делает операцию обратимой.
 В MapToLocal префикс отрезается при вычислении локального пути хранения.
 В ListFilesystems тот же префикс приклеивается обратно к пути, который сообщается планировщику — так путь снова совпадает с тем, что говорит sender, и сопоставление работает.
 Как применить
 Пересоберите бинарник, затем поменяйте имя переменной в systemd-юните (старое ZREPL_STRIP_SENDER_PREFIX_COMPS=1 → новое, с явным значением префикса):
 ini
 Environment=ZREPL_STRIP_SENDER_PREFIX=hdd_9XG1J9J1
 bash
 systemctl daemon-reload
 systemctl restart zrepl_rcv
 Датасеты, скорее всего, снова придётся снести (zfs destroy -r chunk_9XG1J8T0/nomad) и дать реплицироваться с нуля, раз в прошлый раз всё было в неконсистентном состоянии из-за бага. После рестарта проверьте:
 bash
 journalctl -u zrepl_rcv -n 100 --no-pager | grep -iE "conflict|receive failed|error"
 Ключевой признак, что теперь всё работает правильно: строка "conflict automatically resolved" с пустым sorted receiver versions должна появиться только один раз (при первой инициализации датасета), а не повторяться на каждом цикле после успешного приёма.
 а где патч?
 Показываю итоговый diff патча
 Показываю итоговый diff патча
 дай исправленные функции
 Показываю итоговый код исправленных функций целиком
 Показываю итоговый код исправленных функций целиком
 Вот три исправленные функции целиком (файл internal/endpoint/endpoint.go):
 1. Новая вспомогательная функция (строки 638–654):
 go
 // senderPrefixDatasetPath returns the configured literal sender-side prefix
 // (e.g. "hdd_9XG1J9J1") to strip from incoming filesystem paths, or nil if
 // ZREPL_STRIP_SENDER_PREFIX is unset/empty. Using a literal string (rather
 // than a component count) lets us losslessly reconstruct the sender's
 // original path later in ListFilesystems, which is required for the
 // sender/receiver filesystem-matching in the replication planner to work.
 func senderPrefixDatasetPath() *zfs.DatasetPath {
 	s := envconst.String("ZREPL_STRIP_SENDER_PREFIX", "")
 	if s == "" {
 		return nil
 	}
 	dp, err := zfs.NewDatasetPath(s)
 	if err != nil || dp.Length() == 0 {
 		return nil
 	}
 	return dp
 }
 2. MapToLocal (строки 656–672):
 go
 func (f subroot) MapToLocal(fs string) (*zfs.DatasetPath, error) {
 	p, err := zfs.NewDatasetPath(fs)
 	if err != nil {
 		return nil, err
 	}
 	if p.Length() == 0 {
 		return nil, errors.Errorf("cannot map empty filesystem")
 	}
 	// PATCH: strip a configured literal prefix (e.g. the source zpool name)
 	// from the *sender's* path before appending it under root_fs.
 	if prefix := senderPrefixDatasetPath(); prefix != nil && p.HasPrefix(prefix) {
 		p.TrimPrefix(prefix)
 	}
 	c := f.localRoot.Copy()
 	c.Extend(p)
 	return c, nil
 }
 3. Кусок ListFilesystems, где формируется Path (строки 712–734, остальная часть функции без изменений):
 go
 		a.TrimPrefix(root)
 		// PATCH: re-prepend the literal sender prefix that MapToLocal strips,
 		// so that the Path we report here matches what the sender reports
 		// for the same filesystem (see replication_logic.go's Path-based
 		// sender/receiver matching). Without this, the planner can never
 		// find a matching receiverFS and will treat every cycle as if the
 		// filesystem doesn't exist yet on the receiver.
 		reportPath := a.ToString()
 		if prefix := senderPrefixDatasetPath(); prefix != nil {
 			if reportPath == "" {
 				reportPath = prefix.ToString()
 			} else {
 				reportPath = prefix.ToString() + "/" + reportPath
 			}
 		}
 		fs := &pdu.Filesystem{
 			Path:          reportPath,
 			IsPlaceholder: ph.IsPlaceholder,
 			ResumeToken:   token,
 		}
 		fss = append(fss, fs)
 Больше в файле ничего не менялось — MapToLocal вызывается ещё и из Receive/ListFilesystemVersions/DestroySnapshots, но там аргументом всегда приходит исходный путь сендера (req.Filesystem), так что дополнительных правок там не требуется.
@@ -1,122 +0,0 @@
 #!/usr/bin/env bash
 #
 # zrepl_last_snapshot_age.sh
 #
 # Показывает, сколько минут назад был создан последний снапшот
 # (по умолчанию с префиксом "zrepl_") на резервной (приёмной) стороне.
 #
 # Использование:
 #   ./zrepl_last_snapshot_age.sh <датасет> [<датасет2> ...]
 #   ./zrepl_last_snapshot_age.sh -r <датасет>            # рекурсивно по всем дочерним
 #   ./zrepl_last_snapshot_age.sh -p zrepl_ -r chunk_9XG1J8T0/nomad
 #
 # Опции:
 #   -r            рекурсивно проверять все дочерние датасеты
 #   -p PREFIX     фильтровать снапшоты по префиксу имени (по умолчанию "zrepl_")
 #   -w MINUTES    задать порог в минутах: если снапшот старше — exit code 1
 #                 (удобно для мониторинга / cron)
 #   -q            тихий режим: не печатать построчный вывод, только итог
 #
 # Коды возврата:
 #   0 — всё ок (снапшот найден, и если задан -w, он не старше порога)
 #   1 — снапшот найден, но старше порога -w
 #   2 — снапшот не найден вообще (для датасета/выборки)
 #   3 — ошибка использования / zfs недоступен
 set -euo pipefail
 PREFIX="zrepl_"
 RECURSIVE=0
 WARN_MINUTES=""
 QUIET=0
 usage() {
    grep '^#' "$0" | sed -e '1d' -e 's/^# \{0,1\}//'
    exit 3
 }
 while getopts ":rp:w:qh" opt; do
    case "$opt" in
        r) RECURSIVE=1 ;;
        p) PREFIX="$OPTARG" ;;
        w) WARN_MINUTES="$OPTARG" ;;
        q) QUIET=1 ;;
        h) usage ;;
        *) usage ;;
    esac
 done
 shift $((OPTIND - 1))
 if [[ $# -lt 1 ]]; then
    echo "Ошибка: укажите хотя бы один датасет." >&2
    usage
 fi
 if ! command -v zfs >/dev/null 2>&1; then
    echo "Ошибка: команда zfs не найдена." >&2
    exit 3
 fi
 now_epoch=$(date +%s)
 overall_newest_epoch=0
 overall_newest_name=""
 any_found=0
 for dataset in "$@"; do
    zfs_opts=(-H -p -o name,creation -t snapshot -s creation)
    if [[ $RECURSIVE -eq 1 ]]; then
        zfs_opts+=(-r)
    else
        zfs_opts+=(-d 1)
    fi
    # -H  без заголовков, -p значения в "машинном" виде (creation -> epoch seconds)
    mapfile -t lines < <(zfs list "${zfs_opts[@]}" "$dataset" 2>/dev/null | grep "@${PREFIX}" || true)
    if [[ ${#lines[@]} -eq 0 ]]; then
        if [[ $QUIET -eq 0 ]]; then
            echo "[$dataset] снапшотов с префиксом '${PREFIX}' не найдено"
        fi
        continue
    fi
    # последняя строка — самый свежий снапшот (сортировка -s creation)
    last_line="${lines[-1]}"
    snap_name="${last_line%%$'\t'*}"
    snap_epoch="${last_line##*$'\t'}"
    age_seconds=$(( now_epoch - snap_epoch ))
    age_minutes=$(( age_seconds / 60 ))
    any_found=1
    if (( snap_epoch > overall_newest_epoch )); then
        overall_newest_epoch=$snap_epoch
        overall_newest_name=$snap_name
    fi
    if [[ $QUIET -eq 0 ]]; then
        printf "[%s] последний снапшот: %s (%d мин. назад)\n" \
            "$dataset" "$snap_name" "$age_minutes"
    fi
 done
 if [[ $any_found -eq 0 ]]; then
    echo "Ни одного снапшота с префиксом '${PREFIX}' не найдено ни в одном из указанных датасетов." >&2
    exit 2
 fi
 overall_age_minutes=$(( (now_epoch - overall_newest_epoch) / 60 ))
 echo "---"
 printf "Самый свежий снапшот в выборке: %s (%d мин. назад)\n" \
    "$overall_newest_name" "$overall_age_minutes"
 if [[ -n "$WARN_MINUTES" ]]; then
    if (( overall_age_minutes > WARN_MINUTES )); then
        echo "ВНИМАНИЕ: последний снапшот старше порога ${WARN_MINUTES} мин." >&2
        exit 1
    fi
 fi
 exit 0