appveyor.yml

#
# Network UPS Tools: AppVeyor CI build recipe: NUT for Windows with MSYS2/MinGW x64
#
# Copyright (C) 2022-2026	Jim Klimov <jimklimov+nut@gmail.com>
#
# Inspired by docs:
# https://www.msys2.org/docs/ci/
# https://www.appveyor.com/docs/appveyor-yml/
# https://www.appveyor.com/docs/build-configuration/
# https://www.appveyor.com/docs/windows-images-software/

version: 2.8.5.{build}-{branch}

# base image
image: Visual Studio 2022

# branches to build
branches:
  # whitelist
  only:
    - master
    - FTY
    - fightwarn
    - /Windows/

platform: x64

# https://www.appveyor.com/docs/build-cache/
environment:
  # Note: APPVEYOR_SAVE_CACHE_ON_ERROR and APPVEYOR_CACHE_SKIP_SAVE are set
  # in the init section below, depending on the branch and PR status.
  # Store cache entries without (re-)compression: the data is mostly already
  # compressed (ccache objects, pacman *.zst packages), and single-threaded
  # LZMA at -mx=6 took 1-2 minutes per run for the ccache entry alone.
  # The cache limit for PR builds is 1GB total across all entries, so
  # ccache max_size is capped below (see install steps) to leave a margin
  # for the other entries (pacman packages ~100MB, nut-win-deps, nut-ci):
  APPVEYOR_CACHE_ENTRY_ZIP_ARGS: -t7z -mx=0
  CCACHE_DIR: /home/appveyor/.ccache
  CCACHE_MAXSIZE: 750MiB
  DO_CLEAN_NUTCI_CACHE: no
  DO_USE_NUTCI_CACHE: yes
  MSYS: "winsymlinks:nativestrict"

# Customize GH notification message template
# TODO: for repetitive commits in a PR consider a clean-up GHA like
#  https://github.com/marketplace/actions/delete-pr-comments-during-subsequent-ci
notifications:
  - provider: GitHubPullRequest
    on_build_success: true
    on_build_failure: true
    on_build_status_changed: false
    template: |
      {{#passed}}:white_check_mark:{{/passed}}{{#failed}}:x:{{/failed}} [Build {{&projectName}} {{buildVersion}} {{status}}]({{buildUrl}}) (commit {{commitUrl}} by @{{&commitAuthorUsername}})
      * [artifacts{{#failed}} (if any){{/failed}}]({{buildurl}}/artifacts)
      {{#passed}}
        {{#jobs}}
          {{#messages}}
      * {{message}}
          {{/messages}}
        {{/jobs}}
      {{/passed}}

# https://github.com/networkupstools/nut/blob/Windows-v2.8.0-1/docs/config-prereqs.txt#L951
# or look for the chapter in nearby lines in later (current) revisions.
# Note: not using `time` in scripts currently - they did upset
# AppVeyor console log scanner with a /^sys.*/ match (apparently)
# AppVeyor kills the job at 60 minutes, without saving the build cache:
# a cold-ccache build then never gets warm, and an interrupted cache save
# can even lose the old entry. So we set a deadline at 57 minutes after
# the job start, and our build/test/bundle steps fail (via `timeout`) if
# they would run past it. With APPVEYOR_SAVE_CACHE_ON_ERROR this leaves
# a few minutes for the 7z step, artifact upload and saving the cache
# (about 1.5 minutes in total as of 2026-10: 7z ~45s, cache save ~40s).
# PR builds do not save the cache (see below), so they get no deadline
# and can use the whole hour.
# Each bash step below computes its time left like this (no CMD special
# characters, so it passes through CMD intact):
#   R=$(( ${CI_DEADLINE_EPOCH:-EPOCHSECONDS+3600} - EPOCHSECONDS ))
init:
  - ps: |
        if (-not $env:APPVEYOR_PULL_REQUEST_NUMBER) { Set-AppveyorBuildVariable -Name CI_DEADLINE_EPOCH -Value ([DateTimeOffset]::UtcNow.ToUnixTimeSeconds() + 57 * 60) }
  # There is one build cache per project, shared by all branches and PRs,
  # and each save replaces the previous entries. So PR builds only restore
  # the cache (warmed up by master builds), but do not save it: otherwise
  # each PR build pushes objects of master (and other PRs) out of the
  # limited cache space, and spends time to upload it all again. This is
  # only a convenience, since a PR can change this file: to really keep
  # PR builds from altering the cache used by master builds, untick "Save
  # build cache in Pull Request builds" in the AppVeyor project settings.
  # Builds of master also save the cache when they fail (e.g. by the
  # deadline above), so that a cold cache gets warm over several builds;
  # other branch builds only save it when they succeed. Note that for PR
  # builds APPVEYOR_REPO_BRANCH is the target branch.
  - ps: |
        if ($env:APPVEYOR_PULL_REQUEST_NUMBER) { $env:APPVEYOR_CACHE_SKIP_SAVE = "true" } elseif ($env:APPVEYOR_REPO_BRANCH -eq "master") { $env:APPVEYOR_SAVE_CACHE_ON_ERROR = "true" }

# The steps below are kept in scripts named after each section, under
# .github/scripts/appveyor/ (.bat for `cmd`, .ps1 for `ps` steps),
# numbered by the position of their section among the AppVeyor
# pipeline script phases (01-init, 02-install, 03-before_build,
# 04-build_script, 05-after_build, 06-before_test, 07-test_script,
# 08-after_test, ...), see https://www.appveyor.com/docs/build-configuration/#build-pipeline
# Note that `init` above runs before the repository is cloned, so it
# has to stay inline.
# NOTE for the .bat scripts: cmd.exe only treats double quotes as quoting,
# not the single quotes around `bash -lc '...'` arguments, so any `|`, `||`,
# `&`, `&&`, `<` or `>` outside of double quotes there is interpreted by
# cmd itself (e.g. `a || b` runs `b` only as a cmd command if bash failed,
# and silently drops everything after it otherwise). Use `if ! a ; then
# b ; fi` style constructs (or double-quote such characters) instead.
install:
  - cmd: call "%APPVEYOR_BUILD_FOLDER%\.github\scripts\appveyor\02-install.bat"

before_build:
  - cmd: call "%APPVEYOR_BUILD_FOLDER%\.github\scripts\appveyor\03-before_build.bat"

build_script:
  - cmd: call "%APPVEYOR_BUILD_FOLDER%\.github\scripts\appveyor\04-build_script.bat"

after_build:
  - cmd: call "%APPVEYOR_BUILD_FOLDER%\.github\scripts\appveyor\05-after_build.bat"

test_script:
  - cmd: call "%APPVEYOR_BUILD_FOLDER%\.github\scripts\appveyor\07-test_script.bat"

after_test:
  - cmd: call "%APPVEYOR_BUILD_FOLDER%\.github\scripts\appveyor\08-after_test.bat"
  - ps: '& "$env:APPVEYOR_BUILD_FOLDER\.github\scripts\appveyor\08-after_test.ps1"'

# Note: AppVeyor saves the build cache after on_success scripts (and for
# failed master builds right after the failing step, before on_failure), see
# https://www.appveyor.com/docs/build-configuration/#build-pipeline
on_success:
  - ps: '& "$env:APPVEYOR_BUILD_FOLDER\.github\scripts\appveyor\12-on_success.ps1"'

artifacts:
  - path: 'NUT-for-Windows*.7z'
    name: Bundle of binary files and FOSS dependencies of NUT for Windows

  - path: config.log
    name: config.log of recent build of NUT for Windows

  - path: config.nut_report_feature.log
    name: config.nut_report_feature.log of recent build of NUT for Windows

# Example optional cache (depends on file change):
# - C:\msys64 -> appveyor.yml
cache:
  - C:\projects\nut\config.cache
  - C:\projects\.cache\nut-ci
  - C:\msys64\home\appveyor\nut-win-deps
  - C:\msys64\home\appveyor\.cache\nut-ci
  - C:\msys64\home\appveyor\.ccache
  - C:\msys64\home\appveyor\ccache # likely missing, no problem - but the name is reported in ccache status
  - C:\Users\appveyor\AppData\Local\ccache # may be default in newer versions
  - C:\Users\appveyor\AppData\Local\.ccache # may be missing, but for completeness like above
  - C:\msys64\var\cache\pacman\pkg

# Below we tried to stash binaries of MSYS2 environment
# so VM deployment is faster on subsequent builds
# (update/install "from scratch" costs about 3 min),
# but unstashing the archive takes comparable time
# and often leads to conflicts in pacman book-keeping,
# while creating/updating the archive costs ~10 min.
  #- C:\msys64\var\lib\pacman
  #- C:\msys64\var\lib
  #- C:\msys64\mingw64
  #- C:\msys64\mingw32
  #- C:\msys64\ucrt64
  #- C:\msys64\clang32
  #- C:\msys64\clang64
  #- C:\msys64\clangarm64
  #- C:\msys64\usr
  #- C:\msys64\bin
  #- C:\msys64\etc
  #- C:\msys64\*.*
  #- C:\msys64\installerResources