* Update VM base tools:
* Intel OneAPI updated
* Powershell Updated
* VS updated to install current Windows SDKs, and atlmfc for arm, and CMake/ninja
* Update pools.
* Remove passing-remove-from-fail list port.
* Update vcpkg-tool to 2022-03-24
* Hook up https://github.com/microsoft/vcpkg-tool/pull/345
* Hook up https://github.com/microsoft/vcpkg-tool/pull/442
* Update vcpkg-tool to 2022-03-25
* Analysis of failures.
* [Most recent nightly build failed](https://dev.azure.com/vcpkg/public/_build/results?buildId=69427)
* [Validation of this tool update failed](https://dev.azure.com/vcpkg/public/_build/results?buildId=69417)
## Common to both:
PASSING, REMOVE FROM FAIL LIST: chartdir:x64-windows (.\scripts\ci.baseline.txt)
PASSING, REMOVE FROM FAIL LIST: chartdir:x64-windows-static-md (.\scripts\ci.baseline.txt)
PASSING, REMOVE FROM FAIL LIST: chartdir:x86-windows (.\scripts\ci.baseline.txt)
Probably fixed by https://github.com/microsoft/vcpkg/pull/23701
PASSING, REMOVE FROM FAIL LIST: gmp:x64-uwp (.\scripts\ci.baseline.txt)
PASSING, REMOVE FROM FAIL LIST: gmp:x64-windows-static-md (.\scripts\ci.baseline.txt)
Probably fixed by https://github.com/microsoft/vcpkg/pull/23466 ?
REGRESSION: colmap:x64-windows-static-md failed with BUILD_FAILED. If expected, add colmap:x64-windows-static-md=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
I don't know exactly what changed. I observe that
* this thing depends on a *lot* of stuff
* on March 14 we didn't even attempt to build this
* the x64-windows ones are already in the baseline
so I skipped it.
REGRESSION: qtdeclarative:x64-windows. If expected, add qtdeclarative:x64-windows=fail to .\scripts\ci.baseline.txt.
REGRESSION: qtdeclarative:x64-windows failed with BUILD_FAILED. If expected, add qtdeclarative:x64-windows=fail to C:\a\1\s\scripts\azure-pipelines/../ci.baseline.txt.
REGRESSION: qtdeclarative:x64-windows failed with BUILD_FAILED. If expected, add qtdeclarative:x64-windows=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
REGRESSION: qtdeclarative:x64-windows failed with BUILD_FAILED. If expected, add qtdeclarative:x64-windows=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
REGRESSION: qtdeclarative:x64-windows failed with BUILD_FAILED. If expected, add qtdeclarative:x64-windows=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
REGRESSION: qtdeclarative:x64-windows failed with BUILD_FAILED. If expected, add qtdeclarative:x64-windows=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
REGRESSION: qtdeclarative:x64-windows failed with BUILD_FAILED. If expected, add qtdeclarative:x64-windows=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
REGRESSION: qtdeclarative:x64-windows failed with BUILD_FAILED. If expected, add qtdeclarative:x64-windows=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
This is a reporting change: The new world order also includes host build failures which is why it's duplicated.
See also https://github.com/microsoft/vcpkg/issues/23714
See also https://github.com/microsoft/vcpkg/issues/23490
I'm nervous about baslining this because it seems most of the qt world is built on top of this port
I filed https://github.com/microsoft/vcpkg/issues/23824 about this and @Neumann-A indicated this should be fixed by https://github.com/microsoft/vcpkg/pull/23755
REGRESSION: nettle:x64-uwp. If expected, add nettle:x64-uwp=fail to .\scripts\ci.baseline.txt.
REGRESSION: nettle:x64-windows-static-md. If expected, add nettle:x64-windows-static-md=fail to .\scripts\ci.baseline.txt.
REGRESSION: nettle:x64-uwp failed with BUILD_FAILED. If expected, add nettle:x64-uwp=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
REGRESSION: nettle:x64-windows-static-md failed with POST_BUILD_CHECKS_FAILED. If expected, add nettle:x64-windows-static-md=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
Didn't analyze, probably fixed by https://github.com/microsoft/vcpkg/pull/23519 ?
REGRESSION: libgpg-error:x64-uwp. If expected, add libgpg-error:x64-uwp=fail to .\scripts\ci.baseline.txt.
REGRESSION: libgpg-error:x64-uwp failed with BUILD_FAILED. If expected, add libgpg-error:x64-uwp=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
This was broken by VS2022 update:
```
C:\Program Files\Microsoft Visual Studio\2022\Enterprise\MSBuild\Microsoft\VisualStudio\v17.0\AppxPackage\Microsoft.AppXPackage.Targets(892,25): error MSB4086: A numeric comparison was attempted on "$(TargetPlatformMinVersion)" that evaluates to "" instead of a number, in condition "'$(TargetPlatformMinVersion)' >= '10.0.17200.0'". [C:\Dev\vcpkg\buildtrees\libgpg-error\x64-uwp-rel\error-1.42-2324ddbc71.clean\SMP\libgpg-error_winrt.vcxproj]
```
REGRESSION: libmikmod:x64-osx. If expected, add libmikmod:x64-osx=fail to .\scripts\ci.baseline.txt.
REGRESSION: libmikmod:x64-osx failed with BUILD_FAILED. If expected, add libmikmod:x64-osx=fail to /Users/vagrant/Data/work/2/s/scripts/azure-pipelines/../ci.baseline.txt.
Broken between [2022-03-16](https://dev.azure.com/vcpkg/public/_build/results?buildId=68947) and [2022-03-18](https://dev.azure.com/vcpkg/public/_build/results?buildId=69051). Unfortunately I don't see obvious reasons why. Nothing else depends on this and nobody has noticed in 2 weeks, so I'm baslining it for now. (Will investigate shortly...)
## Only broken in tool update:
REGRESSION: mesa:x64-windows failed with BUILD_FAILED. If expected, add mesa:x64-windows=fail to C:\a\2\s\scripts\azure-pipelines/../ci.baseline.txt.
```
-- Downloading https://gitlab.freedesktop.org/mesa/mesa/-/archive/mesa-21.2.5/mesa-mesa-21.2.5.tar.gz -> mesa-mesa-mesa-21.2.5-1.tar.gz...
-- Extracting source /Users/vagrant/Data/downloads/mesa-mesa-mesa-21.2.5-1.tar.gz
-- Applying patch swravx512-post-static-link.patch
-- Applying patch swr-msvc-2.patch
-- Applying patch swr-llvm13.patch
-- Applying patch radv-msvc-llvm13-2.patch
-- Applying patch d3d10sw.patch
-- Using source at /Users/vagrant/Data/buildtrees/mesa/src/esa-21.2.5-2df234d2b1.clean
Traceback (most recent call last):
File "<string>", line 1, in <module>
ModuleNotFoundError: No module named 'mako'
CMake Error at ports/mesa/portfile.cmake:85 (message):
Python package 'mako' needs to be installed for port 'mesa'.
Complete list of required python packages: setuptools;mako
Call Stack (most recent call first):
ports/mesa/portfile.cmake:91 (vcpkg_get_python_package)
scripts/ports.cmake:145 (include)
```
Looks like this is being tracked by https://github.com/microsoft/vcpkg/pull/23089 ; perhaps that we don't have as aggressive a recycling strategy for macos boxes as we do for the others has let different machines give different results?
## Only broken without tool update:
REGRESSION: chromium-base:x64-osx. If expected, add chromium-base:x64-osx=fail to .\scripts\ci.baseline.txt.
This one has been constantly flaky; I baselined it.
REGRESSION: libxml2:x64-osx. If expected, add libxml2:x64-osx=fail to .\scripts\ci.baseline.txt.
This port uses vcpkg_from_git and the upstream server was down during the build.
* Restore chartdir to the baseline, I thought https://github.com/microsoft/vcpkg/pull/23732 had been merged.
* Trigger some test ports from vcpkg.cmake changes
* Bump test port version
* Revert "Bump test port version"
* Use old toolchain version for parent hashes
* Add ABI hashing note to vcpkg.cmake
* Test vcpkg.cmake with cmake instead vpckg-ci-paraview
* Update Linux VM CUDA to 11.6.
* Update Intel OneAPI
* Update Visual Studio to 2022, and remove deployment of v140 and v141 compilers.
* Update CUDA to 11.6 on Windows.
* Update pools.
* cix ci.baseline.txt
Co-authored-by: nicole mazzuca <mazzucan@outlook.com>
* Update vcpkg-tool to 2022-03-08.
* Back off required git version to something actually in the build lab.
* Remove hiding of the output of --parent-hashes.
* Update 7zip reference as indicated in https://github.com/microsoft/vcpkg-tool/pull/413
* Add triplet to name of main CI task
* Fix osx bootstrap task name
* Hide output from switch to parent hash
* Bump test port version
* Hardcode osx+linux triplet names
* Improve logging order
* Add error handling to each tool download
* Revert "Bump test port version"
* Don't prefetch 7zip
* Some of the Windows runs have started to exceed 24 hours now as well as the MacOS ones as we have added more ports.
* We are no longer trying to constantly "psychic debug" why ports are failing; CI passes somewhat regularly now, and the failures we get seem to be much more deterministic.
* We're still budget constrained on Azure resources (and have been a bit over budget)
To that end, we're going back off nightly CI runs to Monday Wednesday Friday (like osx already is), so we no longer need separate bits for them.
* Minor bugfixes to MacOS deployment readme.
* Update vsts-agent to 2.198.3
* Remove rimraf of ~/vagrant/vcpkg-eg-mac
* Add fetch and branch switch before running Install-Prerequisites.ps1 to make sure it's updated
* Clarify "the fileshare" as ~/vagrant/share
* Clarify that the date `Setup-VagrantMachines.ps1` wants is the box version date.
* `Setup-VagrantMachines.ps1` parameter to be DevopsPat rather than Pat
* nicole updates
Co-authored-by: nicole mazzuca <mazzucan@outlook.com>
* add moar ram (no PR)
* Provision autoconf-archive on osx (PR #22872)
* [mpfr] Support Linux and OSX (PR #22845)
* VM software versions
* fix minor schema bug
* update pool
Co-authored-by: nicole mazzuca <mazzucan@outlook.com>
Co-authored-by: Kai Pastor <dg0yt@darc.de>
Co-authored-by: JackBoosY <yuzaiyang@beyondsoft.com>
* Revert "incorporate changes from microsoft:master"
* Revert "Revert "incorporate changes from microsoft:master""
* preliminary approach to solve pip issue
* fix python venv
* fix macOS build command again
* bump version
* x-add-version
* Fix spelling of prerequisites.
* Add python3-venv
* Update linux pool.
Co-authored-by: Joachim Gehweiler <joachim@Joachims-iMac.local>
Co-authored-by: Billy Robert O'Neal III <bion@microsoft.com>
* Rewrite to single invovation of 'vcpkg ci'
* Split caching args from common args
* Use parent hashes for pull request CI builds
* Completely switch to parent revision state and back
* [macos ci] move out tensorflow
* rename ExtraPorts and CorePorts
Co-authored-by: nicole mazzuca <mazzucan@outlook.com>
Co-authored-by: Billy Robert O'Neal III <bion@microsoft.com>
* [ci] Move formatting and doc checks into x86-windows to avoid burning a whole job for that 1 run, don't run file lists generation in PRs.
* Deduplicate x86-windows and restore PR file lists, as requested/suggested by @dg0yt
* More quotes
Co-authored-by: nicole mazzuca <83086508+strega-nil-ms@users.noreply.github.com>
Co-authored-by: nicole mazzuca <83086508+strega-nil-ms@users.noreply.github.com>
* Decouple the storage account used for binary caching from the images.
We are being forced to change infrastructure somewhat more frequently than we would like, so change the binary caches to use a shared storage account like we already use for asset caching.
* blob blob blob
* Add quotes
* Avoid name stomping.
* update 2021-07-27
macOS -> 11.5.1, XCode -> 12.5.1
* update azure-pipelines
* update sha of macfuse
* change how macos-ci-base works
* fix build errors
* fix itpp:linux
* more fixes
* remove tab
* allow version changes in all the remove/rename-version patches
* fix libunistring for real
* robert CR
Now that vcpkg has artifact caching, meaning downloading artifacts usually doesn't have to leave the datacenter, there's no need to potentially dirty different runs with leftover downloads from each other.
This resulted in some "impossible" results in https://github.com/microsoft/vcpkg/pull/18687 where a previous PR run accidentially downloaded the Windows version of nodejs to /mnt/vcpkg-ci/downloads/node-v14.17.4-linux-x64.tar.xz
Remove no longer used "unstable" scaffolding.
Use ephemeral OS disks for better latency.
Add ability to change whether temp disks are used or explicitly provisioned disks are used; explicitly provisioned disks turned out to be more expensive than expected before.
Disable Windows Updates in the scale set because the machines are very short lived; working on a mechanism of patching the images instead.
