Store glib build log on build failure

The Windows CI builds are intermittently failing. Even though a prebuilt vcpkg
is already cached, it will sometimes choose to rebuild lots of packages anyway.
On some of those rebuilds, there's a failure when building glib.

On the times that it does fail, the failure is often enough glib that checking
the build log manually seems more reasonable than checking what failed.

An alternative would be to look at why it sometimes chooses to rebuild, but
that would mean remotely debugging a CMake config. I'm hoping the build log
will have a clear compiler error pointing to a race condition in the order that
packages are built.
This commit is contained in:
Steve Cotton 2023-02-03 04:50:15 +01:00 committed by Steve Cotton
parent ee0830f53f
commit 4c22f736ec

View file

@ -287,6 +287,15 @@ jobs:
run: |
cmake -DCMAKE_BUILD_TYPE=%CFG% -DENABLE_GAME=true -DENABLE_SERVER=true -DENABLE_CAMPAIGN_SERVER=true -DENABLE_TESTS=true -DENABLE_MYSQL=false -DENABLE_NLS=false -DVCPKG_TARGET_TRIPLET=x64-windows -DCMAKE_TOOLCHAIN_FILE=D:/a/wesnoth/vcpkg/scripts/buildsystems/vcpkg.cmake -DCMAKE_GENERATOR_PLATFORM=x64 -G "Visual Studio 16 2019" .
# Trying to find the cause of the intermittent build failures. This assumes that the failure was during building glib,
# but that seems regular enough that I'll check manually to see if it was.
- name: Store glib build log
uses: actions/upload-artifact@v3
if: failure()
with:
name: glib-log
path: D:/a/wesnoth/vcpkg/buildtrees/glib/package-x64-windows-*-out.log
# delete buildtrees directory to free up space after cmake invokes vcpkg to build the dependencies
# otherwise the job was failing when trying to write a .obj file
# building vcpkg on the more spacious C drive didn't work since for some reason vcpkg decides to not create the pango/cairo pkgconfig files there