Gitlab CI artifact storage fails
First: I'm not sure this is really an issue with our install or some bug with GitLab's S3 storage. My gitlab CI artifacts have what looks like S3 verification chunks in them. For example this raw job log has:
section_end:1530509100:build_script
[0Ksection_start:1530509100:after_script
[0Ksection_end:1530509102:after_script
[0Ksection_start:1530509102:archive_cache
[0Ksection_end:1530509104:archive_cache
[0Ksection_start:1530509104:upload_artifacts
[0K[32;1mUploading artifacts...[0;m
builddir/html/: found 229 matching files [0;m
Uploading artifacts to coordinator... ok [0;m id[0;m=6715 responseStatus[0;m=201 Created token[0;m=-A_T3aiq
section_end:1530509109:upload_artifacts
[0K[32;1mJob succeeded
[0;m
0;chunk-signature=1eec42b554458a11e6c991298185f8e0e9f8c3448a7013ca628eb1df54127db3
When the artifact is downloaded, the chunk-signature is inside the zip, corrupting the zipfile. This happens with the acacia runner but also with a second test runner set up for libinput.
This seems to break the libinput CI pipeline because we rely on artifact passing to generate docker image (see e.g. here).