Reduce image tags for Piglit uprev
The image tags for Piglit uprevs included a short hash. Those tags are fields to build the docker image name and, with the new field of the pkg_repo_rev, the total size exceeded the string length limit. This change tries to help here by reducing 9 characters.
Image tag in .gitlab-ci/image-tags.yml have the structure {{ date }}-{{ topic }}. This is meant to point to an existing image or, in the build jobs, if it doesn't exists build a fresh one. They are not only rebuilding the "topic", when there is a change in one of the dependencies, a change in the tag rebuild everything including the modified dependency.
To uprev Piglit, we have to modify three tags (aka, DEBIAN_X86_64_TEST_GL_TAG, DEBIAN_X86_64_TEST_VK_TAG, and KERNEL_ROOTFS_TAG). The "topic" field was set to piglit-{{ short hash }} with 8 characters to be sure we are building with the new piglit if there was already an image build today with a previous Piglit.
But with a recent attempt to uprev, some build jobs failed. One sample of a failing build job show a docker image name registry.freedesktop.org/gfx-ci-bot/mesa/debian/x86_64_test-gl:2024-01-18-piglit-814046fe--2024-01-14-runner--67f2c46b--v6.6.4-for-mesa-ci-e4f4c500f7fb--d5aa3941aa03c2f716595116354fb81eb8012acb that means 193 characters. One image that is working is registry.freedesktop.org/mesa/mesa/debian/x86_64_build:2024-01-04-find--2024-01-14-runner--67f2c46b--d5aa3941aa03c2f716595116354fb81eb8012acb that is 141 characters long. Current last pre-merge pipeline that build images has created registry.freedesktop.org/mesa/mesa/debian/x86_64_test-gl:2024-01-16-kmod--2024-01-14-runner--67f2c46b--v6.6.4-for-mesa-ci-e4f4c500f7fb--d5aa3941aa03c2f716595116354fb81eb8012acb that is 176 characters long.
I think the name fields (for gl but similar for others) are:
${DEBIAN_X86_64_TEST_GL_TAG}--${KERNEL_ROOTFS_TAG}--${PKG_REPO_REV}--${KERNEL_TAG}--${MESA_TEMPLATES_COMMIT}Which I guess a maximum size of 192 characters (because is in the range is the closes power of 2). The size is 128 as @eric mentioned.
This is not a truly solution as far as we don't have a equitable distribution with other field of the available space. But it means a reduction of what it is not really needed for the Piglit's uprev.