On Mon, Sep 14, 2020 at 12:53:00PM +0200, Philippe Mathieu-Daudé wrote: > +Michael & Bruce > > On 9/14/20 11:54 AM, Gerd Hoffmann wrote: > > Hi, > > > >> The CI idea is to have reproducible builds if possible. > >> When the submodule is updated (or the QEMU scripts containing the > >> -D defines), it triggers the 'build-edk2' job which produce these > >> same binaries. > >> My original idea was to push the tag on GitLab that triggers the > >> job, then download the produced binaries, test them, then commit > >> them. > > > > Now with CI in place we maybe should step back and have a look at the > > big picture. > > > > Should we simply stop committing firmware binaries into git and provide > > them as CI artifacts instead? > > What I'm not sure about is how to keep the built artifacts forever.
Dunno. gitlab has support for various packages and container images, but not for tarballs or rpms or raw binaries. Possibly we could publish them as gitlab pages. > Also we need to modify the source tarball generator script to fetch > the artifacts and include them, isn't it? An firmware download script certainly makes sense. Whenever we want continue including the firmware in the tarballs when we dropped it from the git repo is up for debate ;) take care, Gerd
