Collaborative Working Sessions - RB tooling

State of RB tooling

reprotest it builds pacakge twice and runs diffoscope TODO: teach about more build types make into a github/CI action ALSO: expand it to also do variance testing, such as: toolchains, time and place, flags and settings, single thread vs multi thread build

diffoscope big files make it go OOM output can be hard to read which flags do we pass to diffoscope? TODO: output of all the flags, into an html output, with check boxes in that output to show/hide pieces mirroring the CLI flags

trydiffoscope slower than running diffoscope locally TODO: print out the command line arguments used to produce the output

MKOSI & systemd-repart tiny differences between invocations reproducible inputs, non-reproducible output

rebuilderd easy to setup setting up the backends is difficult, different for each backend TODO: documentation needs work one instance & one database & many distros = RULE THE WORLD WANT: integration across the tools store the build logs outside of the database stop the log truncation some pieces need optimization: queue, data access, storage distribute builds based on ahead-of-time resource allocation

reproduce containers diffoci will show the differences between containers diffoci can also see and compare the layers of a container

package management installing A then B vs installing B then A? does the date/time of installation matter? (post-install scripts, etc) ALSO: reproducible ISOs reproducibility of firmware for flashing check reproducibility status of packages when installing them only allow the installation of packages which are reproducible

data sharing who? format? when? inputs?

uncategorized check reproducibility as part of the upload process in distros or language repositories reproducible status of packages, checked recursively is the package still reproducible if cross-built strip-nondeterminism with more output, to show the nondeterminism in the package