diff options
author | Christopher Baines <mail@cbaines.net> | 2023-11-02 11:08:56 +0000 |
---|---|---|
committer | Christopher Baines <mail@cbaines.net> | 2023-11-02 11:08:56 +0000 |
commit | 0b566c95c301201e1f4c7fcfb43a8b2582dd81ce (patch) | |
tree | 9dad8d7372aac5f543338dcb25d123e02715fcb3 | |
parent | d7f2691a42457483027b0d9f21e3f282e37c6e0a (diff) | |
download | qa-frontpage-0b566c95c301201e1f4c7fcfb43a8b2582dd81ce.tar qa-frontpage-0b566c95c301201e1f4c7fcfb43a8b2582dd81ce.tar.gz |
Add a few more items to the TODO list
From the discussions at the Reproducible Builds summit.
-rw-r--r-- | README.org | 21 |
1 files changed, 21 insertions, 0 deletions
@@ -75,6 +75,12 @@ Builds are now happening on bordeaux.guix.gnu.org, once it's caught up, enabling i586-gnu should be possible. There might be issues with a lack of build resources (like x86_64-linux). +**** TODO Test reproducibility + +This depends on getting more data on builds back to the data service. + +Submit multiple builds for the same derivation. + *** DONE Improve patches page **** DONE Add more statuses **** DONE Support filtering by status @@ -87,6 +93,12 @@ package. *** data.qa.guix.gnu.org **** TODO Move away from Chris renting beid (the machine it runs on) **** TODO Make processing revisions faster and require less resources +**** TODO Support accepting build reports/buildinfo + +This can include the hashes of the output and be factored in to the assessment +of reproducibility. This will allow using the results of multiple builds per +build server, rather than just being able to look at a single substitute. + *** Monitoring and observability **** TODO Move away from Chris hosting Prometheus/Grafana for this purpose @@ -110,3 +122,12 @@ in the process. Maybe the QA Frontpage has a part to play in helping this to happen. **** DONE Add mark patches as "reviewed-looks-good" feature +** Reproducible builds +*** TODO Improve issue display, show severity and separate by status +*** TODO Record and expose percentage of packages that build reproducibly + +Compute the percentage, ignoring packages where the status is unknown. Also +ignore systems with little data available. + +Expose in Prometheus metrics, and record in the database. This could be +backfilled by looking through the older data service data. |