holger at layer-acht.org
Mon Mar 6 12:54:16 UTC 2017
On Mon, Mar 06, 2017 at 12:21:29AM +0100, Andreas Beckmann wrote:
> Andreas Beckmann (6):
> p: fall back to on-the-fly generation if --end-meta file is missing
> p: pass the full chroot state to restore_selections
> p: factor out Chroot.get_state_meta_data()
> p: rename Chroot.save_meta_data() to Chroot.get_tree_meta_data()
> p: keep history of 'apt-cache dumpavail | md5sum'
> p-s: use the --save-end-meta and --end-meta options
> this should save some time on upgrade tests
> the metadata file automatically gets updated (in the next test) if
> discrepancies are noticed
> so far I didn't run into a failure due to outdated metadata
> there is one metadata file kept per slave instance and config section
> the first 4 packages could be used as refactoring without new features
I don't understand…
> the first patch is for copy+paste to rerun a piuparts command from a
> logfile (which usually uses -E) *without* an existing metadata file
> (from the *correct* upgrade path) in $(pwd)
yay! This should be made clear very clearly in the logs, "copy+paste this to
re-run this test on your machine" or so…
> PS: now that we have some testing power, what do you think about adding
> [oldstable2stable2testing] ?
I like it!
When thinking about adding more archs to piuparts last night before falling
asleep I thought that the bottleneck is probably piuparts-reports now and
I wondered whether we'd better rewrite this, using storing results in a
postgresql db and then creating the results pages, quite similarly what
we're doing for tests.reproducible-builds.org/debian - maybe even sharing
some code there?
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 811 bytes
Desc: Digital signature
More information about the Piuparts-devel