Details

Reviewers

jlduran
jrm
ashish

Summary

Compact and verify the target pkg database after manual edits
(metalog fixups, trigger runs, etc).

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Build Status

Buildable 76086
Build 72969: arc lint + arc unit

Event Timeline

senguptaangshuman17_gmail.com created this revision.Thu, Aug 20, 12:22 PM

Herald added a subscriber: imp. · View Herald TranscriptThu, Aug 20, 12:22 PM

senguptaangshuman17_gmail.com requested review of this revision.Thu, Aug 20, 12:22 PM

Harbormaster completed remote builds in B75840: Diff 184511.Thu, Aug 20, 12:22 PM

senguptaangshuman17_gmail.com added a parent revision: D59045: nanobsd: Add _xxx_pkg_add_var_db_files_to_metalog.Thu, Aug 20, 12:33 PM

senguptaangshuman17_gmail.com added a child revision: D58961: nanobsd: Clean up target pkg database and metalog after install.

Address review comments: normalize comment to # XXX:

Herald added a reviewer: scheduler. · View Herald TranscriptThu, Aug 20, 7:03 PM

Harbormaster completed remote builds in B75914: Diff 184600.Thu, Aug 20, 7:03 PM

lwhsu removed a reviewer: scheduler.Thu, Aug 20, 8:17 PM

I think this should only do a vacuum (integrity check).
So far we have not been able to create deterministic pkgbase images.

In D59046#1354947, @jlduran wrote:

I think this should only do a vacuum (integrity check).
So far we have not been able to create deterministic pkgbase images.

As per my understanding, I'd like to keep the dump rather than dropping it, the function just dumps the db as a precondition for someday diffing/reproducing images. Once pkgbase builds are reproducible, this becomes the hook that lets us verify it, without another patch reintroducing it then.
Please let me know your view on this, then will change accordingly.

senguptaangshuman17_gmail.com updated this revision to Diff 184834.Sun, Aug 23, 11:02 PM

Harbormaster completed remote builds in B76020: Diff 184834.Sun, Aug 23, 11:02 PM

In D59046#1356180, @senguptaangshuman17_gmail.com wrote:

In D59046#1354947, @jlduran wrote:

I think this should only do a vacuum (integrity check).
So far we have not been able to create deterministic pkgbase images.

As per my understanding, I'd like to keep the dump rather than dropping it, the function just dumps the db as a precondition for someday diffing/reproducing images.

OK. Then, this must be stated in the commit message, and clarified in the comment above the function. I'm not sure if that is even possible.

Once pkgbase builds are reproducible, this becomes the hook that lets us verify it, without another patch reintroducing it then.

The pkgbase builds are currently reproducible. The installations are not. Since we are installing (pkg install), and that process populates an sqlite3 database, the resulting database file is not deterministic.

senguptaangshuman17_gmail.com updated this revision to Diff 184963.Tue, Aug 25, 6:11 AM

Harbormaster completed remote builds in B76086: Diff 184963.Tue, Aug 25, 6:11 AM

senguptaangshuman17_gmail.com retitled this revision from nanobsd: Add _xxx_pkg_db_dump_or_vacuum to nanobsd: Add _xxx_pkg_db_vacuum.Tue, Aug 25, 6:14 AM

senguptaangshuman17_gmail.com edited the summary of this revision. (Show Details)Tue, Aug 25, 6:26 AM

Rebase stack on top of D59043 update

Harbormaster completed remote builds in B76122: Diff 184999.Tue, Aug 25, 8:00 AM

senguptaangshuman17_gmail.com abandoned this revision.Thu, Aug 27, 9:31 PM

senguptaangshuman17_gmail.com removed a child revision: D58961: nanobsd: Clean up target pkg database and metalog after install.