One month ago Petr Šabata posted a module build proposal here on this
I'd like to gather some ideas about the very first of his key points as
it poses some problems with our current buildsystem:
> - building a module means building the components it contains and
> composing a module deliverable (a repository, an image, other)
Rebuilding compoments is necessary as the buildroot of modules can be
quite different from package builds which usually just take whatever
package is tagged last into a koji tag. For modules it will depend on
the modules definition.
But we can't just simply rebuild the same RPM (same
name-version-release) in koji for a module. koji correctly doesn't allow
To make it worse, that same RPM might be required by different modules
with different buildroots, thus needing a rebuild there as well.
The current idea is to use the dist tag for this and let privileged
users or koji itself (based on some rules) overwrite the %dist tag for
builds of module components. This lets us automate those builds fairly
Another approach would be to modify the package release for each
rebuild. That still allows us to automate rebuilds to some extend, but
for example finding an upgrade path might get a little messy.
Does anyone have any other ideas for this problem ?
I did update from 2.30.2 to 2.31.0 and didn't notice that there was
API breakage, but soname was still -> 0. Upstream breaks API each
X.Y.0 release so I did some changes in their buildsystem and now
soname is -> 2.31 and will change each X.Y.0 release.
Looks like that nothing except SuperTuxKart uses it (which I will rebuild soon).
Thanks for attention ;)
Today's Rawhide compose contained no images at all (which is why there
are no openQA tests for it, in case you wondered). At least some of the
compose fails (I haven't checked them all) seem to be caused by a
I suspect that complete borkage in libselinux-python may affect quite a
lot of stuff, though I don't know specifics ATM, so I'd probably
suggest avoiding updating libselinux for now.
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
there seems to be problem with COPR ATM because it reports an error
> (psycopg2.ProgrammingError) column copr.unlisted_on_hp does not exist
> LINE 2: ..., copr.build_enable_net AS copr_build_enable_net,
> copr.unlis... ^ [SQL: 'SELECT count(*) AS count_1 \nFROM (SELECT
> copr.id AS copr_id, copr.name AS copr_name, copr.homepage AS
> copr_homepage, copr.contact AS copr_contact, copr.repos AS
> copr_repos, copr.created_on AS copr_created_on, copr.description AS
> copr_description, copr.instructions AS copr_instructions,
> copr.deleted AS copr_deleted, copr.playground AS copr_playground,
> copr.auto_createrepo AS copr_auto_createrepo, copr.user_id AS
> copr_user_id, copr.group_id AS copr_group_id, copr.forked_from_id AS
> copr_forked_from_id, copr.webhook_secret AS copr_webhook_secret,
> copr.build_enable_net AS copr_build_enable_net, copr.unlisted_on_hp
> AS copr_unlisted_on_hp \nFROM copr JOIN "user" ON "user".id =
> copr.user_id LEFT OUTER JOIN "group" ON "group".id = copr.group_id
> \nWHERE copr.deleted IS false AND copr.unlisted_on_hp IS false ORDER
> BY copr.id DESC) AS anon_1']