pkgsrc-Bugs archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
pkg/59406: pbulk: upload should omit restricted packages from pkg_summary
>Number: 59406
>Category: pkg
>Synopsis: pbulk: upload should omit restricted packages from pkg_summary
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: pkg-manager
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Thu May 08 12:50:00 +0000 2025
>Originator: Taylor R Campbell
>Release: current
>Organization:
The NO_NETBSD_ON_FOUNDATION
>Environment:
>Description:
pbulk will:
1. build _all_ packages (assuming SKIP_LICENSE_CHECK=yes),
2. record NO_BIN_ON_FTP notes in presolve,
3. generate a pkg_summary for _all_ built packages with pkg_info -X, and
4. upload that pkg_summary, together with the non-restricted ones based on presolve (determined by compute-packages.awk).
The resulting uploaded pkg_summary will have entries for binary packages that weren't uploaded (if there are any NO_BIN_ON_FTP packages).
Later, a cron job messes with the pkg_summary, via the (sometimes mangled) .pkgcache it also uses to generate readmes, so that it doesn't list the restricted packages. (But sometimes it just lists mangled results, as in PR pkg/59365: binary package repo wrong, https://gnats.netbsd.org/59365)
This is bonkers. What we upload, and publish at known URLs, should be a consistent snapshot, and should not be mangled in place by a poltergeist provoked by cron.
>How-To-Repeat:
infrastructure inspection
>Fix:
Teach pbulk to limit pkg_summary it uploads to the packages it's uploading too.
Perhaps we should just run compute-packages.awk (via ${packages_script}) in the build script, rather than in the upload script.
Home |
Main Index |
Thread Index |
Old Index