frequent occurrences of tag mismatches were probably triggered by
the import of qt5, which is a long-running nojunk port.
when we're finished with kdeN ports, there's pressure to force junking to
go thru kdeM ports, BUT actual junking *won't* happen in the presence of
a nojunk port, though the current code make it seem as though
we've "succeeded".
Forensics shows:
23826@1431149112: K: x11/kde4/webdev openbsd-2 kde3 vs kde4
23826@1431149112: J: devel/hs-FindBin openbsd-2
23826@1431149119: B: security/p5-Crypt-OpenSSL-RSA
23826@1431149120: J: x11/tellico-kde4 openbsd-2
which made no sense since nothing happened between the K and the J.
But the log of tellico shows the junk not happening.
Still tainted: 1
>>> Running junk in x11/tellico-kde4 at 1431149274
Can't run junk because of lock on x11/qt5,,-main
So synch "can_be_unjunked" for forced junks: it should not succeed if there's
a nojunk port.
through the PLT, so do on OpenBSD what Linux and FreeBSD are already doing.
Fixes build with binutils 2.17 and fixes a potential runtime issue with
binutils 2.15.
ok espie@
Use that when switching groups to enforce full correct list of groups.
Figure out users that must exist locally, and error out if they don't.
Based on feedback by sthen@
- In addition to sodium_bin2hex(), sodium_hex2bin() is now a
constant-time function.
- crypto_stream_xsalsa20_ic() has been added.
- crypto_generichash_statebytes(), crypto_auth_*_statebytes() and
crypto_hash_*_statebytes() have been added in order to retrieve the
size of structures keeping states from foreign languages.
- crypto_box_seal() and crypto_box_seal_open() have been added.
introduced there's no longer need for a log socket in the chroot jail,
also demonstrate use of rcctl rather than asking the user to manually
change rc.conf.local (making the message more generic while there).
OK brad.
that it can be used to bake PIE. The binutils 2.15 linker (incorrectly)
allows direct calls, but binutils 2.17 flags this as an error.
ok Brad (maintainer)
