Placing mdns after dns in the host database resolution can cause
unintuitive failures for users without nss-mdns or mDNSResponder. When
looking up an invalid host, having order "dns mdns" will cause a dns
"notfound" result to fall through to mdns, which will yield
"unavailable" for users without an mdns provider. This causes clients
like ssh to report "Device or resource busy" rather than a more
meaningful message to indicate that the hostname is not valid. Flipping
back to order "mdns dns" will case an "unavailable" mdns result to fall
through to dns, which then returns "notfound" that clients interpret as
expected.
Because both mDNSResponder and nss-mdns offer a generic mdns provider.
The mdns4 and mdns4_minimal providers offered by nss-mdns are subsets of
the generic mdns provider, so there is no value in restoring the
original "mdns mdns4_minimal mdns4" ordering (in fact, this may have
caused failed mdns lookups to be repeated three times). All have been
coalesced into the single mdns provider.
This (partially) reverts commit cc11e17fab.
The version string used by wx to check for a "compatible" ABI includes
the C++ ABI, which changed between GCC 9 and 10 (1013 and 1014,
respectively). This made it so launching a program built with GCC 10
with wxWidgets built with GCC 9 failed. Once wxWidgets was rebuilt with
GCC 10, applications built with GCC 9 would be the ones to fail.
When on an NFS volume, the shell redirection `: > $TIMEFILE` seems to
short-circuit the empty write when the file is already empty, so the
mtime is never updated. Using `touch` yields expected behavior.
Đoàn Trần Công Danh
John
mobinmob
Jürgen Buchmüller
Benjamin Slade
Bleznudd
Giuseppe Fierro
Leah Neukirchen
Andrew J. Hesford
Matthias von Faber
q66
Érico Rolim
Roberto Ricci
Paper
teldra
Toby Merz