#tmux wykrzacza się z #libevent 2.2.1 (alpha). Niestety, komunikat błędu wypisywany jest na nowym terminalu, a tmux od razu go zamyka i czyści ekran.

Próbowałem wydostać tekst poprzez przekierowanie wyjścia, poprzez logi tmuksa, nie udało się nawet przy pomocy strace (!). Dopiero #asciinema dało radę.

Kolejne zastosowanie dla tego świetnego programu. Szczerze polecam!

https://github.com/tmux/tmux/issues/3572

#Gentoo

#tmux crashes with #libevent 2.2.1 (alpha). However, the assertion message only gets printed inside the new terminal, and tmux closes it immediately and clears the screen.

I couldn't manage to get the message via redirecting output, via tmux logs, even via strace (!). Finally, I've figured out to use #asciinema.

Another great use case for this great program. Kudos!

https://github.com/tmux/tmux/issues/3572

#Gentoo

Today's #Gentoo fun:

tmux fails to build with #libevent 2.2.1. It fails because of incompatible declaration of forkpty() in compat.h.

Looking into compat.h. forkpty() is declared there if HAVE_FORKPTY is false.

Looking into config.log. forkpty() is false because configure tries to link random test programs to -levent_core@CMAKE_DEBUG_POSTFIX@.

At this point, you can guess what the problem is: https://github.com/libevent/libevent/issues/1459

#CMake #autotols

#glibc 2.36 introduces arc4random(). This makes #libevent switch from bundled to system ARC4 when rebuilt. Good so far.

However, glibc doesn't support arc4random_addrandom(), so libevent… removes evutil_secure_rng_add_bytes() and changes its ABI, effectively breaking installed #Tor (#Gentoo is especially affected).

Upstream argues that leaving the function as no-op is bad for #security However, when it's gone, Tor just skips the call.

https://github.com/libevent/libevent/issues/1393
https://gitlab.torproject.org/tpo/core/tor/-/blob/main/src/lib/evloop/compat_libevent.c#L483-485

Oh, groovy!

Also thanks to @eborisch #libevent in #MacPorts now has a commit merged with a patch that is a *little* bit ahead of libevent's versioned release in order to ameliorate some issues with #LibreSSL noticed on versions above 3.5+ (this is related to what I was testing earlier today):

https://github.com/macports/macports-ports/pull/16681

Hopefully that helps folks who were running into issues such as documented in this Trac ticket: https://trac.macports.org/ticket/66096

(Casually wonders why #MacPorts has
#libevent
https://ports.macports.org/port/libevent/ [more or less current with upstream]

and
#libevent1

https://ports.macports.org/port/libevent1/ [what is this doing here? I am so confused. Though obviously has some different things which depend on it such as honeyd, also by Niels Provos. So: maybe old and crufty stuff needs older libevent1?] )

There's still this #libevent diff that I have been meaning to test: https://github.com/macports/macports-ports/pull/16681/files

I have definitely noticed some occasional oddities with #LibreSSL vis-à-vis libevent, so hopefully that rectifies them?

#MacPorts

Bug report submitted to #libevent to discuss adding a callback API for accessing the global namespace for better #FreeBSD #Capsicum support: https://github.com/libevent/libevent/issues/601

This will be needed to complete the Capsicum support in #Tor.