Alejandro Colomar

One detail: Only the first `@` is transformed into a `\0`. If the user specifies further `@`s, they will be treated literally. If the user specifies `\0`s in the middle...

What I've already tested is configuring the following: ```json { "listeners": { "unix:@magic": { "pass": "routes", "client_ip": { "header": "X-Forwarded-For", "source": "unix" } } }, "routes": [{ "action": { "share":...

This is yet another even simpler alternative to #670 and #735 .

I'm trying to test this with curl, for which I need a socket name that doesn't end in `\0`. For that, I'm trying to do `sa->socklen--;`, but for some reason,...

> 1. `nxt_sockaddr_unix_parse` Yeah, it seems it wasn't entering my conditional. Heh! Well, then, we have still some issue. There's a trailing `\0`, even though the code also did `socklen--;`...

> Just take a glance. > > 1. In `nxt_sockaddr_unix_parse()`, there is already handling with abstract sockets. > > > ``` > #if (NXT_LINUX) > > /* > * Linux...

> before that if check makes things work. Although I think we may have tickled a bug somewhere as now ss only shows @magic as magic, even though the bind(2)...

Digging, I found a lot of dead code that was accidentally left in some of Igor's refactors. I removed it all.

I found more or less why we are getting the extra length. See the following debug log: ``` alx@asus5775:/usr/local$ sudo cat unit.log | tr '\0' '#' | strings | grep...

> s/behavior/behaviour/ Wontfix :) > s/alive/to live/ > s/bug/error/ > s/be more/be no more/ > s/an/a/ Fixed. Thanks!