This is a new error that just started cropping up. I suspect it's related to recent SSH changes. This might help a little with context, when it happens again.

Totally not sketchy, trust me.

Interestingly, when they fail, Synchronet appears to crash.
This is fine.

These should've been static to begin with, like the other servers. Explains why FTP error message are with "term" as the server.

(such as ec22519), unconditionally fallback to password auth instead
of the previous behaviour of diconnecting.

Much of the "password then pubkey" mentality is showing in these
bits of Cryptlib.

For Nightfox and his newsgroup reader thing: The strings that end up in the editor drop file (e.g. MSGINF) can now be passed as optional arguments to consol.editfile() and the underlying C++ API.

Allow maxlines (default:10000) to be overridden with an optional consol.editfile() argument too.

On Win32, size is explcitly "ignored" when target is NULL, so we
should set it to the size of the allocation.

Previously, once a pubkey was attempted, you could not use a password.

When we're closing the socket, always delete channels and don't wait
for the remote to confirm.

Fixes assertion failure on hangup from SSH.

Was spinning at 100% CPU on a lost connection if sftp was active.

328:25: warning: ‘pubkey’ may be used uninitialized

DDMsgReader: Changed rescan hotkey for the indexed menu to Ctrl-R, since Ctrl-S won't work for Wyse terminals (Ctrl-S is for xon/xoff toggling and can't be changed)

See merge request !389

Eric Oulashin authored 1 year ago and Rob Swindell committed 1 year ago

DDMsgReader: Changed rescan hotkey for the indexed menu to Ctrl-R, since Ctrl-S won't work for Wyse terminals (Ctrl-S is for xon/xoff toggling and can't be changed)

TLS 1.2.  For static services and JS sockets, add a new tls_minver
socket property which can be set to use a lower minver (down to the
wildly insecure TLS 1.0).

TLS 1.0 and 1.1 have been not reccomended since 2015, and deprecated
since 2021.

Update cryptlib to 3.4.7

See merge request !388

Even on filesystems where you can do it, you shouldn't.  Stop it.

The server is a lot thinner, with callbacks for each operation.

The client_socket_dup is overwritten after with the return value of accept(),
so this was likely a socket/handle leak (since commit 74470573, 17 years ago).

Noticed this only after logging a new error in an SSH connection when this
call to DuplicateHandle() failed for some reason (WinError 6), but it seems
like it's been redundant code for a long time now. So just ace it and possibly
fix a resource leak.

Merge branch 'ddmr_indexed_mode_mark_all_read_updates_and_newscan_force_indexed_menu_opts' into 'master'

DDMsgReader: Updated behavior for indexed menu R (mark all read) when doing a newscan. Ctrl-S does scan refresh in indexed menu. New R "snap-to" option for the indexed menu. Also, command-line options for forcing use of the indexed mode menu when doing a

See merge request !387

Eric Oulashin authored 1 year ago and Rob Swindell committed 1 year ago

DDMsgReader: Updated behavior for indexed menu R (mark all read) when doing a newscan. Ctrl-S does scan refresh in indexed menu. New R "snap-to" option for the indexed menu. Also, command-line options for forcing use of the indexed mode menu when doing a

Broken 11 months ago with commit ae44ab15: the unit argument was ignored (e.g. converting to kibibytes) and the free/total disk size/space checks were reversed.

Reported by Max (WESTLINE) via usage of JS properties: system.freediskspace and system.freediskspacek.

The JS global functions dir_freespace() and disk_size() were also broken due to this bug.

This fixes ssl.c build for Windows

scfgnet.c:1463:25: warning: result of comparison of constant 100000 with expression of type 'uint16_t' (aka 'unsigned short') is always true