Keys sent with send-keys -K keep KEYC_SENT when they reach input_key, so
most of them never get to the pane. Remove the flag early before trying
to process. From Martin Trojer.
rpki-client: avoid a double free in state file parsing
Properly transfer ownership of nca and ca to the state trees. Otherwise
an ferror() after a successful RB_INSERT() could result in a double free
in the error path.
ok claudio job
httpd: use the host of an absolute-form request-target
For "GET http://host/path" the host from the URL replaces the Host header, as
RFC 9112 3.2.2 requires, and the path becomes the request path. Only http(s) are
accepted, and URLs with userinfo are rejected. Also reject requests with more
than one Host header.
http_path_orig now holds the origin-form path, so the access log and
REQUEST_URI show "/path" instead of "http://host/path".
parse_url() is copied from relayd.
Feedback and ok armani@
pppoe(4): account for packets sent through the direct transmit path
pppoe_enqueue() bypasses ifq_enqueue() with the default priq, leaving
outbound data packets uncounted. Count packets and their pre-PPPoE
length after successful transmission; sppp_output() already accounts for
the framing bytes.
From Steffen Beyer.
OK dlg@
httpd: check access settings of the directory index location
When you ask for a directory, the index file can end up being served
by another location, and we never looked at its header rules, body
limits, block or authenticate. With authenticate on a fastcgi location
this even took the whole daemon down.
Move these checks into server_http_access() and call it from
server_file(), so every location that serves a request gets checked.
Hang on to the Authorization header until the request is done, so
each location can check it, and wipe it before handing the request
to fastcgi.
With input and help from armani@, OK armani@, kirill@
Reported by Artus KG
Add support for head -c, standardized in POSIX 2024
Ports patches will follow when amd64 snaps ship with head -c.
Input from ian@, tests from matthieu@, ok op@%
Fix endless loop in udp_output() on 32 bit architectures.
As CMSG_ALIGN() can overflow the integer value, do the length check
with and without alignment.
from Acts1631; OK mvs@
rpki-client: do not leak cert if auth_insert() fails
If certid reaches CERTID_MAX or the depth exceeds the maximum cert depth,
cert is not inserted into the auth tree. In this case we need to free the
cert.
Reported by Ties de Kock
ok claudio