| Age | Commit message (Collapse) | Author |
|---|
|
checksum of a NATed UDP or TCP packet in the payload of an ICMP
packet anymore. Disable all test that rely on this feature.
Check that the router's interface mtu has been set to 1300.
|
|
interface mtu has been set to 1300. New netcat needs -N for shutdown.
The IPv4 and IPv6 addresses for the af-to tests must have corresponding
host numbers.
|
|
ok reyk
|
|
ok reyk
|
|
remote test machine.
|
|
softly if the setup is incomplete. Link them to the build.
|
|
run over 4 machines. Add a check-setup target to verify that
interface addresses and routes are properly set up.
|
|
|
|
OPENSSL_NO_X509_VERIFY. We're not going to build with these and the same
removal has already been done for libssl.
|
|
makes it talk to a Go TLS server.
|
|
ok reyk
|
|
ok bluhm@ benno@
|
|
files. We do not want to put a license header into each subtest.
suggested by reyk@
|
|
|
|
common web servers). Add a related regress test.
OK benno@
|
|
one fresh daemon is started with an individual config. It is talking
OSPF via a tun interface. At the process side of the tun device a
client is running. This test programm is parsing the OSPF packets
and generates new ones. It simulates other OSPF daemons.
For now only hello packets are tested. Each test has a task list
which drives the ospfd through its interface state machine. The
plan is to extend the framework for the other OSPF packet types and
state machines. Eventualy the generated kernel routing messages
could also be checked.
I have developed this test suite together with Florian Riehm.
|
|
RT.
|
|
with an empty line and a header line.
|
|
OpenSSL trunk
|
|
request.
|
|
The echo reply has to be fragmented.
|
|
- Move the tests from splice to new sosplice directory for consistent naming.
- Split the API tests and the TCP splicing tests into separate directories.
- Create some tests for the upcoming UDP splicing.
- Tests can be run in obj directories now.
- The API tests can run both on a local and on a remote machine now.
- Fix the forking TCP tests which splice and read or write simultaneously.
- Bunch of little fixes for races in the tests.
- Deduplicate code, move checks into common functions.
|
|
|
|
executed tests. This allows to micro benchmark relayd.
based on a diff from andre@; OK reyk@
|
|
with associated lists instead of the complicated lookup table and
"others" list. This might add a little malloc overhead for common
headers but also fixes some issues like the handling of repeated
headers - for example, handling of multiple "Set-Cookie" headers.
ok bluhm@ (regress part)
ok benno@
|
|
|
|
ok beck@
|
|
Noticed testing with clang.
|
|
Solaris and OS X clobber the signal stack when returning to the main
stack, which caused the original testing strategy (inspecting the
signal stack once we're back on the main stack) to fail.
To be compatible with this behavior, the regress test now inspects the
signal stack space while we're still executing on it. This is a bit
iffy because we might clobber it ourselves while inspecting it, but we
as long as its not completely clobbered we should be okay.
thx bcook for the Solaris test account
|
|
|
|
|
|
suggested by reyk@
|
|
per default for the default length value in check_logs(). If a
test doesn't want this check, it has to disable it explicitly with
nocheck. Also remove unneeded nocheck switches for deeper testing.
|
|
|
|
From RFC 7230, section 3.2.2:
"A sender MUST NOT generate multiple header fields with the same field
name in a message unless either the entire field value for that
header field is defined as a comma-separated list [i.e., #(values)]
or the header field is a well-known exception (as noted below)."
|
|
that offers multiple requets per HTTP connection is run twice.
First all requests are run over a single TCP connection. Then for
each requests a separate TCP connection is established. If a
requests during a persistent connection fails, a reconnect is done
as relayd cancels the whole connection. This allows stricter tests
for filters.
OK reyk@
|
|
run on altstack.
|
|
comparisons
|
|
As discussed with bluhm@ deraadt@
|
|
Rename _waitpid() to safewaitpid() to avoid POSIX reserved identifier
namespace.
KNF nit: return value expressions should be surrounded by parentheses,
per style(9).
Ensure SIGCHLD is set to SIG_DFL, not SIG_IGN. POSIX allows (and
requires under XSI) that terminated child processes not leave zombies
if SIGCHLD is set to SIG_IGN, and it also allows execve() to leave
SIGCHLD set to SIG_IGN.
|
|
OS X's sigaltstack() fails with ENOMEM if ss_size < MINSIGSTKSZ even
if SS_DISABLE is specified in ss_flags. Rather than add code to try
to cope with this stupidity, just don't bother restoring the original
signal stack.
|
|
filtering rules. Some of the filenames are based on the old grammar,
we keep them for now and just update the contents to have a
before/after comparison of the features.
Most of the work was done by Andre de Oliveira (andre@)
OK bluhm@ benno@
|
|
ok jsing@
|
|
OS X fails to restore the old signal stack because the signal stack is
not enabled by default. This causes sigaltstack(2) to fail with ENOMEM
as ss_size is 0, < MINSIGSTCKSZ.
ok jsing@
|
|
This avoids compiler warnings for always true/false conditionals and makes
the code readable. Also avoid pulling in modes_lcl.h, which should not be
used outside the library.
|
|
ok beck@ jsing@
|
|
ok jsing@
|
|
ok miod@
|
|
ok beck@ jsing@
|
|
Hierarchical naming and mention macro names in messages.
|
