serf 1.0.0 released

[Blair Zajac]

Hi all,
I'm pleased to finally announce the 1.0.0 release of serf!
The serf 1.0.x series is intended to support Apache Subversion 1.6.x
and the upcoming 1.7.x for their supported lifetimes.

Nice!
So any version of Subversion 1.6.x, assuming it is recompiled, will
work with serf 1.0.x? Or do we need to wait for 1.6.18?

I honestly don't know. I haven't investigated that, although I believe
others have.

Just tried, svn 1.6.x doesn't compile against serf 1.0.0 due to
configure looking for serf-0 instead of serf-1. If I s/serf-0/serf-1/
g in configure, than it doesn't compile due to the removal of
serf_bucket_snapshot().

Given that serf-0 and serf-1 can be simultaneously installed, it
doesn't appear necessary for us to make 1.6.x work with serf 1.0.x,
but it may be nice for some distributions. For MacPorts, it looks
like we'll just provide svn 1.7.x and let Mac OS X provide 1.6.x

One other thing. With 0.7.2, make check passed all tests, but 1.0.0
fails 7 tests on Mac OS X 10.6 against MacPorts' provided apr:

== Running test_all ==
FFFFFFF..........

There were 7 failures:
1) test_serf_connection_request_create: test/test_context.c:166:
expected <0> but was <47>
2) test_serf_connection_priority_request_create: test/test_context.c:
265: expected <0> but was <47>
3) test_serf_closed_connection: test/test_context.c:404: expected <0>
but was <47>
4) test_serf_setup_proxy: test/test_context.c:505: expected <0> but
was <47>
5) test_keepalive_limit_one_by_one: test/test_context.c:656: expected
<0> but was <47>
6) test_keepalive_limit_one_by_one_and_burst: test/test_context.c:810:
expected <0> but was <47>
7) test_serf_progress_callback: test/test_context.c:932: expected <0>
but was <47>

Is there anything different I should do, or should the unit test just
work?

Blair

[Peter Samuelson]

[Blair Zajac]

Post by Blair Zajac
One other thing. With 0.7.2, make check passed all tests, but 1.0.0
== Running test_all ==
FFFFFFF..........
expected <0> but was <47>

This is serf issue 78, http://code.google.com/p/serf/issues/detail?id=78.

I've seen 3 variations, 3 distinct apr errors. You get 47, I've also
seen 22 and 270002. From what I can tell, they map to:

47, SOCEAFNOSUPPORT "Address family not supported by protocol family"
22, SOCEINVAL "Invalid argument"
670002, EAI_NONAME "getaddrinfo: Name or service is unknown"

This seems to have something to do with the 'localhost' test using an
IPv6 socket, when 'localhost' does not map to an IPv6 address (should
be ::1). But, given the 3 distinct errors I've seen, it's not clear to
me that it's just one bug, there may be 2 or 3 related bugs here.