was not only inefficient (writing out to a file and then reading back!),
but also had a bug in its handling of va_args.
The most obvious ill-effect was the daemon crashing shortly after
start-up on amd64. The inefficientcy was less obvious, but still there
on all platforms.
Although log.c contained HAVE_VSNPRINTF ifdef-s, the configure script
does not check for the function's presence and never sets the define.
Add the -DHAVE_VSNPRINTF as a simple work-around... Bump PORTREVISION.
Approved by: maintainer
While here, ditch the ancient socks5.sh.sample in favor of the proper
USE_RC_SUBR script based on the maintainer-supplied code.
also doesn't require manual downloading, so remove the code to prompt
for that.
Note that although the checksum has changed, there are no differences
in the distfile save for the compression. I compared the uncompressed
tarball against one from http://www.etn.yz.yamagata-u.ac.jp/ipv6arc/socks/
which corresponded to the old checksum and they were identical (md5).
I chose to use the SGI MASTER_SITE instead since it would presumably
have more bandwidth.
Approved by: Josh Paetzel <josh@tcbug.org> (MAINTAINER)
just as in other ports which require "manual" downloading of distfiles.
(also correcting the extranous '.')
PR: 26719
Submitted by: Dimitry Andric <dim@xs4all.nl>
A diff of the original .tgz file and the new one shows that they
reformatted their copyright notice, added the year 2000 to it, removed
one file that had an identical copy of the copyright notice and
license and took the opportunity to slightly reformat the copyright
notice and license. No code changes were in evidence by this change.
Approved by: Mike Tancsa <mike@sentex.net>
zombies on the system. Without this change, the socks5 daemon would
work great for a short period of time, then hang because no more
children could be created. It appears from debugging that this is
because gravedigger was only called once. I reestablish the signal
handler in gravedigger to force SIGCHLD to be unmasked.
A more proper fix might be to fix Signal to have multi-shot signals,
or to see what might be broken with the signal masks. Since I'm in a
hurry and this port's daemon is broken without this, I commit this
kludge knowing that it will live approximately forever, less three
days.
r11 as well as parts of patch-ak. 1.0r10 is no longer available at
the NEC web site. Verified the URL was still the right place to go to
get socks5 1.0r11.
Noticed by: Jim Paterson <jpaterso@paterson.org>
previous commit message to bsd.port.mk, which said INSTALL_SHLIBS. Boo.)
Line up the rhs of variable assignments nicely. Remove a couple of extra
whitespaces while I'm here.
Suggested by: sobomax
Remove patches integrated since r9.
1/2 of patch-aa was integrated, the second half has been retained.
Add new patch (patch-ak) from web site for s5fakehost files.
Add new patch (patch-al) to fix long standing telnet failure to build (it
is a kludge, so hasn't been submitted to the socks maintainers).
The deleted patches were merged into the program.
Obtained from: http://www.socks.nec.com/patch-socks5.html
(the new patches, that is)
NOTE: I'd love to have used PATCH_SITES and its cohorts,
however the patches from the web site were directory-ambigous
and adding them to our repository is the best way to avoid a
lot of glue and uglyness in the Makefile.
Patches-ad and ae from
Andre Albsmeier <andre.albsmeier@mchp.siemens.de>
I've not had a chance to test this as much as I'd like, but it has
become obvious to me that a possibly minorly broken port is better
than a definitely broken port.
Thanks to obrian for the gentile reminder...
FreeBSD 2.x dynamic loader. This makes more programs, including emacs,
work with runsocks.
This patch will work at least as far back as 1.0r3.
Submitted by: Aaron Smith <aaron@sigma.veritas.com>
