The changes in the patch files are almost the same as in Eugene's PR,
although I arrived at them mostly independently :)
Reformat all patch files now that I use quilt to manage them.
Fix two complaints from portlint: needless use of CFLAGS and differentiation
between NOPORTDOCS and NOPORTEXAMPLES. Thanks, Eugene!
PR: 132358
Submitted by: Eygene Ryabinkin <rea-fbsd@codelabs.ru>
Yes, there's been a new 7.19.3 version for a couple of days now,
but I figure you folks have been waiting for me much too much as it is :(
Apologies all around!
Prompted by: several PR's and more people (and portmgrs) than I dare admit
Fix the (still present in 7.16.3) libssh2 problems, at least so that
cURL works with the libssh2 0.17 in our ports tree.
PR: 114215 (the basic update)
Submitted by: pesho.petrov@gmail.com
- add a LIBSSH2 option for SCP and SFTP support using security/libssh2;
- add a patch from the cURL CVS repository to fix an expired cookie in
test 62.
PR: 109670
Submitted by: Eygene Ryabinkin <rea-fbsd@codelabs.ru>
Bump PORTREVISION of all dependent ports.
Fix the build errors in the few ports that still use the long deprecated,
and now obsoleted, cURL options.
Thanks to everyone who took the time to look over the patch!
Discussed on: -ports
really making OpenSSL support optional. Now OpenSSL and GnuTLS
are on level ground, both controlled by the OPTIONS setting.
Fortunately, it turned out that the curl_multi_socket() API change
does not affect any of the ports that use ftp/curl.
PR: 102871, 103126
Submitted by: pav, vd, Scot Hetzel <swhetzel@gmail.com>
Explicitly disable libidn support so that it is not even probed if libidn
is installed; I will look into enabling it after some more testing.
Testing paid off on: ref4 (lib/select.c), sledge (lib/sendf.c)
For some reason, this segfaults badly on sparc64, or at least on the
only sparc64 machine that I have access to (thanks mike!). I'll deal
with this later; for the present, let the x86 and Alpha users benefit
from the new cURL features and bugfixes :)
PR: 39873
Submitted by: Ying-Chih Kuo <yckuo@yckuo.org>
"reallocate" memory even when the programmer only asks for a resizing
of the region.
Reported by: Steve Ames <steve@energistic.com>
While I'm here, change the way SSL header and library paths are handled
to avoid specifically referencing /usr/lib and /usr/include; while
-STABLE's GCC shrugs this off, GCC 3.1 whines loudly about an explicit
-I/usr/include (and rightly so, IMHO).
