as disable socket.test in this case. Some of the TCL's self-tests attempt to
open a socket to (or request http-data from) a "dead" port on the localhost.
These tests expect a quick failure, but get a hang, when the blackhole is on.
Tests in http.test use a timeout, but some of those in socket.test do not...
Pointed out by: confusion on ports@
. add recognition of the BSD-specific SIGINFO signal (to be
submitted to authors);
. avoid even extracting the compat/ subdirectory for fear
of picking up a wrong declaration for a function defined
in a system library;
. force to use the system strstr() instead of building TCL's
own (from compat/strstr.c);
. similarly, use the system memmove() instead of the (identical)
bcopy;
. run the vendor's tests as part of the build -- the test failures
are not fatal, but may help investigate problems;
. fix up the two failing tests so that none currently fail on my two
test systems (FreeBSD-6.1-STABLE i386 and amd64) (fixes submitted
to authors);
. quiet down the noisy warning, triggered in the thread-case by
the re-#define of inet_ntoa() in tclUnixPort.h.
Bump PORTREVISION.
patch-tclUnixChan.c appears to have been a no-op all along, since
it is comparing literal device names, and we have no "/dev/cua".
patch-tclUnixSock.c was introduced on the basis that the nodename
returned by uname(3) was limited to 32 characters, which does not
appear to be the case now.
PORTREVISION not bumped since there should be no functional change.
PR: ports/92334
Submitted by: Donal Fellows <donal.k.fellows@manchester.ac.uk>
(Tcl Core Team)
Tcl sources to be available at Tk build-time. Only a few headers
are actually needed and they are installed by Tcl port under
$LOCALBASE/include/tcl8.x
FreeBSD only provides the __fpsetreg() used by the following two
for the GNU Compiler. When using, say, Intel's icc they break.
Submitted by: root@celsius.bychok.com
