FWIW, checkout of these things took 5+hrs, staying on the local
.freebsd.org net w/o hitting the 'net at all.
As promised,
$ time cvs ci
real 67m51.701s
user 0m1.250s
sys 0m5.345s
${MACHINE_ARCH}--freebsd${OSREL} is now passed to CONFIGURE_ARGS if
GNU_CONFIGURE is defined. Take the target out of CONFIGURE_ARGS of
some ports that added it explicitly; define it as
${MACHINE_ARCH}--freebsd if the port doesn't like the ${OSREL} part;
define it as something else (such as ${MACHINE_ARCH}--freebsdelf if
the port requires that; define it as an empty string if the port
doesn't like it at all.
The last might be a sign that a GNU_CONFIGURE port actually doesn't
use GNU's version of configure at all; but I don't have time to go
look at them all, we'll fix them as time goes on.
At least we've got much fewer "-unknown-"s in the tree as the result. :)
bsd.port.mk rev. 1.304 for details on the change.
The fix here is one of the following.
(1) Define USE_BZIP2 instead of BUILD_DEPENDS on bzip2 and redefining
EXTRACT_* commands.
(2) Change ${EXTRACT_CMD} to ${TAR} when the command is obviously
calling the "tar" command (i.e., arguments like "-xzf" are spelled
out).
(3) If ${EXTRACT_CMD} is called directly with ${EXTRACT_BEFORE_ARGS},
add ${EXTRACT_AFTER_ARGS} to the command line as well.
(4) If any of EXTRACT_CMD, EXTRACT_BEFORE_ARGS or EXTRACT_AFTER_ARGS
is set, define the other two too.
to fix it today. Asked Thomas (Gellekum) for help ...
Only this friday to fix it. The patching stage fails ...
The mess started with the new organization starting from
the egcs source base ...
- Changed PATCH_SITES, added some more mirrors and now use ftp sites
- Autoconf is needed in FETCH_DEPENDS, since I have to re-create
the configure script just before the pre-configure target
- pre-configure target now does the suggested maintainer-clean
- post-install renames c++, g++, gcc to pc++, pg++ and pgcc
to avoid conflicting usage with the systems default compiler.
- updated PLIST
Please note, that I left out the very last diff, because after
that configure doesn't work properly anymore.
PR:
Reviewed by:
Submitted by:
Obtained from:
want higher optimization levels result in unsafer binaries).
Fix bug, that causes core dumps when using higher optimization levels
than -O2.
Results with dhrystone 2.1 (1000000 loops):
stock cc (2.7.2.1 from current) : 136k (136xxx)
pgcc-current with -O2 : 138k
pgcc-current with -O6 : 172k
And no coredump anymore. Reason for coredump:
With -O3 cc1 does an register-to-register-copy optimization and it tries
to write some debugging information about what it is doing into a file
that does not exist unless you have instructed cc1 to dump all its
information about its internal state. (From: Sascha Blank)
BTW: PLIST is really coming soon, it's on my todo list ;-)
BTW2: Is Sascha already in the contributors list ? Have to go to work now.
Submitted by: Sascha Blank <blank@sliphost37.uni-trier.de>
optimizations. The motivation to make a separate pgcc port is, to
provide two versions, an official release, which is missing the
leading edge of developement and an interesting current version,
which fixes some things and might breake (or not) others.
Fact is, that one is able to compile and run a complete FreeBSD-current
kernel without trouble and so closes 2 PR's.
>Number: 1221
>Category: bino
and possibly
>Number: 1222
>Category: ports (should be tested by the submitter.
