(Sorry Jordan, but your other idea (${CATEGORIES}) was a major hit.)
Also remove the keyword field in the INDEX line and replace it with
two columns: build-time dependencies and run-time dependencies. They
are both list of package names (minus the ".tgz").
(1) people can have weird paths and it will still work, and
(2) if you really need to use /usr/local/bin/cp instead of /bin/cp,
you can do that by changing only one line.
Submitted by: wosch
/var/db/pkg/${PKGNAME} exists. ("make install" will do nothing
because this is not a critical error and the installation is
treated as successful.)
Closes PR 870.
prototypes don't go missing again. Also added -Winline so that some
doubtful (non-)inlines get fixed.
bsd.kmod.mk:
Also added `-Wreturn-type -Wimplicit -Wnested-externs' to catch up
with the kernel.
1. ${ROFF} is run in ${.OBJDIR}.
2. the preprocessor prefixes ${SRCDIR}/ to relative pathnames in `.so'
statements.
This is useful when running ${ROFF} in the source directory isn't
convenient.
Added dependencies on ${EXTRA} and ${OBJS}. These are usually for files
that are sourced indirectly. ${OBJS} is for files that are built.
4.4lite has decentalized incomplete dependencies on ${EXTRA} and ${DPADD}.
These were broken by are centralized handling of the roff targets.
too. Basically, if the name starts with a "/", it's tested with "test -e";
otherwise, it's tested with "witch -s".
Reviewed by: the ports list (well at least nobody complained)
misplaced extern declarations (mostly prototypes of interrupt handlers)
that this exposed. The prototypes should be moved back to the driver
sources when the functions are staticalized.
Added idempotency guards to <machine/conf.h>. "ioconf.h" can't be
included when building LKMs so define a wart in bsd.kmod.mk to help
guard against including it.
failed when there was an obj directory. Use .PATH.n for installing too
so that make can find the source files. This allows the source files to
be in several directories (the old method using cd only works well for
a single directory). The dependencies are on the source files even for
the compressed case, although it would be more flexible to depend on the
files being installed, so that `make install' doesn't attempt to build
things
Force COPY to -c for the NOMANCOMPRESS case. Then the files to be installed
are always sources, so they must not be moved.
use .PATH.n to get the dependencies right and to avoid some shell tests.
Remove bogus dependency of individual compressed man pages on MANDEPEND.
Use for loops to avoid duplicated code.
Combine some rm steps in installation of links. Linking still takes too
long.
They are called from the fetch, extract and install targets,
respectively.
Also, only RUN_DEPENDS is put into the @pkgdep list of the package.
EXEC_DEPENDS is still supported (for now), it is copied into
BUILD_DEPENDS and RUN_DEPENDS. This will go away after we finish
fixing all the ports Makefiles.
This change fixes the following bug/features:
(1) "make fetch" building and installing all the dependencies
(2) Programs needed for building only (e.g., gmake) put into the
packages' dependency lists ("why does the emacs package depend on
gmake?")
Reviewed by: the ports list
FORMATS can be defined as an empty string to suppress generation or
installation of any files. Previously setting it to "null" had that
effects.
Now uses MANOWN, MANGRP and MANMODE for installation instead of BINOWN,
BINGRP and BINMODE.
the file is fetched or not. Apparently Jordan fixed it a long time
ago but it was broken again at import of the new version of ncftp.
Which means even if we fix it, it may break again and we may need to
fix it again, and (imagination here, please)....
Instead, move the file existence check into the for loop for
MASTER_SITES/PATCH_SITES and break out with "continue 2" when the file
is found. This is actually a cleaner logic than before if you ask me,
because instead of assuming the file is fetched on a 0 exit status
from ncftp AND checking for the existence of the file after the loop,
the check is done exactly once for each iteration and nowhere else.
do-extract target depending on defined(EXTRACT_ONLY) or not, simply
set EXTRACT_ONLY ?= ${DISTFILES} and always use ${EXTRACT_ONLY} as
the extraction list.
change, but I've been testing this on thud and silvia for quite a
while, also I haven't gotten any bug reports from the ports list, so
I'm going to let it loose!
It cleans up this file quite a bit, now I can go in and start adding
some more "interesting" things.... ;)
of replacing it. This way you can point it to a site close to you
that carries many distfiles, and still let it go fetch from the
original site if the distfile is not there.
Original idea by: mmead@Glock.COM
to do something else than "install". For example,
make IS_DEPENDED_TARGET=fetch fetch
will fetch the required distfiles including those of the dependencies
without actually building and installing dependencies.
Also document ECHO_MSG.
Requested by: paul
Reviewed by: paul, jhs and others
ftp://ftp.freebsd.org/pub/FreeBSD/distfiles/
as our distribution point for distfiles and patches. Other than
cosmetic changes (freebsd.cdrom.com -> ftp.freebsd.org), the
omission of "ports" is important. I would like to move this
directory completely out of the ports tree (on the ftp site),
so that people who do "get ports.tar.gz" won't get a bogus distfiles
-> ../distfiles symlink (which will make "make fetch" fail).
Sometime around the 2.1 release, the distfiles link will be deleted.
set permissions and ownerships of PREFIX (usually /usr/local). This
is the default if USE_IMAKE or USE_X11 is set.
This should be useful for machines like thud, where we want to keep
the /usr/local subtree writable to a group ("ports" in our case). Anybody
who installs stuff in /usr/local should have this set in the environment.
Note this won't affect anything the pkg_* suite does.
Note that the two "touch"s I took out from do-patch shouldn't have
been there in the first place.
This target may give incorrent results if two separate patches deal
with the same file, and their hunks overlap. (But having those kinds
of patches are bad, and they should be merged anyway.)
Reviewed by: hsu
".../packages/All". The "all" category that was automatically added
for every package is gone.
Note that bsd.port.mk requires category names to start with lowercase
names, otherwise it may get confused.
Reviewed by: jkh
By the way, here is a small script to convert your local package
hierarchy. Run it in bash, as /bin/sh not only will bark at the
$(.) command substitution but will also botch the [a-z]*/*.tgz
expansion (long-standing and annoying bug, reported before).
cd /usr/ports/packages
mv .packages All
for i in [a-z]*/*.tgz; do
j=$(basename $i)
/bin/rm $i
ln -s ../All/$j $i
done
one of the key components of the system, but I'm sure that this:
===
- ${ECHO_MSG} "===> Registering installation for ${PKGNAME}"; \
+ ${ECHO_MSG} "===> Registering installation for ${PKGNAME}"; \
===
change has absolutely no chance to screw us up, right? :)
Ports for which we can't build packages should define NO_PACKAGE but
still prepare pkg/* files. The user who really wants a package and
clear of the legal problems can say FORCE_PACKAGE from the command line
to build a package anyway.
package: check installation, build package, create links,
touch cookie
repackage: ditto but don't check cookie
package-noinstall: just build package from installed stuff, no cookies
involved at all
package-links create the symbolic links only
delete-package: delete package and symbolic links
delete-package-links: delete links only
These should make the management of the spaghetti of package links
a little friendlier. :)
too (otherwise the chain won't work).
(2) If NO_WRKDIR is set, "make clean" removes "./.*_done" (assuming
these are cookies...or should I list all the cookies?)
build, install) are now all skeletons and do nothing but
(1) Call pre-* target (if exists)
(2) Call scripts/pre-* script (if exists)
(3) Call do-* target
(4) Call post-* target (if exists)
(5) Call scripts/post-* script (if exists)
The do-* targets do all the work. The pre-* and post-* targets/scripts
don't exist by default. The main targets check for the cookies too, so
porters shouldn't have to worry about them at all.
NOTE: THE MAIN TARGETS IN THE PORTS MAKEFILES SHOULD GO AWAY. We need
to fix this before wcarchive comes back up. Change the names to do-*,
rip out the cookies, rip out the calls to pre-* etc. and most of them
should work.
Also, reorganize the whole thing so that similar targets are together
and add more comments. Surround section header with 64 #'s (C-u C-u
C-u # in emacs :).
Hopefully this will be the last major change to bsd.port.mk. Now let
the Makefile-hacking begin.
rule.
2. Have all non-X11 prefix using packages include the BSD.local.dist mtree
file for initialization of /usr/local. I'm still not sure if this is
A Good Thing(tm) but I'll see what the users say. It's easily overridden.
3. Standardise on ${PKG_DBDIR} as pointer to /var/db/pkg or local preference.
"Building for WWW" (pops up in two different ports) "Installing for
web2c-6.1" (ditto), which aren even't reminiscent of the port's real
name.
Sorry jmz, please don't go fix the print Makefiles' own messages.
We are going to take them out after we do the great bsd.port.mk
update anyway.
the top level and have the build-package sequence of each port work
together.
For the old behavior (i.e, just go ahead and blindly pack everything up,
regardless of the contents of work/), there is a new target "repackage".
Since "build" depends on "configure", which depends on "patch", etc.,
this shouldn't disrupt any Makefile that doesn't break the dependency
chain.
The old behavior was very annoying because when I did a "make -k",
it would still try to go configure and build even if the extraction
failed.
all .tgz files go to /usr/ports/packages/.packages, and a relative
symlink is created for every item in CATEGORIES...i.e., if "CATEGORIES
= foo bar", then /usr/ports/packages/{foo,bar}/pkgname.tgz both point
to /usr/ports/packages/.packages/pkgname.tgz.
Suggested by: jkh
*Really* strip out unused local symbols from shared objects.
This was a typo on my part caused by an assumption that the profiled
libraries stripped symbols that same way as the non-profiled libraries.
Cut-n-Paste strikes again.
Obtained from: NetBSD
New variables:
PATCH_SITES: patch equivalent of MASTER_SITES, overridable with
. MASTER_SITE_OVERRIDE.
PATCHFILES: Additional files to fetch and give to patch before
. applying the ones in patches/patch-*. If name ends
. with ".gz" or ".Z", it will be piped through zcat first.
Plus PATCH_DIST_STRIP and PATCH_DIST_ARGS that serve the same functions
as PATCH_STRIP and PATCH_ARGS for patches in patches/patch-*.
In the documentation and echo messages, I used the term "distributed
patches" and "FreeBSD patches" to refer to ${PATCHFILES} and patches/patch-*.
If you can come up with better names, by all means go ahead and fix them.
"grep PATCH /usr/ports/*/*/Makefile" reveals seven ports (mule, jless,
jtcl, jtk, dgd, less, color_xterm, gee I wonder why I'm the one who
implemented this) that can benefit from this. I'm now diving headlong
into /usr/ports to fix their Makefiles.
