| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
since according to the comment in check/bsd.check.mk, they belong there.
Added a new check for all C and C++ header files to make sure they don't
contain strings like ${prefix} or ${exec_prefix}, which is currently a
problem with sysutils/dbus and has been noticed in PR 35019. This check
is disabled by default since I don't know anything about possible false
positives, but I plan to enable it for PKG_DEVELOPERs after some
testing.
Added two names for hooks that are placed in the configure and in the
build phase. Now the checks look more like becoming something one could
call a framework, sharing a common structure and a documented interface.
|
|
another process. It's a little more verbose now, but it helps
understanding what goes on.
|
|
|
|
|
|
|
|
Default value is kept but now set in mk/defaults/mk.conf.
|
|
don't use mkdir, as the latter uses the group permissions of the
parent directory under BSDs, resulting in group leaks for
USE_DESTDIR=yes. Discussed with agc@.
|
|
|
|
list for bulk builds. (Revert previous on bootstrap; the PKG_PRESERVE
hack didn't work as expected.)
|
|
wrong about some of its results when using implied privileges (e.g., an
account in group Administrators where the directory is writable by Users).
|
|
Assume those two can handle their own dependencies.
|
|
|
|
|
|
is not supported at all by pkgsrc, print an error message in that case.
|
|
|
|
|
|
SETUID_ROOT_PERMS.
|
|
|
|
|
|
the case of modular Xorg and the latter is more specific.
|
|
|
|
the fullly conditionalised bsd.prefs.mk still needs a considerable
amount of time and moving the include guard out kills most of that.
Preliminary tests with renaming bsd.prefs.mk and using the same logic
to point to the real files show 5%-10% reduced run time for
"make show-options" in x11/kdebase3 and no measurable regression in
simpler places. This versions allows incremental conversion of the
use in buildlink3.mk. Some care should be used in mk/ itself, as e.g.
bsd.pkg.mk is included from first level hierachies as well.
|
|
|
|
or [] is better, I have used [ to keep the code shorter. Noticed by wiz.
|
|
overwritten when the package-provided files are (possibly dead)
symlinks. Since on Mac OS X, ls(1) does not return the status I
expected, I investigated a little in the available options of test(1),
and found that the "-h" option should be quite portable. So let's try
that instead.
This fixes PR 34865.
|
|
|
|
mk/defaults/mk.conf.
|
|
doesn't match the *.mk pattern, and that's why it had been excluded up
to now, accidentally.
|
|
For that purpose, BUILD_DEFS_EFFECTS has just been introduced.
|
|
the user to affect how packages are built (BUILD_DEFS) and the effects
that those variables have (BUILD_DEFS_EFFECTS). The latter variable has
been introduced to clearly separate these two issues.
While here, reduced the indentation of the directives a little bit. One
visible change is that "make build-defs-message" will always show the
message when called directly. Before, it had been shown only once, which
makes debugging a bit more difficult.
|
|
|
|
|
|
need to specify INSTALLATION_DIRS itself. Instead, the list is generated
from the static PLIST files. Like for INSTALLATION_DIRS, mtree is
disabled when this variable is set.
|
|
|
|
just ${PREFIX}.
|
|
deinstall. The semantic of deinstall should stay as "remove the
installed version" and the old behaviour force a repackaging.
|
|
are identical. For DESTDIR builds, the package is not installed to
PREFIX as part of the build, so package-install does exactly that after
package is done. Change bin-install to call package-install.
|
|
Patch provided by Hrvoje Habjanic in http://mail-index.netbsd.org/pkgsrc-users/2006/11/02/0003.html
|
|
|
|
sysutils/checkperms.
Regular files are now checked if they needlessly have the executable bit
set.
|
|
|
|
|
|
|
|
|
|
|
|
cvs: ----------------------------------------------------------------------
|
|
|
|
common MAKE_FLAGS instead of overriding them, to align with the new
definition of INSTALL_MAKE_FLAGS.
The packages that override BUILD_MAKE_FLAGS don't have to be changed,
since they prepended MAKE_FLAGS to BUILD_MAKE_FLAGS, except for less
than five cases. In those cases, it may well have been unknown to the
package author that MAKE_INSTALL_FLAGS had been an overriding variable
instead of one whose value is just appended to something else.
Some packages (postgresql, some compilers, Java environments, qcad) are
using code copied from build.mk. These packages will have to be adjusted
to the new variable definitions. To make this change easier, two new
variables BUILD_MAKE_CMD and TEST_MAKE_CMD have been introduced that
contain the complete command that makes up the call to make(1). This
prevents further code duplication.
|
|
passing the lists literally. This avoids calling printf(1) with only one
argument, since POSIX and bash differ in the implementation.
|
|
"make install". That way, packages don't have to say:
INSTALL_MAKE_FLAGS+= ${MAKE_FLAGS} foo=bar,
instead it suffices to say
INSTALL_MAKE_FLAGS+= foo=bar,
which is also more appropriate since multiple files (Makefile,
Makefile.common, options.mk, *.mk) don't need to care about whether
INSTALL_MAKE_FLAGS already includes MAKE_FLAGS or not.
Note: It is no longer possible to override MAKE_FLAGS completely. But
except for some exotic platforms, the default value is empty anyway.
|
