As a consequence, explicitely disable the FAM backend.
This bring a huge improvement compared to what we've been using before
(gio-fam+gamin): file / directory monitoring and notifications are next
to instantaneous without any cpu overhead like it was before.
This means we can also drop the bootstrap FLAVOR of gamin that was
needed because of the cyclic dependency (gamin->glib2->gamin).
also annotate gamin for bootstrap property, and mark haddock,no_deps as
a "real flavor", since it's only used for bootstrap anyways AND significantly
changes the build.
Committing now (along with upcoming atk, pango and gtk+3 updates) during
the hackathon where we have time to fix all the eventual fallout (bulk
running).
ok jasper@ robert@
* use MODPY_ADJ_FILES
* use REGRESS_FLAGS instead of 4 patches...
* use post-install to remove files we don't ship, 2 less patches
* install gtester-report(1) -- it's a python script but we do not
enforce a run dependency on python
PLIST and remove remove everything under the @sample'd directory instead
of the directory itself to prevent a warning from pkg_delete(1) trying
to remove a non existing directory and to help preventing left-over
files and directories.
implement the "intended" SCM_CREDS stack as if we had support for that
in the kernel (by-pass it almost completely).
send/recv a single null byte without creds, but on recv, just do a
getsockopt(SO_PEERCRED) and return that as if it coming from the cmsg.
This works as long as creds are not retreived from an fd which has
already been handed over to a different process via SCM_RIGHTS. It will
probably not be enough in the future but we'll see then.
all this work done by eric@ (thanks!) and tested by myself
Enable support for g_credential*
Fix a couple of warnings.
ok eric@ jasper@
Fixes the following:
* 644465 undefined reference to `_usleep'
* 644552 g_timeout_add_seconds(1, ...) may have a latency...
* 644607 correct internal definition of C_()
