from removing them (so that we do less delete/re-add cycles, the more
cores you have the more relevant this becomes).
again, found on Yandex machines, thx to 12 cores or more.
at the end of fake
- sort available cores so that if all hosts fire at the same time, the
load will be fair (found out on yandex machines)
- new iamroot property: for chroot distant hosts, it's simpler to connect
as root, then you can directly chroot -u user and not need the extra sudo
stage.
- more specific "walk_same_host_jobs" to use to signal jobs, for instance.
stylistically, I tend to prefer ``, but since we use a lot of fragments,
the parenthetical nature of $() is better.
- give a more complete name to portstree-/inst- depends lists, so that several
"make package" can run in the same cache.
- create temp dependency files atomically by using mktemp/chmod/mv
(global cache)
(to be referenced by documentation, to stop confusing crypto signatures
with other signatures).
(old name kept over release, to be scraped completely afterwards, user-visible
change only)
okay sthen@
user improvement:
with an extra -d, also scan the directories for "current" pkgnames,
and display a ! in pkgnames that are current while reporting conflicts.
ports. Given that you import port once but update it a few times, make
the update mode (-CU) default for portcheck, and provide a "new port"
switch (-N) instead.
This commit updates portcheck(1) and portimport(1): code, documentation and
tests. The WWW stuff will be updated separately.
Note: regression suite fails now as it detected an actual misconsistence
in portcheck's output. This will be handled ASAP, too.
Initial prodding by naddy@
you cite in email, but noone stops you to !fmt the text. This removes the
corresponding parameter from check_long_lines() because it's not needed
anymore.
inspired in chat by bcallah@, idea okayed by sthen@, espie@ and naddy@
usually added by corresponding MODULE, which, of course, does his work
right. So catch common "kdeui" and "kio" libraries, too. Already run into
this a few times.
because of missing depends. Keep in mind that build_path_list has
to be extensive to avoid LISTING errors...
So, if we meet a stub there, we should do our best to avoid it.
Some of these checks are probably slightly overkill.
problem noticed by landry@
RUN_DEPENDS =
...
RUN_DEPENDS =
This won't catch such problems if they are spread in different files, but
should be helpful anyway now: idea from my recent typo in x11/kde4/runtime.
problems, and make some other automated checks ineffective. And we
try to have are ports more declarative than imperative, anyway.
The idea was supported by espie@ a while ago.
