Commit Graph

247 Commits

Author SHA1 Message Date
espie
6b8c2c3025 typo 2010-03-20 18:30:04 +00:00
espie
74539029e0 changes that I worked on after lock.
- avoid logging incomplete builds.
- put stuff in more sane directories
- avoid exiting early
- create unique machine objects and use them
- allow for a script at init/auto ncpu determination
- global build size in /tmp for a given host.
- better terminal handling.
2010-03-20 18:29:18 +00:00
espie
40399089e7 actual timeouts 2010-03-05 09:05:46 +00:00
espie
198d43e5ad forgot the manpage 2010-03-05 09:05:30 +00:00
espie
a3ca1adf1f simpler fix: do the \n printing always for glitching terminals 2010-03-05 07:50:15 +00:00
espie
a14f2a2255 fix newline handling. thanks naddy@ for the hint 2010-03-04 20:48:12 +00:00
espie
609847a7d3 slightly more evolved: run du twice,to get build size vs. fake size.
Yep, I should be able to use this information later on.
2010-03-04 14:23:01 +00:00
espie
6ccdda1670 log signature errors
move the logs/locks yet again: those are dpb logs, so they belong in
logs dir (note the s). Use one single option (-L) to move both.
remove old -t mode, use -t/-T for ssh/display timeouts
Change the "mem" option to "parse size file", dependent on show-size.
2010-03-04 13:56:09 +00:00
espie
ff14b4c7c3 pass host names around for comparison messages 2010-03-04 13:51:48 +00:00
espie
6fdf400298 add computation of workdir size on -s (assumes a patch to bsd.port.mk
that's not yet there)
2010-03-02 18:20:45 +00:00
espie
d2abbd61aa better control logic.
prepare to du port dir.
2010-03-02 02:33:15 +00:00
espie
9a2f0331f4 set random behavior after we parse the option, so this is actually more
random... problem noticed by naddy@
2010-03-01 18:11:11 +00:00
espie
a8e8acea0d fix for non-terminal 2010-03-01 18:01:11 +00:00
espie
61b975e3b2 tweak logic of run slightly, so that we actually check stuff in finalize,
as should be. This also simplifies infinite jobs slightly, since we build
the tasks we want directly
2010-03-01 17:59:49 +00:00
espie
a58639ad7a if make clean isn't enough, do it again with sudo, but log it.
(also, bug in notime, really don't take into account what we shouldn't)
2010-03-01 17:57:25 +00:00
espie
53d76dd037 less flickering: compute what to display, and do it at once 2010-02-28 11:49:45 +00:00
espie
f71d67db02 don't stop at the first error, report all 2010-02-27 10:05:42 +00:00
espie
0118b624ce let myprint work with -x 2010-02-27 09:53:44 +00:00
espie
d17585f86b activate library signature code. now we check that machines DO match 2010-02-27 09:53:09 +00:00
espie
c8942d3783 fix watch: if we use the offset too soon, it changes all the time, and we
never notice
2010-02-27 09:28:47 +00:00
espie
abbba7caf8 oops, fetch should not be accounted in the log either. 2010-02-27 08:30:44 +00:00
espie
81cdf2d3a4 do something sensible if run without any options 2010-02-27 08:30:01 +00:00
espie
a435ac28e6 a few improvemennts:
- give more responsibility to tasks, so that they can time themselves without
any higher level support
- record errors the opposite way, so that more recent errors show up first
- better terminal display: use fullwidth, clamp to terminal height
- less logsize fuzz, always trim preparation stages
- dpb's pid - dependent ssh socket names, so that 2 dpb can talk to the same machine.
- pid display for dpb itself, and ssh masters.
2010-02-26 12:14:57 +00:00
espie
a0aa57c2e3 being clueless shouldn't fork-bomb.
featuring tedu@ as crash-test dummy ;-)
2010-02-26 12:11:55 +00:00
espie
61d822e8d5 work-in-progress: new distributed ports builder.
will replace old dpb once we're satisfied it works.
some features:
- no more waiting, starts building right away, the dependencies discovery
process is just another job,
- monitor display that makes it easy to know when jobs get stuck, or machines
get down.
- loads of logs,
- locks that stay around in case of error, and that you can remove when
the error has been fixed (dpb picks it up),
- smart choice of which package to build,
- can take into account previous build times,
- builds on machines of differing speed, offloading "lighter" jobs to slower
boxes.

(again, ports is still locked for most people)
2010-02-24 11:33:31 +00:00
espie
8461975175 band-aid for out-of-date, as noticed by various people
(there are probably nicer ways to do this now)
2010-02-16 20:50:32 +00:00
espie
87a403339e adapt to new api 2010-01-02 12:54:22 +00:00
steven
218487feca spacing 2009-11-15 22:05:36 +00:00
steven
d343c92dff look for .la file associated with -lfoo argument in the cwd too
seems to fix a handful of ports
2009-11-15 21:51:59 +00:00
espie
27a3d543cb a few tweaks I should have committed long ago.
- better at handling progress meter incomplete lines and sanitizing them
before output
- less greedy about maintaining a context, let post-error messages stuff
end up in the default.log, instead of seeing huge swaths of entering directory
after errors.
- -s silent option, for use when we see the output but don't need to duplicate
it.

After work with NicM, portslogger is useable with tmux. From within tmux,
just run something like:
tmux pipe-pane 'perl /usr/ports/infrastructure/build/portslogger -s /usr/ports/log'

to log all ports building activity...
2009-11-08 09:24:11 +00:00
jasper
a296057f8c - add a -d flag to show the directory in which the outdated port resides
original diff by giovanni@, reworked by me to not alter default output
ok giovanni@ sthen@
2009-10-22 08:44:44 +00:00
steven
685bd523ac cleanup unused variable 2009-10-18 08:32:19 +00:00
steven
16d0c32f3c separate lists of static and shared libraries
this makes xenocara build.
2009-10-17 14:40:22 +00:00
steven
84737a0d1c clean up a bit, and save some time 2009-10-15 15:20:34 +00:00
steven
de328a6e74 better regex 2009-10-15 13:48:11 +00:00
steven
3628fb6ff2 try to cope with people passing /path/libfoo.so.X.Y instead of -lfoo
if a static library has been found, don't search for it again.
2009-10-15 13:46:23 +00:00
steven
3197dd1add no need for fiddling with deplibs again here 2009-10-15 11:26:08 +00:00
steven
90acfab4e5 revert some of the previous; we need this to deal with convenience libs 2009-10-15 11:16:10 +00:00
phessler
3c719e581d allow us to run make update-patches when WRKDIST has spaces in it.
OK espie@
2009-10-15 10:48:05 +00:00
steven
d514f1eb1d remove unneeded extract_archive function 2009-10-15 09:39:45 +00:00
steven
7b6089698d cleanup; this seems to work better 2009-10-14 22:03:16 +00:00
steven
1adbca4d86 shut up if not in debug mode 2009-10-14 18:22:03 +00:00
steven
24132ef048 use find method here as well, as the object data might not have been filled 2009-10-14 16:03:02 +00:00
steven
869d85c921 make guess_implicit_mode work again 2009-10-14 14:28:22 +00:00
steven
dd74eb33f1 register directory where static lib can be found
add missing newline
2009-10-14 14:14:21 +00:00
steven
cb183eb8c6 run find unconditionally 2009-10-14 14:05:33 +00:00
steven
fa7822ef2f sigh, install programs do not always take options 2009-10-14 13:42:41 +00:00
steven
faaa9351fc if (defined $a) instead of (if $a) for the -release option, as it may be 0 2009-10-14 12:28:44 +00:00
steven
38fc564f05 move some library finding code out of the parser
store the .la file path in the library object for later use
fix a bug where we lost some of the command line arguments
this makes more stuff build again and improves static linking
2009-10-14 11:31:50 +00:00
steven
e6877d2f2a read architecture from the perl configuration 2009-10-14 09:06:17 +00:00