Commit Graph

1106 Commits

Author SHA1 Message Date
espie
9fd4058fbf also tweak the lock names to end in .lock so that they can't get confused
with core dumps ;)
2023-01-30 18:18:41 +00:00
espie
d04a59d00e fix typo 2023-01-25 13:53:48 +00:00
espie
56e460b091 be more thorough in flagging bad stat lines.
doesn't happen to me, but I don't break my file systems very often
as noticed by tb@
2022-11-06 09:29:12 +00:00
espie
eab8549545 remove the badly named, undocumented alias for "print-update-signature" 2022-11-01 10:55:54 +00:00
kmos
45837f04bf Tweak license handling to handle more common licenses or variations
of the same licenses.

ok giovanni@
2022-10-15 00:11:56 +00:00
espie
2ca8bcbc63 refactor by adding an extra "factory" class in order not to overload
things horribly. No actual functional change
2022-09-11 08:40:40 +00:00
espie
ee19561663 do something slightly horrible to match base through plistreader subclasses 2022-09-11 08:27:21 +00:00
kirby
4b25157323 Stop adding RCS ID tag on the top of Makefile.
looks right to afresh@
2022-06-18 19:32:19 +00:00
espie
5f4d472f4e tweak special behavior for quirks: instead of putting it aside on normal
build, use a new DPB_PROPERTIES "later" that informs the Grabber to build
it during a second stage (with the exact same rules as before).

And thus, reconnect quirks into the "normal" build.

Main difference: other databases like sqlports or pkglocatedb will index
quirks again
2022-06-07 16:13:16 +00:00
espie
8c5ddca555 bump the polling to a lot more (120s) to give realistic time to enter
a passphrase, get it wrong, and try it again
2022-06-03 07:45:37 +00:00
espie
504904a5fe allow the ssh master to start interactively, so that it's even possible
to run without ssh keys or with a passphrase.

allow up to 10 seconds for startup so that the prompts do not
overlap each other (after that, the host will be dead as usual).

paraphrased from naddy@ "it's unusual to allow root ssh with a password,
but that means that problems with the configuration will show up straight
away" which is a good thing.
2022-06-03 07:44:33 +00:00
espie
7bac54f790 force the creation of the ssh master from the distant host creation.
as notice by naddy@, "just in time" master creation means the ssh master
was created from deep within Core/Init.pm (for the whoami task basically),
and actually *after* forking, so the parent wouldn't know about it, and
create a new master from the main process.
2022-06-02 08:52:50 +00:00
espie
4c9bfe12fb tweak the logic of building quirks last slightly: if we built any package
during the current run, we rm any quirks packages that could be left over
from a previous build first.

This ensures quirks gets rebuilt if some packages did change, so that
the updateinfo cache is as fresh as possible.
2022-04-14 19:39:37 +00:00
espie
9b2a2be4f9 let recheck_errors return the value it's supposed to. This was broken
5 years ago, also: add a sleep to avoid dpb busy-polling on locks in
case there is nothing else to do.
2022-03-15 14:33:29 +00:00
espie
b091283ada fix refactoring botch introduced in rev 1.60: check_path does the right
sanity check for the build_path_list.
The extra is_stub check is non-sensical, since check_path does a full
(correct) check.

Sorry for breaking www/squid

Note that missing info for pkgpaths end up at the end of equiv.log
2022-03-11 18:38:52 +00:00
espie
45f0a735ce slightly cleaner: attach stuff to the CVSTag, BUT don't copy it since we
won't do anything with it.
2022-03-07 16:34:26 +00:00
espie
e4b3169766 Final fix (hopefully) for newuser and newgroup not copying over as reported
by aja@ and landry@

My error was that I did think I needed a placeholder in the new list to
tack new stuff to, whereas I actually need the placeholder in the old list,
so that it gets copied along with the stuff that's attached to it...
2022-03-07 08:57:10 +00:00
espie
45e5be99fd kill the remaining scaffolding wrt cvstags 2022-03-01 11:01:17 +00:00
espie
0d435662c6 forgo outputting CVSTags, okay sthen@ 2022-02-28 22:38:30 +00:00
espie
a20dd924fb move summary.log generation into the subengine, because it's absurd
to create it in dpb modes that don't build anything (such as -F and
HISTORY_ONLY). No functional change.
2022-02-17 12:42:37 +00:00
espie
024ce525d7 oops missed one 2022-02-13 17:42:28 +00:00
espie
d7d3228526 automatically add @sample for login class files. 2022-02-13 12:42:46 +00:00
yasuoka
5824a4ab8d Override "build_user" with a command argument if any.
ok espie
2022-01-26 14:28:13 +00:00
espie
30fe5eaa7f glue for recognizing ocaml files...
(misses the fragment dispatch)

generalize "tweak_other_paths"  to run *after* the normal recognition, since
ocaml paths are linked in more generic ways that just "parent directory for
fonts", so they don't necessarily occur in the right order, and we shouldn't
care.
2022-01-21 10:55:18 +00:00
espie
edf8647ae7 slightly better comments 2022-01-21 09:38:48 +00:00
espie
7dfad28b86 move all the logic that handles file dispatch, actual backsubstitution and
state handling into a separate file
2022-01-21 09:36:00 +00:00
sthen
568a72ef8d adjust for CommonPlist.pm renaming 2022-01-20 08:18:47 +00:00
espie
f897e884e8 rename file to a more intuitive name 2022-01-19 14:46:00 +00:00
espie
cf1897620d move a bit of the main code into a different file, because this is starting
to grow a bit too much.
2022-01-19 14:38:48 +00:00
sdk
4ba8953556 Use the File::Copy cp() alias instead of copy(), because it preserves file
permissions. This is closer to system cp -a, which was used before.
2022-01-05 21:23:18 +00:00
sdk
4f3cce6109 Use File::Copy instead of system cp in portgen
ok espie@
2022-01-05 18:28:25 +00:00
espie
c486524600 silence CPAN::Meta::Dependency warning by passing an explicit minimal version
of 0 if none is provided

okay afresh1@
2022-01-05 09:13:40 +00:00
espie
0e522668c5 add -m VAR=value option to pass stuff directly to make.
okay afresh1@
2022-01-05 09:12:50 +00:00
ajacoutot
9b0787094c update-plist: add support for '@tag ibus-write-cache'
ok espie@
2021-11-26 16:47:30 +00:00
espie
f6b4392369 log_same_ts already appends a space 2021-11-15 14:48:17 +00:00
espie
d1fa45d8d6 forgot a wee little tidbit for bad paths in external to NOT break things
oops
2021-11-15 14:39:45 +00:00
espie
c3271c33f7 fix the error control flow so that we don't quit dpb on wrong input on
the control socket.

noticed by solene@
2021-10-05 12:31:34 +00:00
espie
f963059f7a scrape the old ldd code, which hasn't been used since we got rid of a.out 2021-06-21 15:16:52 +00:00
afresh1
d6b4517057 Have portgen(1) strip leading "v" on perl ports that need it
Noticed by Mischa <obsdml () high5 ! nl>
2021-06-13 19:00:06 +00:00
afresh1
ac796829f9 Have portgen(1) use newly set DEPENDS and CONFIGURE_STYLE
If we don't clean, then the upcoming make build won't notice any changes
that we actually added to the Makefile, such as setting the CONFIGURE_STYLE
to "modbuild tiny".

Noticed by Mischa <obsdml () high5 ! nl>
2021-06-13 18:30:50 +00:00
espie
0039b6d7da add support for FETCH_CMD as a config file option/command line override
that will replace whatever comes from mk.conf
2021-05-03 07:16:46 +00:00
espie
133c3cebf5 grab FETCH_CMD value from bsd.port.mk, as discussed and tested with
Dimitri Karamazov

TODO: add Config.pm support for this as well
2021-05-02 06:08:53 +00:00
espie
a6ebec8550 properly set up a watched in the Job::BasePort class instead of doing
it manually in each factory method in the builder.

gets rid of warnings under some install/wipe circumstances

okay naddy@
2021-04-16 20:14:47 +00:00
abieber
f10c854458 Allow an optional @version suffix (@v1.0.0 for example) to be specified when
generating a Go port with 'portgen go'. Patch from Josh Rickmar. Ty Josh!
2021-03-23 13:17:41 +00:00
espie
03db828d59 refactor: make the builder responsible for telling us whether we should
clean at the end of a build.

Add a new global config option: NEVER_CLEAN
so that no path ever gets cleaned (to be used ONLY if you have lots of disk
space).

I've actually done this manually in the past, to test various tools on
lots of objdirs...
2021-03-21 19:17:34 +00:00
gnezdo
c68d711c25 Remove GhcConf packing element as ghc-pkg-recache is no more
These were previously used to automatically add tags to PLIST for
Haskell library ports. Since library ports no longer exist, such
tagging only ever triggers for ghc package where it is now harmful.

ok kili
2021-03-07 19:30:16 +00:00
abieber
e27a7fd632 Fix a bug that was preventing version comparisons from working.
While here try using @v/list first, if that fails use @latest. This should
fix some of the "not being able to see the latest version" issues that
happen when upstreams re-tag.

Tweaks / OK from afresh1
2021-03-03 03:08:30 +00:00
espie
6576f7d21a use the new cleaner interface of pkg_add's state that decouples
continue and window size change for readability
2021-02-26 07:54:11 +00:00
sthen
6311f228e2 Record timings at the bottom of the individual build log (it's just the
same information as in summary.log but I find it useful to have here too0.
ok espie@
2021-02-21 10:48:04 +00:00
abieber
0666597e06 Update Go to 1.16.
OK jsing@
2021-02-20 14:56:34 +00:00