Re: sqlite 3.16.2

2017-01-26 Thread Marc Espie
On Wed, Jan 25, 2017 at 08:27:28PM +0100, Landry Breuil wrote: > Hi, > > firefox 52 requires sqlite 3.15.1 at least; so let's jump to 3.16.2 > directly. Bumped the major to be on the safe side, but nm showed no > differences between libs, but there are always defines lurking around in > headers...

HEADS UP: pkg_add supports installpath

2017-02-06 Thread Marc Espie
just committed, will be in the next series of snapshots. so, basic installs create /etc/installurl now, pkg_add uses it "by default" to create an installpath. pkg.conf(5) takes precedence (user defines are stronger than automatic defines): installpath = anything in pkg.conf will replace the inst

Re: add ports infrastructure support for aarch64

2017-02-07 Thread Marc Espie
On Tue, Feb 07, 2017 at 11:47:51AM +1100, Jonathan Gray wrote: > On Mon, Jan 23, 2017 at 10:14:15AM +, Stuart Henderson wrote: > > On 2017/01/23 11:55, Vadim Zhukov wrote: > > > 2017-01-23 4:04 GMT+03:00 Peter Hessler : > > > > Here are some bits to add arch defines to the ports infrastruture f

Re: add ports infrastructure support for aarch64

2017-02-07 Thread Marc Espie
On Tue, Feb 07, 2017 at 10:02:57AM +, Stuart Henderson wrote: > The simple fix for your issue would be to add a ${BASE_CXXLIB}, stdc++ > on base-is-gcc arches, c++ on base-is-clang ones, and replace all > stdc++ in WANTLIB with that variable. > > But this is part of a wider problem that we don

spring cleanup, old scripts removal

2017-02-20 Thread Marc Espie
I'd like to remove extract-dependencies find-build-order (dpb does it and more) portslogger (haven't used it in ages, again, dpb) and maybe retrieve-index does anyone actually use that one ? (personally, I do either grep on INDEX, or I use sqlports whenever I really need something)

Re: spring cleanup, old scripts removal

2017-02-21 Thread Marc Espie
On Mon, Feb 20, 2017 at 11:47:00PM +, Nigel Taylor wrote: > On 02/20/17 09:44, David Coppa wrote: > > On Mon, Feb 20, 2017 at 10:13 AM, Marc Espie wrote: > >> I'd like to remove > >> extract-dependencies > >> find-build-order > >> (dpb doe

more spring cleanup

2017-02-21 Thread Marc Espie
A few oddities in bsd.port.mk - does anyone still use WRKOBJDIR= (which used to create w-* directly in each port's directory) - CDROM_SITE ? - link-categories/unlink-categories ? - homepage-links ?

ports tree dependency introspection

2017-02-21 Thread Marc Espie
Since we have caching running, "modern ports" uses a mechanism like 'show-run-depends' to yield a recursive trimmed dependency list quickly. We also have an older mechanism (*-dir-depends) that yields a full list suitable for tsort. Con: show-run-depends does NOT show the parent of each dependenc

Re: more spring cleanup

2017-02-22 Thread Marc Espie
On Tue, Feb 21, 2017 at 04:11:18PM +, Christian Weisgerber wrote: > On 2017-02-21, Marc Espie wrote: > > > A few oddities in bsd.port.mk > > > > - CDROM_SITE ? > > This is somewhat useful. Not for actual CDROM distribution, but > for opportunistically get

HEADS UP: fetch/checksum/makesum tweaks

2017-02-22 Thread Marc Espie
I've recently streamlined a bit the fetch/checksum/makesum code... too many tests. One *wanted* side-effect is to make it slightly harder to have distfiles without checksum. Doesn't change anything for normal port users. When you're creating a new port, however, you will probably want to run "ma

Re: HEADS UP: fetch/checksum/makesum tweaks

2017-02-22 Thread Marc Espie
On Wed, Feb 22, 2017 at 07:37:09PM +0100, Jeremie Courreges-Anglas wrote: > I'd like to point out that it harms a process I have as a port user. If > projects published signatures for their releases, I want to check them, > because I can have a trust relationship with upstream. > So the process t

Re: HEADS UP: fetch/checksum/makesum tweaks

2017-02-23 Thread Marc Espie
On Thu, Feb 23, 2017 at 02:48:30PM +0100, Jeremie Courreges-Anglas wrote: > Marc Espie writes: > > > On Wed, Feb 22, 2017 at 07:37:09PM +0100, Jeremie Courreges-Anglas wrote: > >> I'd like to point out that it harms a process I have as a port user. If > >>

Re: HEADS UP: fetch/checksum/makesum tweaks

2017-02-23 Thread Marc Espie
On Thu, Feb 23, 2017 at 04:05:30PM +0100, Alexander Bluhm wrote: > On Thu, Feb 23, 2017 at 03:21:22PM +0100, Marc Espie wrote: > > Yes, "make extract' wouldn't deal with it, but the file would still be there > > and extractible manually. It is harder to do now. >

Re: sysctl set up in packages

2017-03-01 Thread Marc Espie
On Tue, Feb 28, 2017 at 10:21:05PM +, Stuart Henderson wrote: > Send diff :) Rather, do not. Removing keywords from pkg tools is somewhat tricky. There are some steps you have to take to keep pkg_add -u working.

Re: sysctl set up in packages

2017-03-02 Thread Marc Espie
On Wed, Mar 01, 2017 at 11:36:44PM +, Stuart Henderson wrote: > On 2017/03/01 18:07, sven falempin wrote: > > On Wed, Mar 1, 2017 at 10:25 AM, Marc Espie wrote: > > > > > On Tue, Feb 28, 2017 at 10:21:05PM +, Stuart Henderson wrote: > > > > Sen

Re: sysctl set up in packages

2017-03-02 Thread Marc Espie
This is what the actual proper removal code looks like when you don't miss the finer points of the framework. Index: OpenBSD/Add.pm === RCS file: /cvs/src/usr.sbin/pkg_add/OpenBSD/Add.pm,v retrieving revision 1.172 diff -u -p -r1.172

Re: ghc.port.mk (was: Ports with hardcoded "gcc", "g++")

2017-03-03 Thread Marc Espie
On Thu, Mar 02, 2017 at 10:03:29PM +0100, Karel Gardas wrote: > On Thu, Mar 2, 2017 at 9:23 PM, Matthias Kilian > wrote: > > Also, I've no idea wether --with-clang=${CC} will work after the > > switch to clang. On the other hand, the diff is just a workaround, > > > +MODGHC_SETUP_CONF_ARGS +=

Re: imake: honor CC, CFLAGS, CXX, CXXFLAGS

2017-03-07 Thread Marc Espie
On Mon, Mar 06, 2017 at 09:01:53PM +0100, Christian Weisgerber wrote: > Ports that use CONFIGURE_STYLE=imake don't actually honor the CC, > CFLAGS, CXX, CXXFLAGS settings. They just pick up whatever is > hardcoded in devel/imake-cf. This behavior has grown historically > and there is really no go

streamlined pkg_add messages

2017-03-07 Thread Marc Espie
I just spent a few hours removing some extra error messages from pkg_add e.g., $ doas pkg_add screen-- will just say: http://ftp.fr.openbsd.org/pub/OpenBSD/6.0/packages/amd64/screen-4.0.3p6.tgz: unsigned package (signify(1) doesn't see old-style signatures) instead

Re: Firefox and Chrome and linkedin account create/recaptcha

2017-03-13 Thread Marc Espie
On Mon, Mar 13, 2017 at 12:07:29PM +, Stuart Henderson wrote: > There's also an annoying UK webshop (broadbandbuyer.co.uk) that also tries > to force a recaptcha if I use it from OpenBSD (I guess they're doing some > user-agent check or something), I get a similar thing where the recaptcha > ju

Re: openbsd ports of perl modules

2017-03-13 Thread Marc Espie
On Mon, Mar 13, 2017 at 04:00:06PM +0300, Sergey Bronnikov wrote: > Hello, > > some time ago I made an OpenBSD port for Tapper testing system. > This port depended from huge amount of Perl modules and I made port > for each of them. Right now Tapper development is discontinued > and I removed Tapp

Re: make an additional category in ports tree

2017-03-27 Thread Marc Espie
On Fri, Mar 17, 2017 at 05:12:51PM +0300, Sergey Bronnikov wrote: > OpenBSD ports contains many applications useful for different kinds of > testing: load, performance, functional, fuzzing etc. > Load and performance tests locate in benchmarks subdir mostly, > but other tests locate in different su

Re: Question about updating packages

2017-03-27 Thread Marc Espie
On Mon, Mar 20, 2017 at 08:57:42PM -0500, mitchell wodach wrote: > Someone correct me if I'm wrong but wouldn't you want to pkg_add -Uu. Which > updates dependent packages also? Nope. pkg_add -U exists for when you want to install a package in a hurry, and don't want to update your whole base of

HEADS up: how to cope with pre-release testing with pkg_add(1)

2017-03-31 Thread Marc Espie
Just so that the subject makes it clear. This info already made it in various messages ont the mailing list. Right now, if you install OpenBSD, it says 6.1 without beta. So pkg_add will look for the release directory, and not find it yet because it's not out yet. pkg_add -Dsnap will make pkg_ad

Re: libc++ / libstdc++ WANTLIB

2017-04-08 Thread Marc Espie
Oh I missed the Cc to ports. Well, I got a more sturdy patch for check-lib-depends (to be tested/documented). The gist is to be able to generically be able to add some lib variables into WANTLIB and have check-lib-depends recognize them. Basically, this would need some support in bsd.port.mk as

Re: textproc/agrep: "fix" or kill?

2017-04-12 Thread Marc Espie
On Tue, Apr 11, 2017 at 02:41:26PM +0100, Stuart Henderson wrote: > With this PERMIT_* I'm not sure if it's really worth bothering with. > > > Index: Makefile > === > RCS file: /cvs/ports/textproc/agrep/Makefile,v > retrieving revisi

Re: textproc/agrep: "fix" or kill?

2017-04-12 Thread Marc Espie
On Tue, Apr 11, 2017 at 02:41:26PM +0100, Stuart Henderson wrote: > With this PERMIT_* I'm not sure if it's really worth bothering with. > > > Index: Makefile > === > RCS file: /cvs/ports/textproc/agrep/Makefile,v > retrieving revisi

Re: textproc/agrep: "fix" or kill?

2017-04-12 Thread Marc Espie
On Wed, Apr 12, 2017 at 07:46:28AM -0600, Todd C. Miller wrote: > We already have the superior TRE version of agrep in ports. Why > do we need the old agrep? > > - todd We don't really need it, indeed, but quirks should have an update path then.

Re: amd64-clang port build failures

2017-04-16 Thread Marc Espie
On Fri, Apr 14, 2017 at 04:49:20PM +0200, Christian Weisgerber wrote: > * In the clang module, also link /usr/local/bin/clang and clang++ > into the leading PATH directory. Should that actually be default ? it will make sure the clang from ports is used on clang architectures, if that's the in

Re: make an additional category in ports tree

2017-04-16 Thread Marc Espie
On Sun, Apr 16, 2017 at 01:17:43PM +0300, Sergey Bronnikov wrote: > ok to commit? > > On 22:46 Tue 28 Mar , Sergey Bronnikov wrote: > > On 08:58 Tue 28 Mar , Antoine Jacoutot wrote: > > > On Tue, Mar 28, 2017 at 01:20:06AM +0300, Sergey Bronnikov wrote: > > > > > Call it test, rather. > > > > > >

Re: devel/boost vs. clang

2017-04-18 Thread Marc Espie
On Tue, Apr 18, 2017 at 08:24:55PM +0200, Christian Weisgerber wrote: > I tried to get boost to build with clang, but failed pretty miserably > because I don't know what I'm doing. > > If anybody wants to look into this, here's what I got: > * Configure with --with-toolset=clang > * Fix boost_has_

Re: databases/mongodb build error

2016-05-09 Thread Marc Espie
On Mon, May 09, 2016 at 08:29:36AM +0100, Stuart Henderson wrote: > On 2016/05/08 23:15, Andrew Aldridge wrote: > > On 2016-05-07 17:37, Christian Weisgerber wrote: > > > databases/mongodb failed to build in my latest amd64 bulk build: > > > > > > scons: *** > > > [build/opt/third_party/mozjs-38/p

HEADS UP: PLIST_DB replaced with PLIST_REPOSITORY

2016-05-20 Thread Marc Espie
A few days, we changed from PLIST_DB to PLIST_REPOSITORY, so that it looks a lot more like PACKAGE_REPOSITORY. (specifically, having to figure out MACHINE_ARCH in one case and not in the other was turning to be a pain for some scripts and some installations). If you didn't even know about PLIST_D

HEADS UP: important tweaks in proot(1)

2016-06-04 Thread Marc Espie
By default, it now does unpopulate_light. I'm now reasonably confident the avoidance mechanisms for not deleting important stuff are good enough, so it will wipe chroots of anything that doesn't belong there. See the man page which was updated accordingly. Some fringe case scenarios may need to us

Re: Sandboxing Haskell Port Builds: ghc.port.mk diff + example converters/pandoc

2016-06-15 Thread Marc Espie
For those interested, I added a man page called bulk(8) to ports/infrastructure recently. It does explain in some details how to setup a cluster for package building. This is *de facto* the standard you have to deal with for all ports. Official and development build clusters are built using tho

Re: Unknown fragment error reporting

2016-06-21 Thread Marc Espie
On Tue, Jun 21, 2016 at 11:24:30AM +0200, Adam Wolk wrote: > The error report is quite deep in Subst.pm handle_fragment which is > called from PkgCreate.pm. Is there a way to obtain the ports/package > name or folder in a nice way inside handle_fragment? Adding a package > name to the error report

heads up: @option is-branch

2016-06-28 Thread Marc Espie
I've added some new annotation for plists (@option is-branch) that's going to be used by pkg_info (soon) to figure out stuff like autoconf%2.13 (to display for quick reinstalls instead of full pkgnames). All fast arch snapshots have the new pkg_add code already

packages branch info

2016-06-30 Thread Marc Espie
I committed a set of "useful features" that go together and will be in OpenBSD 6.0. - branch info choice on package stems. when you say pkg_add php you end up having to choose which php you want. now you can say pkg_add php%5.5 this "selects the 5.5 branch", which is already encoded as pkgpath w

Re: [new] exfat-fuse 1.2.4

2016-07-09 Thread Marc Espie
On Sat, Jul 09, 2016 at 05:02:31PM +0200, Antoine Jacoutot wrote: > On Sat, Jul 09, 2016 at 04:04:40PM +0200, Dmitrij D. Czarkoff wrote: > > YASUOKA Masahiko wrote: > > > > > I'd like to add 'exfat-fuse'. > > > > > > It is an exFAT file system implementation working on FUSE. It's > > > useful t

Re: youtube-dl update?

2016-07-10 Thread Marc Espie
On Sun, Jul 10, 2016 at 09:25:31AM +0200, Remi Pointel wrote: > On 07/09/16 23:06, Ted Unangst wrote: > >The current ports version of youtube-dl cannot handle vine anymore. > >Downloading a new version, 2016.07.09.2, works fine manually. However, > >when I update the port to use this version, it fa

Re: Vulnerable packages in ports 26/07/2016

2016-07-27 Thread Marc Espie
On Tue, Jul 26, 2016 at 05:17:02PM +0100, Sevan Janiyan wrote: > databases/sqlite3 > https://www.korelogic.com/Resources/Advisories/KL-001-2016-003.txt Does not affect sqlite3 in a standard OpenBSD install

Re: net/syncthing and @extraunexec

2016-08-09 Thread Marc Espie
On Tue, Aug 09, 2016 at 02:45:46PM -0400, Michael McConville wrote: > Hi, everyone. > > I was recently discussing this with edd@, who suggested that I pull in > ports@ and Marc. > > Syncthing is a daemon that lets you sync data between devices. By > default, both metadata and user data are stored

Re: net/syncthing and @extraunexec

2016-08-10 Thread Marc Espie
On Wed, Aug 10, 2016 at 12:14:13AM +0200, Antoine Jacoutot wrote: > The problem is that these messages are displayed at update time. > I assume because from pkg_add's point of view it's a delete+add. > IMHO UNMESSAGE and @extraunexec stuffs should not be displayed at update time. UNMESSAGE is alre

what to say to upstream about shared libraries

2016-08-15 Thread Marc Espie
Got the question in private mail. Yes, we need to take control of shared library versions. You may have to explain things to upstream, but they do NOT have full control over the ABI. You may need to explain to them the difference between API and ABI. Even if their prototypes don't change, that d

Re: Getting rid of fortran.port.mk

2016-08-22 Thread Marc Espie
On Mon, Aug 22, 2016 at 11:01:53AM +0200, David Coppa wrote: > > Hi, > > Given that: > > - lang/g77-old devel/libf2c-old are long dead > > - lang/gfortran will go to the Attic in the same way > > - to use the newer fortran (egfortran) from gcc 4.9, it's better > to use the MODGCC4_* macros (

Re: Getting rid of fortran.port.mk

2016-08-22 Thread Marc Espie
On Mon, Aug 22, 2016 at 05:03:16PM +0200, David Coppa wrote: > On Mon, 22 Aug 2016, Christian Weisgerber wrote: > > > On 2016-08-22, Marc Espie wrote: > > > > >> - to use the newer fortran (egfortran) from gcc 4.9, it's better > > >

Re: minidlna

2016-08-25 Thread Marc Espie
On Mon, Aug 22, 2016 at 06:42:36PM -0500, Ed Ahlsen-Girard wrote: > While attempting to trouble shoot minidlna (21 Aug 2016 snap) I found > that the package has no docs, either as minidlna, minidlnad, or > readymedia. Checked for man and info pages both. In addition to not > running on my machine.

Re: WIP: GCC 6.2.0

2016-09-04 Thread Marc Espie
On Sun, Sep 04, 2016 at 01:30:26PM +0100, Pascal Stumpf wrote: > Here's a preliminary port of GCC 6.2.0, tested only on amd64, i386 and > sparc64 as I don't have access to anything else at the moment. > > Go ahead and play with it. I'd vote that you put it in, unlinked, and we can play in tree.

Re: lftp man page broken

2016-09-13 Thread Marc Espie
On Tue, Sep 13, 2016 at 03:32:05AM -0600, Anthony J. Bentley wrote: > Jiri B writes: > > Hi, > > > > lftp man pages is odd, see section for 'mirror' command. > > The lftp(1) manual uses tbl(7) macros, which need extra preprocessing. > pkg_create(1) doesn't do any preprocessing... although that's

Re: error: Can't call method "is_signed" on an undefined

2016-10-04 Thread Marc Espie
On Sat, Oct 01, 2016 at 11:45:24AM -0700, patrick keshishian wrote: > Hi, > > This is on a current system (fresh install): > > # sysctl kern.version > kern.version=OpenBSD 6.0-current (GENERIC.MP) #2510: Fri Sep 30 > 09:49:52 MDT 2016 > dera...@amd64.openbsd.org:/usr/src/sys/arch/amd64/compil

some more info on pkg_add/fw_update changes

2016-10-04 Thread Marc Espie
About a week ago, we switched to the new signing scheme by default. There are good reasons to bury the old signing scheme completely, so this is what's currently happening, there are some rough edges. Technically speaking, the new signatures are "outside", they're in the gzip header, and the only

Re: some more info on pkg_add/fw_update changes

2016-10-05 Thread Marc Espie
On Tue, Oct 04, 2016 at 03:15:18PM +0200, Marc Espie wrote: > - the new scheme is slightly more unflexible with respect to unsigned > data: by default, every .tgz is piped thru signify -Zs, so > pkg_add/pkg_info/fw_update WON'T even see any data if it's not signed. &

a note about pkg_sign

2016-10-06 Thread Marc Espie
Note that if you want to sign packages yourself, your private key *MUST* be name "something.sec", so that the proper comment says something.pub to be able to decode things. This is weird and nonsensical behavior on signify's part, which enforces that, but silently, without any error message when t

Re: DPB can't build mail/sendmail

2016-10-12 Thread Marc Espie
On Tue, Oct 11, 2016 at 08:24:15PM -0400, trondd wrote: > I had this problem in 5.9 and forgot about it until 6.0, but DPB can't > successfully build the sendmail package. > > DPB is using the default _pbuild, etc users so the build is not running as > root. Is there some work around for this?

Re: "pkg_info -S" can't get update signature, "unsigned package"

2016-10-17 Thread Marc Espie
On Mon, Oct 17, 2016 at 08:49:43AM +0200, Andreas Kusalananda Kähäri wrote: > On Sun, Oct 16, 2016 at 09:37:50AM +0100, Stuart Henderson wrote: > > On 2016/10/16 10:27, Andreas Kusalananda Kähäri wrote: > > > Hi, > > > > > > "pkg_info -S" used to show the "update signature" for a package, i.e. > >

Re: poison gettext-tools

2016-10-17 Thread Marc Espie
On Mon, Oct 17, 2016 at 09:34:49AM +0100, Stuart Henderson wrote: > I'm going to do a bulk build with this to see if it shakes anything out. > Any comments on whether it's worth committing? > > Index: bsd.port.mk > === > RCS file: /cv

Re: [s...@spacehopper.org: RFC: USE_LLD=No for default-is-lld arches]

2018-12-27 Thread Marc Espie
On Tue, Dec 25, 2018 at 08:15:55PM +, Stuart Henderson wrote: > Any objections or OKs? The principle is good. > - Forwarded message from Stuart Henderson - > > From: Stuart Henderson > Date: Tue, 30 Oct 2018 11:54:15 + > To: ports > User-Agent: NeoMutt/20180716 > Subject: RFC:

Re: UPDATE: bsd.port.mk - Add GitLab support

2018-12-31 Thread Marc Espie
Adding more fluff to bsd.port.mk to support this style of code is fairly disturbing. I don't like the github stuff too much, it's somewhat error-prone and there is regular traffic on ports-changes proving it. Adding a SECOND source of problems does not seem like the way to go. There's also the i

Re: UPDATE: bsd.port.mk - Add GitLab support

2019-01-02 Thread Marc Espie
On Mon, Dec 31, 2018 at 07:01:16PM +0100, Antoine Jacoutot wrote: > On Mon, Dec 31, 2018 at 05:45:19PM +0100, Marc Espie wrote: > > Adding more fluff to bsd.port.mk to support this style of code is fairly > > disturbing. > > > > I don't like the github stuff too

Re: llvm dep on g++ (was: Re: powerpc bulk build report)

2019-01-04 Thread Marc Espie
On Fri, Jan 04, 2019 at 12:47:26PM +0100, Jeremie Courreges-Anglas wrote: > On Mon, Dec 31 2018, George Koehler wrote: > > On Mon, 31 Dec 2018 00:15:56 -0700 (MST) > > lan...@openbsd.org wrote: > > > >> http://build-failures.rhaalovely.net//powerpc/2018-12-14/x11/qt5/qtbase.log > > > > Fails to #i

Re: [s...@spacehopper.org: RFC: USE_LLD=No for default-is-lld arches]

2019-01-04 Thread Marc Espie
On Wed, Jan 02, 2019 at 10:16:16PM +, Stuart Henderson wrote: > On 2018/12/28 02:53, Marc Espie wrote: > > > +.if ${_LD_PROGRAM} != ${_LD_DEFAULT} > > > +_NONDEFAULT_LD = Yes > > > +.else > > > +_NONDEFAULT_LD = No > > > +.endif > > > +

Re: aarch64: use gcc8 as ports-gcc

2019-01-05 Thread Marc Espie
On Fri, Jan 04, 2019 at 07:15:00PM +0100, Pascal Stumpf wrote: > On Fri, 4 Jan 2019 17:11:01 +, Stuart Henderson wrote: > > On 2019/01/04 17:57, Pascal Stumpf wrote: > > > This switches aarch64 to use GCC 8.2.0 as default. Doesn't do any harm > > > because this is the first GCC port that works

Re: password-store depends on gnupg-1.4.23?

2019-02-07 Thread Marc Espie
On Wed, Feb 06, 2019 at 11:32:54PM +0100, Sebastian Benoit wrote: > $ doas pkg_delete gnupg-1.4.23p1 > can't delete gnupg-1.4.23p1 without deleting password-store-1.7.3 > Delete them as well ? [y/N/a] > > but password-store has security/gnupg2 as dependency, not gnu

Re: go uses too many CPUs

2019-02-08 Thread Marc Espie
On Fri, Feb 08, 2019 at 12:10:02AM -0500, Ted Unangst wrote: > Go tries to use NCPU cpus. Unfortunately, half of them are turned off because > hw.smt=0 by default, and then go spends a lot of time fighting against itself. > > The diff below, against go/src/runtime, changes to use the number of CPU

install(1) broken

2019-02-09 Thread Marc Espie
hey, your commit to install(1) broke something. Specifically lang/go-boostrap now produces a broken package which can't be used to build go. All the go/bootstrap/pkg/tool/openbsd_amd64/* have lost their x bit Relevant fake install information, it definitely looks like the last line is now a no-o

Re: install(1) broken

2019-02-09 Thread Marc Espie
On Sat, Feb 09, 2019 at 05:23:09PM -0500, Ted Unangst wrote: > Marc Espie wrote: > > hey, your commit to install(1) broke something. > > > > Specifically lang/go-boostrap now produces a broken package which can't > > be used to build go. > > > >

Re: install(1) broken

2019-02-10 Thread Marc Espie
On Sat, Feb 09, 2019 at 05:23:09PM -0500, Ted Unangst wrote: > Marc Espie wrote: > > hey, your commit to install(1) broke something. > > > > Specifically lang/go-boostrap now produces a broken package which can't > > be used to build go. > > > >

Re: install(1) broken

2019-02-10 Thread Marc Espie
Something similar to this perhaps ? Not fully tested yet, but it should avoid the race of trying to unlink tempfile several times, and also fix the file name in error messages. Index: xinstall.c === RCS file: /cvs/src/usr.bin/xinsta

Re: install(1) broken

2019-02-11 Thread Marc Espie
On Sun, Feb 10, 2019 at 03:55:41PM +0100, Marc Espie wrote: > Something similar to this perhaps ? > Not fully tested yet, but it should avoid the race of trying to > unlink tempfile several times, and also fix the file name in error messages. That's now been tested (gone thru a

Re: lang/gcc/8 can't build x11/qt4

2019-02-13 Thread Marc Espie
On Tue, Feb 12, 2019 at 11:48:03PM -0500, George Koehler wrote: > There are probably more packages that lang/gcc/8 can't build, but the > only ones that I know are x11/qt4 and games/stone-soup. Our > stone-soup is an old version; an update might fix it. Maybe you should start with stone-soup ? O

Re: lang/ghc

2019-02-28 Thread Marc Espie
On Tue, Feb 26, 2019 at 10:21:07PM +0100, Matthias Kilian wrote: > Some notes: make update-plist seems to remove lib/ghc/unregister.sh > from pkg/PLIST, I've no idea, why. ghc is about the only port left with major @exec/@unexec in the plist. it would be great if those scripts could be abstracte

Re: lang/ghc

2019-02-28 Thread Marc Espie
On Thu, Feb 28, 2019 at 08:43:49PM +0100, Matthias Kilian wrote: > Hi, > > On Thu, Feb 28, 2019 at 04:54:51PM +0100, Marc Espie wrote: > > On Tue, Feb 26, 2019 at 10:21:07PM +0100, Matthias Kilian wrote: > > > Some notes: make update-plist seems to remove lib/ghc/unre

Re: Makefile.template tweaks?

2019-03-04 Thread Marc Espie
On Sun, Mar 03, 2019 at 06:57:51PM +0100, Rafael Sadowski wrote: > On Sun Mar 03, 2019 at 02:16:18PM +, Stuart Henderson wrote: > > On 2019/03/03 14:11, Stuart Henderson wrote: > > > comments/ok? > > > > Also about adding this: > > > > # For ports using C++, except in special cases you should

Re: Makefile.template tweaks?

2019-03-04 Thread Marc Espie
On Mon, Mar 04, 2019 at 01:19:51AM +0100, Antoine Jacoutot wrote: > On Sun, Mar 03, 2019 at 02:11:42PM +, Stuart Henderson wrote: > > comments/ok? > > OK > I wish we could somewhat automate this. More CONFIGURE_STYLE to get rid of more do-gen style targets would be welcome. If you can find so

Re: Makefile.template tweaks?

2019-03-04 Thread Marc Espie
On Mon, Mar 04, 2019 at 03:12:40PM +0100, Antoine Jacoutot wrote: > On Mon, Mar 04, 2019 at 03:04:22PM +0100, Marc Espie wrote: > > On Mon, Mar 04, 2019 at 01:19:51AM +0100, Antoine Jacoutot wrote: > > > On Sun, Mar 03, 2019 at 02:11:42PM +, Stuart Henderson wrote: &

Re: Split games/minetest into main, server and data subpackages

2019-03-07 Thread Marc Espie
On Thu, Mar 07, 2019 at 08:07:25AM +0200, Leonid Bobrov wrote: > On Thu, Mar 07, 2019 at 08:02:19AM +0300, Kirill Bychkov wrote: > > Hi! > > What is the reason to split 15 MiB package and complicate the port? > > Some games using such a split to save space on mirrors, but they > > have more than 10

Re: [UPDATE] www/youtube-dl 2019.01.17 -> 2019.03.01

2019-03-07 Thread Marc Espie
On Thu, Mar 07, 2019 at 04:04:33PM +0100, Charlene Wendling wrote: > Hi Paul, ports, > > Here is an update for youtube-dl. Our current version doesn't > work with Soundcloud anymore. > > Upstream changes [1] includes mostly extractors fixes and > additions/deletions. Curl infinite retries has be

Re: remove devel/{cudf,omake,ounit,ocaml-{cmdliner,cppo,dose,extlib,jsonm,re}} ?

2019-03-14 Thread Marc Espie
On Mon, Mar 11, 2019 at 09:55:53PM +0100, Christopher Zimmermann wrote: > Hi, > > switching sysutils/opam to a source distribution including all > (partly patched) dependencies, the following OCaml libraries are no > longer used by other ports. Since active OCaml users will use opam > instead of O

Re: devel/boost vs. clang

2017-04-19 Thread Marc Espie
On Wed, Apr 19, 2017 at 01:33:40AM +0200, Marc Espie wrote: > On Tue, Apr 18, 2017 at 08:24:55PM +0200, Christian Weisgerber wrote: > > I tried to get boost to build with clang, but failed pretty miserably > > because I don't know what I'm doing. > > > > If an

update-patches improvements

2017-05-04 Thread Marc Espie
If you don't follow ports-changes, you may not have noticed that I've finally given up and completely replaced update-patches with perl code. This allowed me to implement several improvements. - instead of diffing file.orig and file, if update-patches finds a file.beforesubst, it will diff file.o

Re: New: devel/noweb 2.11b

2017-05-04 Thread Marc Espie
On Thu, May 04, 2017 at 04:14:56PM +0200, Jeremie Courreges-Anglas wrote: > modify -> derived work? > > isn't it? I do not think our ports are derived works (maybe some > patches we have really are intrusive?) Very often, it will trigger a move to include our patches upstream, which is a good

Re: UPDATE: textproc/glimpse for clang

2017-05-07 Thread Marc Espie
On Sat, May 06, 2017 at 03:31:24PM -0400, Stuart Cassoff wrote: > Index: Makefile > === > RCS file: /cvs/ports/textproc/glimpse/Makefile,v > retrieving revision 1.27 > diff -u -p -u -p -r1.27 Makefile > --- Makefile 11 Apr 2017 13:57:

Re: Pgtl: Take maintainership

2017-05-09 Thread Marc Espie
On Tue, May 09, 2017 at 09:05:52AM -0400, Stuart Cassoff wrote: > I contacted the person currently listed as maintainer, who replied: > "I am not involved neither with that port nor with OpenBSD anymore". > > I'd like to take maintainership, unless someone already wants it. > Hopefully we won't ne

Re: [PATCH] scummvm and residualvm GCC 4.2.1 miscompilation

2017-05-09 Thread Marc Espie
gcc4 is wildly incompatible with base clang. More precisely the C++ library. So, yeah, MODGCC4 on aarch64 ? not that useful.

amd64 ports clang, status update

2017-05-13 Thread Marc Espie
I've now polished WANT_CXX enough to commit it. It's simply there to choose base clang, or gcc4 according to the architecture. I also introduced LIBECXX so that wantlib can be adjusted without needing a lot of conditionals. lang/clang is more complex so this will happen later. I'm in the process

tweak: allow simulation of register-plist

2017-05-14 Thread Marc Espie
There are actually two patches, one for src, the other for port. 1st patch unconditionally regenerates special files, even in -n mode, so that they can be checksummed. As a result, the printed plist contains @sha256 markers for these. Con: very slightly slower. 2nd patch allows register-plist t

Re: update: textproc/ripgrep 0.5.2

2017-05-14 Thread Marc Espie
On Sun, May 14, 2017 at 03:08:59PM +0200, Sebastien Marie wrote: > Hi, > > An update to textproc/ripgrep 0.5.2. > > See https://github.com/BurntSushi/ripgrep/releases/tag/0.5.2 for > changelog. > > While here, adds gcc4 MODULES in order to use egcc as linker. It should > let ripgrep to build wit

Re: tweak: allow simulation of register-plist

2017-05-14 Thread Marc Espie
Forget hope, I forgot about the symlinks registration and everything. I guess I'll have to do something smarter in register-plist, should be feasible.

Re: tweak: allow simulation of register-plist

2017-05-15 Thread Marc Espie
On Sun, May 14, 2017 at 10:21:21PM +0200, Marc Espie wrote: > Forget hope, I forgot about the symlinks registration and everything. > I guess I'll have to do something smarter in register-plist, should be > feasible. More specifically, my current plans involve being able to &quo

Re: Non-English manual pages in ports

2017-05-16 Thread Marc Espie
On Mon, May 15, 2017 at 11:38:54PM +0200, Ingo Schwarze wrote: > 5. If possible, install to man/language/manN, without any "_", ".", > or "@": KISS. There are rare exceptions, most notably zh_CN > vs. zh_HK vs. zh_TW. > Never include the encoding in the path name, and make sure >

Re: Non-English manual pages in ports

2017-05-16 Thread Marc Espie
On Tue, May 16, 2017 at 05:39:54PM +0200, Ingo Schwarze wrote: > Hi Marc, > > Marc Espie wrote on Tue, May 16, 2017 at 10:38:07AM +0200: > > On Mon, May 15, 2017 at 11:38:54PM +0200, Ingo Schwarze wrote: > > >> 5. If possible, install to man/language/manN, withou

Re: Non-English manual pages in ports

2017-05-17 Thread Marc Espie
On Tue, May 16, 2017 at 11:42:32PM -0300, Diogo Galvao wrote: > On Tue, May 16, 2017 at 1:01 PM, Marc Espie wrote: > >> Then there are a number of /pt_BR/ in addition to /pt/. > >> That looks suspicious, but i don't speak Portuguese, > >> so it may or may

ports clang report

2017-05-24 Thread Marc Espie
So we got a building, working boost now. What I did was fairly stupid: simply copy the gcc autoconfig configuration, and use it for now. We can always activate the remaining stuff afterwards. While running boost, I found two issues: - excessive warnings, because boost hardcode __attribute__((unuse

relocation errors due to clang

2017-05-25 Thread Marc Espie
there are 13 ports that don't link because of the confusion between standard library and internal variables: comms/minicom: R_X86_64_PC32 against `COLS' devel/arm-elf/gdb: R_X86_64_PC32 against `PC' devel/avr-gdb: R_X86_64_PC32 against `PC' devel/msp430/gdb: R_X86_64_PC32 against `PC' editors/el

Re: apcupsd/clang

2017-05-25 Thread Marc Espie
On Thu, May 25, 2017 at 02:27:51PM +0100, Stuart Henderson wrote: > apcupsd tries to link its binaries using cc (rather than c++) with a static > link to -lsupc++, for people who "care about binary portability between > Linux distributions". > > the simplest fix seems to be to just use c++ to link

Re: Handle --tag= in the middle of libtool options list

2017-05-27 Thread Marc Espie
On Sat, May 27, 2017 at 11:36:46AM +0300, Vadim Zhukov wrote: > This allows to handle --tag=... args in the middle of options list. > Unbreaks multimedia/lives with base Clang. I don't see any potential > breakage because compilers and linkers do not accept --tag= anyway. > > Okay? > > -- > WBR,

hidden dependency in games/eduke32

2017-05-27 Thread Marc Espie
while checking eduke32 for clang, I ran into a weird compile error. Turns out eduke32 includes an (old?) version of lz4 that doesn't look quite like the one in archivers/lz4. The following patch makes eduke32 just use archivers/lz4 so it won't miscompile if lz4 happens to be installed. Index: Ma

scheme48 update

2017-05-27 Thread Marc Espie
Here's a small update to scheme48. I'm not sure what changed, but this one actually builds and works with clang. 1.9 does segfault once built: ./build/build-usual-image . \ "`(cd . && pwd)`/scheme" "`pwd`/c" 'scheme48.image' './scheme48vm' \ './build/initial.image-64' Segmentation

[update and improvement] gprolog

2017-05-29 Thread Marc Espie
While trying to get everything to work with clang, I've stumbled on gprolog. Here is an update to 1.4.4. It doesn't make it work with clang either, but at least it's newer (I peeked: FreeBSD does compile gprolog with gcc) The most important improvement in the port is doing OpenBSD detection in c

Re: OpenBSD Package Manager

2017-05-29 Thread Marc Espie
On Mon, May 29, 2017 at 02:16:26PM -0400, H. Ishikawa wrote: > 3. Why so many connections? > When I tried to investigate why the update was so slow, I saw that > pkg_add was making one HTTPS connection per package! Tools like > wget from Linux can reuse a single connection for many downloads. > Cou

<    1   2   3   4   5   6   7   8   9   10   >