with blanks to invoked commands. This fixes clean-up of shared library
backups, where only one file was removed due to a parameter list being
terminated by the first blank within one parameter.
Reported by: tatsuki_makino at hotmail.com (Tatsuki Makino)
Approved by: antoine (implicit)
While there, cleanup, and sort depends.
When build and run dependencies are the same, there are three ways to
avoid duplicating the list while not adding the framework added
BUILD_DEPENDS to the RUN_DEPENDS. In order of preference, they are:
1) use RUN_DEPENDS to set BUILD_DEPENDS:
BUILD_DEPENDS= ${RUN_DEPENDS}
RUN_DEPENDS= foo:bar/baz
2) create another variable and use it:
MY_DEPENDS= foo:bar/baz
BUILD_DEPENDS= ${MY_DEPENDS}
RUN_DEPENDS= ${MY_DEPENDS}
3) use BUILD_DEPENDS to set RUN_DEPENDS and force evaluation:
BUILD_DEPENDS= foo:bar/baz
RUN_DEPENDS:= ${BUILD_DEPENDS}
Sponsored by: Absolight
As usual, it is recommended to rebuild or reinstall all the
dependent ports and the lang/ghc port itself in one of the following
ways:
# portmaster -w -r ghc
or
# portupgrade -fr lang/ghc
In case of pkg(8), it is probably safer to remove all the GHC-dependent
packages along with GHC and reinstall everything from scratch. For
example:
# pkg query "%ro" ghc > ghc-pkgs.txt
# pkg delete -y lang/ghc
In ghc-pkgs.txt, check and remove all the packages that have been moved
on the update, then use this command:
# pkg install -y `cat ghc-pkgs.txt`
Approved by: tcberner (mentor)
Differential Revision: https://reviews.freebsd.org/D16038
From now on, ports that depend on Qt4 will have to set
USES= qt:4
USE_QT= foo bar
ports depending on Qt5 will use
USES= qt:5
USE_QT= foo bar
PR: 229225
Exp-run by: antoine
Reviewed by: mat
Approved by: portmgr (antoine)
Differential Revision: →https://reviews.freebsd.org/D15540
Detailed maintainer log:
- Add TEST-DEPENDS.
- Add indication of the type of dependence by the first characters
in parentheses: (BR) for BUILD_DEPENDS and RUN_DEPENDS together.
- Make it ready for Python 3.
- Fix transition of svn sources from Google Code to SF for gen-dist.
PR: 228361
Submitted by: vladimir.chukharev@gmail.com (maintainer)
* Be less strict in the check for MASTER_SITE_SUBDIR use.
This change checks that the subdirectory used is actually different
than what is defined in MASTER_SITES_SUBDIRS. To make this easier to
do, support has been added to single-quote variables passed to the
get_makevar* functions. [1]
* Now only PY_FLAVOR is recommended as opposed to FLAVOR. [2]
* Remove an errant ':' when suggesting PY_FLAVOR. [3]
PR: 228227 [1]
Requested by: mat [2]
Reported by: koobs [3]
The ports infrastructure does not consistently deal with FLAVOR environment
variables set to an empty string. This caused the dependency check in
portmaster to fail for multimedia/phonon.
With FLAVOR="" in the environment, "make -V FLAVOR" returns "qt4", but the
dependency list produced by "make build-depends-list" assumes "qt5", even
though no user preferences for qt5 have been set.
Reported by: Walter Schwarzenfeld <w.schwarzenfeld at utanet.at>
Approved by: antoine (implicit)
* Update lang/ghc to 8.4.2
* Update the boostrap compiler to 8.4.1
* Update the many hs-* ports
* Bump the rest
Thanks a lot to arrowd for doing all the heavy lifting :)
PR: 227968
Exp-run by: antoine
Submitted by: arrowd
Differential Revision: https://reviews.freebsd.org/D15005
time ago, but decided to not expose it in the man-page, since portmaster
will not necessarily install the latest version if installing from local
Jpackages is requested and there exist multiple package files that only
differ in archive format. But since a non-standard package format allows
to choose between an optimization for speed or for disk space, this is a
useful option that shall be described in the man page.
Approved by: antoine (implicit)
called before being defined.
The issue and cause has been reported by Herbert J. Skuhra. (Thanks!)
PR: 227882
Reported by: Herbert J. Skuhra <herbert@gojira.at>
Approved by: antoine (implicit)
The quoting was messed up by use of "eval" and the cached values did not
match what the ports system expected.
Submitted by: tijl
Reported by: tijl
Approved by: antoine (implicit)
and over for individual ports. This significantly reduces the time to scan
for updates if many ports with USES=compiler are present.
Based on a suggestion and code fragment provided by Tijl Coosemans.
While here commit a few changes present in my local version that should not
cause any functional change.
Submitted by: tijl
Approved by: antoine (implicit)
Fix a regression in the mae makepatch check.
Make sure only the patch header is checked, but make sure
to read in the entire patch file. This regression was
introduced in 2.18.0.
PR: 228203
Soften the py-* dependency error around flavors.
Now the message is a warning and a suggestion. This is pending more discussion,
but this seems like a good compromise for now.
PR: 225654
Requested by: antoine
* Fix the check for duplicate SHEBANG_LANG values [1]
* Check to make sure patch files end with newlines [2]
* Allow flavored *_DEPENDS in the DEPENDS section [3]
* Check that kde:4 OR kde:5 is defined with USE_KDE [4]
* Add a check to make sure pkg-descr is at least a certain number of lines [5]
* Fix a couple of typos [6] [7]
* Allow for any *_DEPENDS item in the DEPENDS section and fix a bug in the
dependency syntax checking [8]
* Add a check to make sure Python dependencies include a FLAVOR [9]
PR: 226709 [1]
226109 [2]
226581 [3]
227255 [5]
227729 [6]
227878 [7]
227556 [8]
225654 [9]
Submitted by: tcberner [4]
0mp [5][6][7]
Differential Revision: D13723 [4]
- jail -u freebsd-update: Auto confirm prompts
- bulk tests: Don't let system perl interfere with the partial tree used.
- compute_deps: Determine all errors before bailing out
- Cleanup code
- Speed up and strictly check package's dependencies
- While here, pet portlint by using pkg-plist instead of PLIST_FILES
PR: 227730
Submitted by: kdeguchi@sz.tokoha-u.ac.jp (maintainer)
Approved by: pi (mentor)
- Add BASH option for completion installation
- ports_env: set UID=0 to avoid id -u lookups.
- Remove USE_PTSORT=no support and only use ptsort.
- Remove POOL_BUCKETS option
- Only set POOL_BUCKET_DIRS from load_priorities()
- SIGINFO: Reverse phase and total times to: phase / total
The original change order was confusing but so is the change. It may
ultimately be reverted.
- USE_CACHED=yes: Block using this as it does not work.
- Fix '3: Bad file descriptor' on crashed build
- Format crashed build with JOB_ID if possible
- MUTABLE_BASE=schg: Exclude immutable paths from mtree comparisons
- markfs: No need to rewrite the mtree file every time, just share it from
MASTERMNT
- Always null-mount ro /usr/lib32
- Remove references to INDEX - support removed in 0bef52955606412ac64aceddde6
- MUTABLE_BASE=schg: Need to track flags for mtree
- MUTABLE_BASE=schg mtree: Don't ignore changes to schg files with testport
or bulk -t
- Remove some shash buckets once no longer needed.
- Fix some error spam in bulk.
- Exclude /root from build_fs_violation.
- MUTABLE_BASE=schg: Allow writing new files to /boot.
- Move ALLOW_MAKE_JOBS_PACKAGES handle to remove misleading DISABLE_MAKE_JOBS
in buildlog make.conf
- bulk: Spawn builds in their own process group.
Block SIGTTOU/SIGTTIN to avoid stopping on tty access.
- kill_job: Don't show jobid warning.
- Close stdin at exit in case an interrupted redirection is referencing the
jail which results in umount warnings.
- processonelog: Detect pkg's conflict line
- processonelog: Try to detect immutable base failures
- Create /home and /usr/home paths in port build startup
- MUTABLE_BASE=schg: Allow writing to /usr/home
- poudriere: Default -e to /usr/local/etc for non-install executions
- options: Don't pass FLAVOR as a make arg - which may break downstream ports
- testport -c: Don't pass FLAVOR as a make argument
- testport: Show log summary at the end
- Support PKG_NOCOMPRESS in make.conf
- CCACHE_STATIC_PREFIX: Add CCACHE_JAIL_PREFIX for overriding /ccache in jail
- MUTABLE_BASE=schg + CCACHE_STATIC_PREFIX: Allow ccache-update-links to work
properly
- Add bash completion script
- Don't consider /var/crash/*.core as build_fs_violations
- distclean: Suck in ports_env which brings a big performance improvement.
- bulk: Add validation to check for dependencies with empty FLAVOR 'foo/bar@'
