File too long (should be no more than 24 lines).
Line too long (should be no more than 80 characters).
Trailing empty lines.
Trailing white-space.
Trucated the long files as best as possible while preserving the most info
contained in them.
COMMENT should not be longer than 70 characters.
COMMENT should not begin with 'A'.
COMMENT should not begin with 'An'.
COMMENT should not begin with 'a'.
COMMENT should not end with a period.
COMMENT should start with a capital letter.
pkglint warnings. Some files also got minor formatting, spelling, and style
corrections.
* In test, match the "make makesum" warning more loosely.
* In test, find packages to run against in $PKGSRCDIR explicitly.
* Bump version.
This is the last commit in a set developed for my "Rehabilitating
pkglint" talk at pkgsrcCon 2013.
Reviewed-at-pkgsrcCon-and-approved-during-the-freeze-by: wiz
Leave the silly self-grep in place, adjusting to also match the
lines that initialize the dispatch table. The underlying code is
better and the tests still pass. (No functional change intended.)
pkglint greps itself (really) to generate this list of types. Test
exhaustively for all types in the list today, so that the tested
code can become more civilized tomorrow. Adds a test-time dependency
on Test::Deep for cmp_bag().
In the configure stage, tell pkglint to find *.map in ${FILESDIR}.
That lets tests run even when pkglint isn't installed (which means
it also makes the tests less wrong when pkglint is installed!).
In the install stage, tell pkglint to find *.map under ${PREFIX},
as before (no functional change to the installed package).
Specify USE_TOOLS=perl to quell pkglint. :-)
These bits of non-main functionality are distracting to the casual
reader, and also to your intrepid refactorer. Since we recently
added some characterization tests that run the pkglint executable
against real packages, we can safely extract bits of code to separate
files and prove nothing broke in the move.
Instead of installing the new Perl modules where pkglint can find
them at runtime, I've added a build step to inline them right back
where they used to be (same order and everything). This is a little
annoying, but less annoying than suddenly needing to deal with Perl
runtime paths just for these few tiny modules. Also, loading modules
at runtime would affect pkglint's performance, and I don't yet have
performance tests to prove that this would be fine (or not), so I
can't safely do it.
No functional change intended.
This is the first commit in a set developed for my "Rehabilitating
pkglint" talk at pkgsrcCon 2013.
Reviewed-at-pkgsrcCon-and-approved-during-the-freeze-by: wiz
nih install:
- Order of packages does matter
("nih install emacs- editors/emacs22" is correct)
- Option -P and new config variable NO_REQPROV_CHECK were added
- Option -L and new config variable NO_REMOVE_LEAVES were added
- "Nothing to be done" message is output if everything is up-to-date
- Download failures and cksum failures are handled differently
- Handling CONFLICTS between installed packages and packages to
install are handled differently. Now package splits and similar
problems are resolved with a help of "fake" packages. Fake
packages are always marked for updates.
New option -f for "nih list" was added in order to list them
- Improvemed download progress messages
nih uninstall:
- Option -L was added, see "nih install -L"
- Options -p and -P were added. See "nih install"
nih leaf:
- Command was reimplemented and simplified
- Option -t option was removed
pkg_status doesn't have -ltRn options anymore
- Option -r was added
nih search:
- Option -I uses pkg_digger_installed provided by pkg_summary-utils
pkg_update_plan:
- Fixes in error messages
nih search|meta|info:
- Option -p was added. With a help of it one may search in
pkgsrc packages offline. Also see "nih refresh -p" and man page
Now nih works correctly if REPOSITORY is a directory
Fixes for shell's "trap" command.
Let to expand libtool archive (.la) automatically, so no need to mainain X11
file list manually anymore (every OS update may include shlib version change),
if the OS provide .la file (and no need to mind shlib type).
stage_upload_pkgs:
- delete outdated packages after uploading
- extra files are removed from remote repo
- no verbose flag for rsync
Serious fixes in stage_init
- PKG_UPDATE_SRC_SUMMARY_CMD: -m is passed to pkg_src_summary by default
Fix in signals handling
Fix in "make-depends" (unknown function is_true)
Fix: distbb -a option works correctly if pkg_summary.txt doesn't exist
upload_pkgs_all_files:
- directories and links are also output in addition to files
distbb checks arguments passed to -x|-X for correctness.
Configuraton variable SUDO was renamed to to DISTBB_SUDO.
This fixed build failure of ruby193-base
Configuraton variable OS was renamed to OPSYS
Stderr of pkg_summary2deps is now a part of the final report
(presumably only under Mac OS X Mountain Lion). "Xquarty" simply doesn't
print out anything if you run it with "-version".
This avoids a lot of warning messages from "bmake" while building packages
that use X11.
itself for certain if/elsif conditions. This combination of cleverness
and large conditional block should be replaced with simple,
table-driven code. But first we have to get the functionality under
test, which I've begun to do here.
Initialize a package global with the contents of $0, and override
it in test (otherwise self-grepping breaks because $0 is the test
script). Extract the dirty Test::Trap details to test_unit(). Test
main() for a few common values of @ARGV. Notice the aforementioned
cleverness because it broke an expected success, and fix it just
enough to test get_vartypes_basictypes() for a couple expected
types.
No functional change intended. I've manually tested the small changes
to pkglint.pl and have also verified that a couple packages lint
the same as before.
* minimally adapt pkglint(1) into a "modulino" for testability
* verify it still runs normally as a program
* create a test script with a few very simple test cases
* hook it up to 'make test'
* verify that the tests really fail if I go breaking the code under test
Meta-addresses PR pkg/46570. New BUILD_DEPENDS, but no functional
change, so no PKGREVISION bump. Approved by wiz@.
- Prefer fcntl FD_CLOEXEC to ioctl FIOCLEX.
- Reset SIGPIPE action to default for the build child, avoids 'yes' spinning
indefinitely on EPIPE in Solaris builds, notably for print/pdf2djvu.
configure tests for memset/memcpy, which are standard and haven't
needed checking in a long time. This makes the configure script
tolerate injection of -Werror.
Fix signed/unsigned mismatches in the sha2 and whirlpool code. This
package now passes -Wall with gcc45.
Bump the package version to 20121220, which is when I made these
changes.
pkg_info(1) - from Bug Hunting in PR pkg/47328. The following text is
taken from Bug's PR:
The pkg_info(1), pkg_admin(1), and pkg_install.conf(5) man pages
from pkgsrc' pkgtools/pkg_install package contain errors, have
inaccurate / incomplete / illogically ordered descriptions, have
incorrect macro usages / macro errors, miss an option description
(`-h', for pkg_info(1)), and / or contain other imperfections.
In short: these man pages can use a big correction update.
Additionally, the `usage' message in main.c from pkg_info(1) contains
an incorrect option order.
I originally was trying to patch set-but-not-used errors but then
quickly found out that a new glib also introduced numerous breakages
due to -Werror breaking on g_thread_init deprecation. Between the
two problems I had already generated 10 patches with no end in site.
Rather than continue, address the real issue and remove -Werror from
this old package.
pkg_online server defaults to dict://dict.mova.org:26280
"onlineinfo" has pkg_summary(5) raw format. Appropriate changes were
made for pkg_online_client(1) and pkg_summary2onlineinfo(1).
pkg_update_src_summary:
- Fix for the case when there are no updated/new packages, but
removals exist.
- Resulting summaeries are postprocessed by pkg_summary2bb_pkgs.
- Option -v was added
pkg_subgraph_deps:
- Fix for option -1
pkg_digger:
- Minor fix. -sf outputs both fields and strategies.
- All pkg_digger(1) backends return plain summaries on -1|-3|-9.
Option -r was removed from pkg_digger_{summary,installed}
pkg_digger_summary:
- Implementation for option -n.
pkg_summary2bb_pkgs:
- Additional error checks.
Man pages for pkg_digger, pkg_digger_summary, pkg_digger_installed
were updated.
pkg_update_src_summary has been completely reimplemented and now
works dramatically faster if files under mk/ directory were not
changed since previous update.
pkg_src_summary:
- PLIST field gathering was significantly speeded-up
- Fix. Option -p works correctly with non-empty PSS_SLAVES
- Fix. If -A is applied, -m is always enabled for recursive runs.
- Minor fix for -m.
- New option -F was added
- Minor fixes for PLIST and ALLDISTFILES
- PSS_PPERS was completely removed
cvs_checksum:
- checksums of CVS files are output by default
pkg_digger:
- pkg_digger_backend has been renamed to pkg_digger_summary,
man page was added for pkg_digger_summary
- new search backend pkg_digger_installed
pkg_list_all_pkgs:
- new options -a and -d were added
New utility makefiles2deps was added.
A lot of new regression tests, updates for man pages, minor code
clean-ups
machine, and machine_arch. Add a helper script to simplify use of
libkver. Set VERSION=0.7.
An earlier version of this change was reviewed by seb@NetBSD.org.
libXext/buildlink3.mk, now that it is included there.
Leave the places where its API version is set or variables from it
are used directly (about 3 packages).
There are at least two packages ("graphics/cairo" and "x11/xlsclients")
which require this. And we definitely want to avoid the situation where
both the based and "pkgsrc" version get used at the same time.
Bump all the packages which use the "x11/libxcb" package except the two
which already require version 1.6. For those package remove the manually
set minimum required version.
Change discussed with Thomas Klausner in private e-mail.
- PR/41321 - handle packages which generate multiple pkgnames from the same
pkgdir - e.g. "py{25,26,27}-sqlite3-0nb2.tgz"
- Add support for pkg_summary.bz2 and pkg_summary.txt. Preferred order is now
pkg_summary.bz2, pkg_summary.gz then pkg_summary.txt
- Drop support for old pkg_chk-summary file
Fix rare failures in "Clean" target. Thanks to David Holland!
New configuration variable SEND_REPORT_FIRST was introduced.
Boolean variables MASTER_MODE, CROSS_COMPILE, SEND_REPORT_FIRST
accepts values true|yes|1 for logical TRUE and false|no|0 for
logical FALSE. This may fix some hacks in single node setups.
During upload rsync may unexpectedly fail due to connection
failure. In order to resolve this problem, upload may be repeated
several times. The following variables were introduced for this
purpose: UPLOAD_PKGS_ATTEMPTS, UPLOAD_PKGS_PAUSE,
UPLOAD_LOGS_ATTEMPTS, UPLOAD_LOGS_PAUSE.
Progress messages now better illustrates internal structure of
the build process.
Option -a was added to distbb(1) in order to build updates for
quarter releases more efficiently.
stage_checks:
- double underline in "checks" filename means _ in the report
text, single underline means space.
- one more check for broken pkg_summary (REQUIRES field). It
checks for libraries under /tmp, /home, /srv and /work
directories.
stage_upload_logs:
- huge pkg_enriched_summary.txt is not uploaded
README was updated
Minor clean-ups
Fix gawk warning message appeared on Linux and other systems
nih install:
- commands like 'nih install _ package-' now works correctly
- much better messages are output when update plan failed to build
- option -p was added in order to skip REQUIRES/PROVIDES checks
'nih status -Aa' doesn't output status of uninstalled packages
Fixes and improvements in nih.1
A small shell script to set up a sandbox (usually for a pkgsrc bulk
build), using null mounts.
The file lived in pkgsrc/mk/bulk/mksandbox, but is worthy of its own
package IMHO. New: with manpage.
Modified default behaviour for Proceed [Y/n]
Sync wip with pkgsrc: pkgin 0.6.2
20120729
Rewrote narrow_match() using version_check() to avoid more
dirty repositories mistakes.
20120715
Preparing pkginteractive integration.
20120530
SQLite is now part of NetBSD 6.0 base, remove it from pkgin's tree.
Dependency will be added in the package.
- Fix PR#46748 Error occurs when pkg_chk is used with both "-g" & "-C conf"
pkg_chk was failing if run outside of a valid pkgsrc/$category/$pkg directory
- Merge in (adjusted) changes from Mike Volokhov to by default not downgrade
packages if installed version is larger than that in pkgsrc or packages.
Will still report, and can override with -B
that we prefer pkgsrc for OpenSSL. While the latter doesn't actually work
yet, it's better that we are deterministic about this and do not pick up
the system OpenSSL regardless.
There in no case on DragonFly where the X11_TYPE variable would correctly
be set to "native" because there is no native X11 option available.
Mask the package so it and the four packages that depend on it will never
attempt to build on DragonFly.
4.84:
- Misparse Makefiles a little better. Check for '.' in the left hand of :S and
:C operations, and also handle :U
- Move files from pkglint/files to lintpkgrsc/files
pkg_summary2deps:
- Fix in 'pkg_summary2deps -dn'
(missing pkgname if multivariant packages are present,
for example py26-gtk2 and py27-gtk2)
pkg_summary2leaves:
- Fix. 'pkg_summary2deps -dpnra' should be used
instead of 'pkg_summary2deps -dpa'
More regression tests for pkg_summary2leaves and pkg_summary2deps
Minor fix in pkg_uniq_summary.1
psu_funcs.awk:
- new function pkgpana2pkgpath
- Use the correct id(1) program. Fixes PR pkg/40691
- Avoid unprintable characters, some implementations of tr(1) have issues
with them. Instead go for what hopefully will be a unique string, as
well as making use of shell parameter substitution. Fixes PR pkg/42197
These changes combined make pkg_alternatives work correctly on Solaris.
Added SQLite build patch from sbd@
Added show-all-categories
unlink() before symlink()
20120525
Dynamically read PKG_DBDIR from pkg_admin
20120524
Do not delete repository entries if the new one is unreachable.
20120429
When a package is marked as TOREMOVE and TOUPGRADE because of a
dependency breakage during upgrade, just follow the usual upgrade
process.
20120427
Handled "no value field" for buggy packages.
20120419
Hopefuly fixed the "dirty repository" bug thanks to bapt
20120416
Added show-{pkg-,}category as requested by Julian Fagir
Added pkg-build-defs as requested by gls@
20120415
Added pkg-content and pkg-descr
Introduced pkg_infos.c to fetch remote informations
Removed -D_FILE_OFFSET_BITS=64 from Makefile and added it in fsops.c
20120414
Fire up pkgin 0.6.0
case: cleanup in summary.c (stacktic)
Packages list queries ordered differently for display and internals
This package would fail when a non-privileged user would try to install it.
That's because it was trying to create /usr/pkg/var/log directory. Another
directory was getting installed into DESTDIR but no further.
Added patch:
1) Removes attempt of Makefile to install ${VARBASE}/log
2) ..and directory ${VARBASE}/cache/PackageKit/downloads creation
3) ..which fixes non-root installations
The INSTALL file was removed, instead of installing directories in ${VARBASE},
it was installing directories at /. The functionality was replaced with
OWN_DIRS+= in the Makefile and verified to work as expected.
- Add the GENERATE_PKG_SUMMARY option to tell pkg_comp to generate a
pkg_summary.gz file in the packages directory. Useful to later be
able to run a packaging tool like pkgin against new binary packages.
Resolves at least some issues with finding the right pyNN-foo, when
replacing packages that don't match PYTHON_VERSION_DEFAULT.
(This was a stray change in my tree not committed for a while, but
it's been working ok.)
Now 0.24.1.
Rewrite OpenBSD.bsd.man.mk, based on current NetBSD.bsd.man.mk and changed to
use mandoc.
Fixes manual pages installation with packages using USE_BSD_MAKEFILE=yes
on OpenBSD-5.0.
* Move some MesaLib files from files/xorg to files/xfree.MesaLib,
or unwanted MesaLib may be buildlinked (and result in build failure).
* Add some files from OpenBSD-5.0.
- Allow specifying packages by their basename, without a category. This
affects the arguments to the 'build' command as well as the list of
packages provided to AUTO_PACKAGES and BUILD_PACKAGES.
===================================================================
-- Version 0.44.0 by Aleksey Cheusov, Sat, 25 Feb 2012 11:43:56 +0300
Improvements for report:
- "^WARNING:" messages are logged
- Segmentation faults are logged
- Clean-ups in report_mini.txt and report.txt
Packages are rebuilt recursively only if it necessary
-- Version 0.43.0 by Aleksey Cheusov, Mon, 9 Jan 2012 23:27:20 +0300
Fix for signal handlers (SIGINT, SIGTERM, SIGQUIT).
Fix in "Depends" stage.
Unnecessary dependencies are not install anymore.
Dependency graph. Latest available binary package is always used for
BUILD_DEPENDS if it satisfies dependencies.
