When fixing the SUBST definitions in a package, it can hapen that the
substitution aborts in the middle. In such a case the cookie should not
be written and the substitution should be retried. Otherwise the build
may continue with half the substitutions done.
The severity now depends only on the setting of SUBST_NOOP_OK. Right now
this means that some former warnings will be reported as info only, but
that will change after switching the default of SUBST_NOOP_OK after
2020Q1. Then they will all be reported as warnings, followed by the final
error saying that the pattern has no effect.
This change makes it easier to detect inconsistencies and outdated
definitions, for example by setting the global SUBST_NOOP_OK=no and
redefining WARNING_MSG to actuall fail.
The diff program is only used to produce informative output in the build
logs, nevertheless its output might be translated if there are lines that
do not end with a newline.
This hack is mostly needed for older software like sysutils/daemontools
that was created when errno was still a global variable. Newer packages
like devel/bglibs don't need that hack anymore. Therefore make it
configurable, to avoid build failures when subst.mk does not find any
error.h to patch.
The variables are now sorted chronologically instead of alphabetically,
to express the order in which the actions happen. The sentences are a bit
shorter than before.
The visual layout has been changed to make the whole variable name stick
out more. Before, the part that was easily visible was the same "EXTRACT_"
for all of them.
Before, variables containing dollar characters displayed so wrong that it
was hard to explain.
To fix the problem, I typed almost random characters into the code until
the output was exactly as expected. I still do not understand:
* why the list variables need 8 dollars to survive the @x@ loop,
* why the code only works if the dollars come from an external variable
instead of being written inline,
* why the backslash in the :C modifier needs to be doubled.
Anyway, the output of "bmake show-all-extract" now contains the shell
variable $${extract_file}, just as it should. The dollars are now doubled
in the output and thereby match the source code from the Makefile
exactly.
This prevents unintended shell word expansion.
It also makes that argument visible in "bmake show-all-extract". Before,
it was left out entirely from the output. Now it is printed as an empty
string literal, which at least gives a hint that "there is something".
The default value of SUBST_MESSAGE is based on SUBST_FILES, and that
variable may use the :sh modifier to list files from WRKSRC, which may
not exist at load time.
In the case of pkglocaledir, the SUBST_FILES are generated by a shell
command. That command assumes that the WRKDIR already exists. Therefore
SUBST_FILES must be evaluated as late as possible.
See mk/configure/replace-localedir.mk; an example package that fails is
devel/gettext-tools.
In a bulk build with very strict settings (WARNING_MSG fails, as well as
no-op substitutions), it became clear that nearly all of the cases where
SUBST didn't replace anything were bugs in the package definition.
Most of them were just outdated, which is no surprise given that some
packages are already over 20 years old.
For backwards compatibility, SUBST_NOOP_OK defaults to "yes" right now.
After correcting the affected packages, the default will change to "no".
This file allowed one to in theory specify a wxGTK version, but only
two programs used it. Both now directly include a preferred version.
Generally, packages should use wxGTK30, unless there is a good reason
to use 28.
A major motivation earlier was not being able to install 28 and 30 at
once, but this has been solved.
As briefly discussed on pkgsrc-users.
In an attempt to resolve problems on SunOS, move EXTRACT_USING to the
early variables file, in the hopes that it will be defined before code
that checks it is executed.
When all files are skipped, the tool dependency is not added as well.
This allows packages to skip the check without defining the user-settable
variable CHECK_PERMS.
Since the portability checks are concerned about bashisms, it doesn't make
sense to apply them to programs that will later be run in bash anyway.
Suggested by gdt via private mail.
These files typically come with a corresponding Makefile.in file, and the
same portability issue is in the .in file as well. For building the
package it is only necessary to patch the .in file. Patching the .am file
as well would mean additional work for the pkgsrc package maintainer.
Nevertheless, such fixes should be reported upstream. The upstream
maintainer is more interested in a patch for Makefile.am, since the .in
file is only generated.
Suggested by wiz via private mail.
Before, file names that start or end with spaces were trimmed by read(1),
as specified by POSIX 2004 and POSIX 2018.
https://pubs.opengroup.org/onlinepubs/9699919799/utilities/read.html
This edge case bug had been present since the very beginning of this
code, in November 2006.
Filenames containing backslashes or newlines are still not supported, and
probably never will be intentionally supported.
Native tar on NetBSD <=8 (and later, if MKBSDTAR is no) does not
handle archive formats increasingly being seen in pkgsrc. bsdtar
("libarchive tar") does, and is natively provided on NetBSD >=9.
pkgsrc already has a mechanism to use /bin/tar as the bsdtar tool when
it is bsdtar, so this change should not cause anything different on
NetBSD >= 9. On <=8, it will depend on archivers/bsdtar, which can
be built without using an extract tool, and which will then be broadly
usable.
As broadly discussed over multiple days on tech-pkg.
This is unnecessary, becuse extract/extract.mk has the same default
assignment. It is harmful, because it prevents setting a value in
platform/Foo.mk.
Change the comment to say this is platform specific (which will be
true soon).
Up to now, these files had been skipped because they didn't start with a
#! line. Still, they contain shell programs, therefore it makes sense to
check them.
The added empty line separates this help topic from the header comment of
this file, since USE_BSD_MAKEFILE does not add any dependencies and
therefore doesn't fit that description.
Since several years, the portability check only checks for shell files.
Instead of opening each extracted file to see whether it is a shell file,
it is more efficient to filter the files based on their names. This saves
a few system calls and in the case of lang/gcc9, half of the total
running time.
The file extensions to be excluded from the check have been sorted by
their count in the PLISTs. This data does not accurately represent the
source files of the package, but it's better than nothing.
Since pkgtools/check-portability is not installed by default, it will not
harm any existing pkgsrc setup. To activate it, it has to be installed
manually. After a testing phase, it will be installed whenever
PKG_DEVELOPER is set.
GFORTRAN_VERSION should match CC_VERSION as closely as possible for
ABI compatibility. This update tries to match GFORTRAN_VERSION
to CC_VERSION if the base compiler is GCC. If base compiler is not
GCC, default to a mainstream version likely to work with base clang/llvm.
This allows easy inspection of the realistic environment in which the
configure scripts are run.
The code is copied from the similar build-env target.
This might help to find the cause for pkg/54894, where "gcc -dumpversion"
is said to output 0 as the version number.
Running "/lib/x86_64-linux-gnu/libc.so.6 --version" on Debian unstable
inside WSL says:
GNU C Library (Debian GLIBC 2.29-9) stable release version 2.29.
In this case, there is a simple period, not a comma after the version
number.
Enlightened Sound Daemon was one of the earlier solutions to the old
"multiple programs can't open /dev/audio at once" problem that was once
a thing we had to worry about.
Eventually, it was adopted as part of GNOME. GNOME lost interest in it
about a decade ago and dropped it in favour of PulseAudio, newer
applications are generally uninterested in supporting it. Last release
was in 2008 and support for newer OS APIs is pretty nonexistent.
Several years ago the original website disappeared.
https://en.wikipedia.org/wiki/Enlightened_Sound_Daemonhttps://tracker.debian.org/news/999428/removed-0241-11-from-unstable/
I am under the impression we use _THING to mean "defined by the
implementation", which would be similar to the C meaning of __ prefix,
rather than "private to this file".
Before, not only files containing an RCS Id were recorded in the
+BUILD_VERSION file but also files containing text that looked similar to
an RCS Id were recorded, even though these didn't contain any valuable
version information.
The effect was that before this change, pkgtools/pkglint was built over
and over again by the bulk builds since pbulk uses a different regular
expression for detecting modified files.
The regular expression for unexpanded RCS Ids is added to record files
that have never been checked in to CVS, just to have them recorded and to
distinguish them from the final committed version.
See https://mail-index.netbsd.org/tech-pkg/2020/01/11/msg022489.html.
Before, not only files containing an RCS Id were recorded in the
+BUILD_VERSION file but also files containing text that looked similar to
an RCS Id were recorded, even though these didn't contain any valuable
version information.
The effect was that before this change, pkgtools/pkglint was built over
and over again by the bulk builds since pbulk uses a different regular
expression for detecting modified files.
The regular expression for unexpanded RCS Ids is added to record files
that have never been checked in to CVS, just to have them recorded and to
distinguish them from the final committed version.
See https://mail-index.netbsd.org/tech-pkg/2020/01/11/msg022489.html.
digests to identify file versions. Defaults to no for now. For digest
mode, recursive into FILESDIR, so that e.g. the majority of pkg_install
itself is recorded as well SMF manifests.
Starting from somewhere around GHC 7.8, Cabal installs packages with a
hashed package key instead of just "{NAME}-{VERSION}". In other words,
the pair of the plain package name and the version is no longer unique
in the package DB, and using it for uninstallation may also remove
packages that we didn't mean to remove.
This is paricularly problematic because GHC comes with several bundled
Cabal packages. Installing and uninstalling a package with the same
name could break GHC itself, if the uninstallation is performed
without hashed keys.
