| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
|
|
The alternate code path which uses POSIX_LOG_PTRACE is now defunct.
Also fix some lint.
|
|
|
|
> Target "cxx_static" requires the language dialect "CXX20".
|
|
Having the preformatted manual page checked in into version control
doesn't make sense. Replacing the placeholders is an ugly hack and leads
to ugly line breaks. If the preformatted manual page is needed, the
better approach is to render it at build time, after the placeholders
have been replaced.
|
|
Having duplicate expanded strings is not as bad as having missed
alternatives, it's only a question of performance.
|
|
|
|
|
|
|
|
|
|
GCC 7 is the lowest common denominator for the parts needed to get Firefox
build going again on NetBSD 8.
|
|
When running 'lintpkgsrc -I tmp/lp-85 -yro', lintpkgsrc couldn't find
the cache file due to the intermediate chdir($pkgdistdir).
|
|
|
|
|
|
|
|
|
|
gopls v0.9.3
Move fast and fix things... it appears that v0.9.2 exposed a latent crasher in
gopls (golang/go#54395). This crash occurs when a new workspace is incorrectly
configured, and so is likely to be rather confusing for new users. On that
basis, the fix warranted a patch release.
As always, thank you for filing issues!
gopls v0.9.2
This release contains many bug fixes, particularly related to problems that
would require restarting gopls.
Note about network usage: among these fixes was a change to allow network usage
when reloading the workspace. Reloading occurs when a go.mod, go.sum, or
go.work file changes on disk. In the past, gopls would only allow network
during the first workspace load. This resulted in confusing behavior when, for
example, a new dependency is added to a go.mod file and gopls could not load
it, but loading succeeded on restart.
|
|
platforms, but the pkgsrc package currently does not yet support
configuring the library for them.
|
|
|
|
* Sync with www/firefox-103.0.2.
|
|
|
|
|
|
CHangelog:
Fixed
* Fixed menu shortcuts for users of the JAWS screen reader.
* Fixed an occasional non-overridable certificate error when accessing device
configuration pages.
* Fixed an issue with Picture-in-Picture displaying in fullscreen on macOS.
|
|
I have no idea how to generate __config_site file and include it.
This patch is obviously wrong, however it fixes wasi-libcxx-14.0.6b build.
|
|
|
|
Changes since 2022.08.12:
In package makefiles, the ':U' modifier is evaluated more correctly than
before, handling the values '' and '0' correctly.
Running 'lintpkgsrc --help' produces a clean usage message.
Distinfo files are checked more strictly than before.
Running 'lintpkgsrc -D .' does not produce Perl warnings anymore when
falling back to running bmake to determine the PKGNAME.
When parsing package makefiles, the debug log for evaluating
conditionals is easier to read.
|
|
In Perl, 0 is falsy, but in bmake, the modifier ':U' only cares about
undefined vs. defined.
|
|
|
|
|
|
There are no distinfo files anymore in which the filename starts with
'./', so remove that check.
Complain about invalid distinfo lines. This flags a few pkgsrc-wip
packages, but none in main pkgsrc.
Extract load_distinfo into a separate subroutine, improve local variable
names.
Document that only the first distinfo algorithm (currently BLAKE2s) is
ever checked.
|
|
It was easily possible to confuse scan_pkgsrc_makefiles with
load_pkgsrc_makefiles, so rename the latter to load_pkgdb_from_cache.
|
|
There is no apparent reason for this code. Lintpkgsrc could already
parse the version of pkg_install, by falling back to running bmake.
When run in debug mode, 'lintpkgsrc -D .' failed because lintpkgsrc used
a wrong regular expression instead of just calling dirname. While here,
fix the variable name 'pkgsrcdir' to simply 'pkgdir', as the previous
name was highly confusing.
|
|
Having a variable named 'false' was confusing. Having a stack of
if-states in which 0 means true and 1 means false was even more
confusing. Using these magic numbers in the debug log without any
explanation was misleading.
The code, as well as the debug log, now uses the strings 'not_yet',
'active' and 'done' for the state of the conditionals. While it still
requires a bit of thought to read the debug log, it's at least possible
now.
|
|
|
|
|
|
|
|
- try to restore SunOS patches which were left obviously broken (untested)
- reduce compiler requirement
- delint
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Changes since 2022.08.10:
Improve parsing of conditions, continuation lines, comments in
makefiles.
Replace hard-coded paths to /etc/mk.conf and /usr/pkg with the proper
paths that correspond to the pkgsrc installation.
If pkglint is run with unknown or invalid options, exit with an error
status.
|
|
|
|
|
|
|
|
|
|
|
