that these directories will be conditionally removed (based on reference
counts), regardless of the value of PKG_CONFIG. Bump the PKGREVISION
for packages that were modified as a result.
comment. Instead, allow arbitrary comments at the beginning of the file.
This change required a major rewrite of the code, which had been quite
inflexible before.
code, I had forgotten one assumption I made. It was that in the ifelse
chain the case ( && ) couldn't occur because
the outer loop already cought that. And then I had inserted code that
violated exactly this assumption. Fixed.
The new code, which is not as nice as the old code, will hopefully allow
the --autofix option to be implemented really easily. If that will not
become true, I will probably revert to the old code. There's an
indentation FIXME in the code to keep this patch small. I will remove
that FIXME in a minute.
them. All pathnames are already interpreted as being relative to the
package's directory. Added a reminder that ONLY_FOR_PLATFORM should be
changed to NOT_FOR_PLATFORM where appropriate.
Fix the automagic tuning of the number of unified diff context
lines. It was broken since previous commit to files/pkgdiff.
Also arguments '-u' and '-U NUM' to GNU diffutils 2.8.1, as seen
in NetBSD 2.0_STABLE and current, are not intended to be used
together. Hence only '-U NUM' is now specified. I did not reverted
to the '-u -NUM' form used before as it seems undesirable.
Please note that if PKGDIFF_FMT is specified in the environment,
for testing or debugging purposes, we skip the automagic tuning of
the number of unified diff context lines. The user probably wants
to try another diff format, as suggested in the commit log of
revision 1.12 of files/pkgdiff. So there is no point in adding '-U
NUM' to PKGDIFF_FMT: it may just not work. See:
$ diff -c -U 1 /dev/null /dev/null
diff: conflicting output style options
diff: Try `diff --help' for more information.
While here save a few basename(1) runs.
- add -m, -o, and -r flags to optionally specify the machine-architecture
operating system, and os version to be displayed in the generated README
file. From Greg Woods via private email.
- White space fixes from Greg Woods.
- Clean up the generated README file as well as several of the messages
to reflect if the collection is a CD-ROM or DVD set. Some patches from
Greg Woods, more from me.
- List the tools used at runtime.
measuring the time spent in the main checking subroutines. Using that
function I discovered that method calls like $line->text are awfully
slow. The subroutine checkfile_PLIST spent half of its time in this
function call. Now it is almost twice as fast as before.
references are needed. As the return value of the checkfile_*
subroutines is not needed, it is not provided anymore. Some other small
changes, which all do not affect the actual output or behavior of
pkglint.
Pull in a patch from Vsevolod Stakhov via FreeBSD:
Because that the 55-bytes input-message becomes just 64 bytes
block after `the message terminator code(0x80)' and
`size-field(8-bytes field)' are added. Perhaps, I think the
block of a lot of one be made at this time.
Fixes PR 30857
