| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
I intend to have xbs-abs have separate libdir / libexecdir s.
|
|
in_array wasn't defined
|
|
The have been at the current locations since 5c7b1e0 (2014-06-20).
|
|
This should have been in 314f2c9b1daac8c47d78754569a7310d0b77e22b.
|
|
|
|
|
|
in the light of my attempts to create a riscv64 parabola port, I would
like to see the following changes made to librechroot.
The patch removes the hard-coded arm cross arch checks in favour of a
more general approach, that works for more architectures. As a side
effect, this now also would behave correctly when creating x86 chroots
on arm, although why anyone would choose to do this is beyond me.
Reviewed-By: Luke Shumaker <lukeshu@parabola.nu>
[LS: Added quotes]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This has been sitting uncommitted in my checkout for a while.
This has been sitting uncommitted in my checkout for a while.
|
|
This has been sitting uncommitted in my checkout for a while.
|
|
This has been sitting uncommitted in my checkout for a while.
|
|
This has been sitting uncommitted in my checkout for a while.
|
|
Rationale: https://lists.parabola.nu/pipermail/dev/2017-October/005936.html
This makes get-rep a general-purpose tool for finding replacements and
providers of any package, not just blacklisted-ones. It works equally
well reading from full blacklist.txt lines, just pkgname lines, and
unwittingly; a single line with many pkgnames.
Some caveats for further discussion:
- This changes behavior from showing just one replacement/provider to
_all_ replacements and providers (including pkgname itself) in the repos.
- By "repos" I mean the repos the executing machine is configured to
access. This could be changed in the future with `expac --config`
if more package databases are desired (e.g. different
architecture).
- Results are shown using the "repo/pkgname ..." format rather than
just "pkgname ...". I think this makes the program more useful.
- Results aren't reordered to ensure that the first provider of
pkgname is pkgname itself. Order will be as expected otherwise, by
virtue of expac parsing repos in order.
|
|
Plenty of functions in libreblacklist are expected to print empty
lines, so as to signify that the archpkg at that line in the input has
an empty accompanying field.
Printing no newline and skipping to the next archpkg is therefore a
bug; one that would pass unnoticed by current tests. This is because
command substitution removes trailing whitespace. I.e. the following
is true:
[[ $(libreblacklist get-rep <<<'') == $(libreblacklist get-rep <<<$'\n') ]]
Process substitution and pipes preserve whitespace, but only the
latter work with #!/usr/bin/env roundup, so we use that.
|
|
The blacklist allows referring to a number of trusted bug trackers to
complement the information given in the reason field. However,
maintainable shorthand identifiers are used instead of full
URLs. Printing the latter based on the former is the job of this new
functionality.
This also introduces corresponding test cases and l10n strings
(together with translations for the `es` locale).
|
|
For about the same amount of code we can save us the hassle of messing
with sed expressions every time the number of fields changes in blacklist.txt.
by knowing the delimiter character(s) and the number of times it's
supposed to appear, a corresponding normalizing expression is
constructed on the fly.
|
|
the blacklist has had a couple extra fields (ref, id) for some time, but
`libreblacklist` was never made aware of the change.
this updates test/lib-blacklist-test.sh for the same reason, and adds
a test for comment stripping under `it_normalizes_correctly()`.
|
|
|
|
|
|
|
|
This changed because I added `set -e`. There's no test for this (yet)
because it would take too long for me to mock the server-size. I'll do it
eventually.
Also, this fixes HOOKPOSTRELEASE running in $WORKDIR/staging, which is the
wrong place for it to run.
|
|
I ran `msgmerge -UFi xbs.po ../xbs.pot` for each message domain, but used
git gui to only stage substansive changes; not any changes to
line-wrapping.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
There are two places where the punctuation was on the outside of the
end-quote; leading to syntax errors. There's one place where the msgid
was partially translated.
|
|
This means that the new `po/files.generate` target can be used to build the
`.pot` files without doing anything else; we can stop requiring translators
to install ruby-ronn. It also means that we can put the `.pot` files in
the source tarball, to make it even easier for translators.
Unfortunately, it does add the extra step of having to create another
symlink when adding a new locale.
|
|
|
|
I had
docdir = $(datarootdir)/doc
pkgdocdir = $(docdir)/libretools
# install docs to $(pkgdocdir)
but the GNU Coding Standards (July 25, 2016 revision), say I should be
doing
docdir = $(datarootdir)/doc/libretools
# install docs to $(docdir)
As the config.mk states, we aren't a GNU package. But let's still follow
their standards here, to follow the principle of least surprise.
|
