Age | Commit message (Collapse) | Author |
|
This reverts commit b233aceb92aed4fcdd6a918647c32c915d4fcc71 while
preserving bd3f96a5171cf797bf8ba59601a1a767ed036a05.
jq, even with -e, does not exit non-zero when given an empty input. This
is apparently a known feature:
https://github.com/stedolan/jq/issues/1497
https://github.com/stedolan/jq/issues/1142
Thus, revert this oversimplification which leads to problems when curl
fails, or otherwise returns an empty response body.
Closes #29.
|
|
|
|
This makes the cache an implementation detail which we can freely
change version to version.
|
|
fixup: 0a6a4d0 Use exit status rather than string emptiness
- git-ls always returns 0 as exit status, so it can't be used
as a check. git-show does return an error status.
|
|
Closes #27.
|
|
|
|
|
|
parsing issue and missing function
|
|
- Error happens with repositories with '-' in the name.
ie: community-testing, the '<repo>-<arch>' string
gets parsed wrong (repo=community,arch=testing-x86_64).
- This patch will hold up at least until an arch with a '-'
in the name appears.
|
|
|
|
|
|
|
|
A second attempt at fixing https://bugs.archlinux.org/task/60027 --
using --shared is much simpler and results in better diskspace savings.
|
|
This reverts commit 33b433898ebd7771ca045338bfca7c910312970c.
|
|
Would be nice if git detected this for us, but sadly it doesn't.
|
|
|
|
While we're here, merge our 'git config' command into the clone.
|
|
just set GIT_DIR appropriately and always operate from the original $PWD.
As part of this, let's cleanup our error handling in initialize() a fair
bit.
|
|
No need to define a whole array for passing -q
|
|
Some packages, e.g. libsigc++-docs, needs to be properly hex-escaped.
|
|
Fixes #22.
|
|
|
|
This removes the use of a tempfile entirely, and only updates the cache
when `git ls-remote` succeeds.
Based on a PR by polygamma:
https://github.com/falconindy/asp/pull/20
|
|
|
|
Closes #17.
|
|
|
|
While we're at it, separate our help output into two sections, since
we're accumulating a growing list of subcommands which operate not on
packages, but on the meta level of the asp repo.
Closes #15.
|
|
We have our nice action__ prefix on all of our action disspatchers, so
let's just use that to our advantage.
While we're here, fix the return code when we encounter an ambiguous
action
|
|
This solves the problem of:
$ asp checkout foo
$ asp list-arches foo
warning: refname 'packages/packages/foo' is ambiguous.
x86_64
i686
Same for list-repos (duh, shared codepath)
|
|
|
|
|
|
bsdtar is common to all Arch systems, but other distros (debian-ish)
might only have tar installed by default. Sadly, bsdtar has slightly
nicer semantics in it's -s than GNU tar's transform when matchin
directories in the tarball.
|
|
This is an implementation detail -- no one should really need to know
or core which repo the package came from. Moreover, there's nothing you
can do with this information. It's conceivable that a package of the
same name could exist in both git repos, but this would be semantically
forbidden by Arch packaging policies, and not well handled by asp
regardless of showing the prefix here.
Fixes GH#14.
|
|
|
|
|
|
The curly braces in '${XDG_CACHE_HOME:-$HOME/.cache}/asp' seem to confuse the parser
|
|
|
|
|
|
Fixup how we do versioning for git builds, and fix the missing package
version in the manpage and the static version in the script.
|
|
|
|
|
|
|
|
|
|
|
|
completions might trigger tracking of a new package. it all goes to
stderr, but let's not crowd the output regardless.
|
|
|
|
|
|
mostly to make shellcheck happy
|
|
Update README.md
|
|
|