On 29/04/18 13:12, Eli Schwartz wrote:
On 04/28/2018 04:31 AM, Vladimir Panteleev wrote:
Currently, makepkg doesn't support (actually, mishandles) sources in a subdirectory of the package directory. Though there aren't many uses for such a feature, it is useful for an otherwise source-less package of Arch- or site-specific scripts, to be installed on one or a few systems, for which it doesn't make sense to have a separate source repository. In this case, it would be nice to either move the source files in a subdirectory, or allow pulling them from somewhere else on the filesystem.
Whether or not this is considered a valid usecase, the current behavior is broken (makepkg computes broken paths and attempts to operate on files that aren't there), so this probably ought to be fixed regardless. Also https://bugs.archlinux.org/task/39718
I'm not sure we have a specified behavior, but I'm personally of the opinion this should be prohibited. IMHO there is no reason to hide files in subdirectories when the entire directory with the PKGBUILD is meant to be the package sources etc.
If we did support it officially, we should make sure --source supports it too, and possibly decide whether we support this in the AUR as well.
An argument could be made BTW, that these source paths are in fact malformed and should possibly be file://$(pwd)/ urls which are downloaded via curl according to the intent of the PKGBUILD author (these would then be downloaded from a local URI). The logical extension of that line of thought, is we should add a lint_pkgbuild check with a more informative error when no protocol is specified, but a non-filename is used.
Yes, file:/// sources work fine. If it isn't specifically local:// then download_file is used, which does DLAGENTS lookup. We don't provide a "file" DLAGENTS protocol, which means we usually just silently exit that function without actually downloading anything... but nothing is stopping people from adding one in the PKGBUILD itself. curl supports downloading from file:///absolute/path/to/file
...
I *really* dislike the inconsistency of permitting subdirs for local files (implying we think this is a reasonable use case) but not permitting them for things we resolve as something to download.
I couldn't find a test suite for makepkg (only pacman and makepkg-template), so, here is the test case:
mkdir dir touch a dir/b /tmp/c cat > PKGBUILD <<'EOF' pkgname=dir-test pkgver=1 pkgrel=1 arch=('any') source=(a dir/b /tmp/c) md5sums=(SKIP SKIP SKIP) package() { cp "$srcdir"/* "$pkgdir/" ; } EOF
How would this handle the second case being, instead, local://dir/b ?
The proto=local semantic is used both for files that don't have a proto, and files which use the local:// proto. The local:// proto is *actually* used, believe it or not. It's how we specify that the user is expected to provide the file themselves, for e.g. AUR packages which require proprietary sources (like games hidden behind paywalls). The AUR checks the proto, to make sure that no-proto files which *are* expected to be provided together with the PKGBUILD are not missing.
...
Why don't people use, for the third case, DLAGENTS+=('file::/usr/bin/curl -C - -o %o %u'); source=('file:///tmp/c') ?
And what sources do we expect to have a guaranteed filesystem location, but are not supplied by a package? I consider cp during build/package to be perfectly adequate for files provided by some other package, and in fact I use this here:
Firstly, we do not consider the AUR when making development decisions. Also, I am thinking you are mixing up --source and --allsource. I am happy for a file:// protocol to be added. We already use it for dealing with local packages/repos in pacman, so it should be supported for pacman. In fact, I thought this had already been done, so was surprised when it did not work! As far as the justification for source=("foo/bar") type sources, I think reasonable arguments can be made for their inclusion. For VCS sources, I think a git submodule is a valid example. Also, consider gcc - if you download a source for (e.g.) gmp into a particular directory in its sources, it will compile it and build against that. And the simple example provided by OP of config files organised in different subdirectories relative to the PKGBUILD also seems reasonable. So, sources downloaded to subdirectories is a feature I would support implementing. However, I will not consider partial solutions - or more accurately, partial fixes that look like that will need partially reverted to implement the complete solution. On the top of my head, the complete solution covers: 1) files in subdirectories of the builddir. 2) standard sources 3) VCS sources 4) Sources with foo/bar::http:// type syntax 5) --allsource Allan