makepkg assumes that the remote git repo is named "origin" at several
places in its handling of git sources. It is possible to set the remote
repo name since git v2.30.0 (with bug fix for bare checkouts in v2.30.2).
Add "--origin=origin" to all git clone commands.
Signed-off-by: Allan McRae <allan@archlinux.org>
This feature makes Git VCS build inputs immutable by adding support for
pinning a Git checkout by a hash of its content using the deterministic
export functionality `git archive`.
This feature aids packagers by allowing them to use simple and
convenient refnames (instead of full commit hashes) in the `PKGBUILD`
while still preserving security implications of immutable build inputs
using a trusted cryptographic hash function of the content.
Previously VCS source downloads have been skipped for `--geninteg` and
`--source` as both options did not need a checkout. This commit changes
this behavior by forcing the download of all sources as integrity checks
and generation requires to have an up to date state.
Signed-off-by: Levente Polyak <anthraxx@archlinux.org>
Commit e017a5975c introduced the GITFLAGS
environmental variable. While ensuring the default of "--mirror" was
kept, there was a capitalisation mistake made. Handle the default for
GITFLAGS directly in the git clone command.
The default flag used to clone a git repository when using makepkg
is "--mirror". However, when working with huge repositories, the use
of different flags during cloning can allow an faster checkout. For
example, using "--filter=blob:none" allows for small checkouts, at
the expense of requiring downloads during the build stage if anything
but the HEAD commit is used for the build. In addition, this example
would serve as a replacement for the often requested (but broken)
addition of --depth=1.
Add support for the environment variable GITFLAG to pass flags for
the git clone command. Note that this overrides the default rather
than adding to it in order to prevent incompatibilities.
In commit 882e707e40 we changed message
output to go to stdout by default, unless it was an error. The plain()
function doesn't *look* like an error function, but in practice it was
-- it's used to continue multiline messages, and all in-tree uses were
for warning/error.
This is a problem both because we're sending output to the wrong place,
and because in some cases, we were performing error logging from a
function which would otherwise return a value to be captured in a
variable using command substution.
Fix this and straighten out the API by providing two functions: one for
continuing msg output, and one which wraps this by sending output to
stderr, for continuing error output.
Change all callers to use the second function.
If something like source=(..."#commit=") is used, e.g. due to failed
variable expansion, we try to check out an empty refspec as nothing at
all, and end up just running "git checkout". This happens because we
fail at variable expansion too -- so let's quote our variables properly
and make sure git sees this as an empty refspec, so it can error out.
Also make sure it is interpreted as a ref instead of a path.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
The right-hand side of the [[ ... = ... ]] keyword is an exception to
the general rule that quoting is unnecessary with [[
This is usually not a problem, e.g. in libmakepkg, lint_one_pkgname will
already fail if pkgname has an asterisk, but it certainly doesn't hurt
to be "more proper" and go with the spec; it is more dangerous in
repo-add, which can get caught in an infinite loop instead of safely
asserting there is no package named 'foo*'.
Reported-by: Rafael Ascensão <rafa.almas@gmail.com>
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
In order to cache sources offline, makepkg creates *two* copies of every
git repo. This is a useful tradeoff for network time, but comes at the
cost of increased disk space.
Normally, git can smooth this over automagically. Whenever possible, git
objects are hardlinked to save space, but this does not work when
SRCDEST and BUILDDIR are on separate filesystems.
When the repo in question is both very large (linux.git for example is
2.2 GB) and crosses filesystem boundaries, this results in a lot of
extra disk space being used; the most likely scenario is where BUILDDIR
is a tmpfs for bonus ouch.
git(1) has a builtin feature which serves this case handily: the
--shared flag will create the info/alternates file instructing git to
not copy or hardlink or create objects/packs at all, but merely look for
them in an external location (that being the source of the clone).
The downside of using shared clones, is that if you modify and drop
commits from the original repo, or simply delete the whole repo
altogether, you break the copy. But we don't care about that here,
because
1) the BUILDDIR copy is meant to be a temporary copy strictly derived
via PKGBUILD syntax from the SRCDEST, and must be able to be
recreated at any time,
2) if the SRCDEST disappears, makepkg will redownload it, thus restoring
the objects needed by the BUILDDIR clone,
3) if the user does non-default things like hacking on the BUILDDIR copy
then deleting and re-cloning the SRCDEST may result in momentary
breakage, but ultimately should be fine -- the unique objects they
created will be stored in the BUILDDIR copy.
While it's theoretically possible that upstream will force-push to
overwrite the base tree from which makepkg is building (which they
should not do), *and* the user deleted their SRCDEST which they should
not do, *and* they saved work in makepkg's working directory which they
should not do either...
... this is an unlikely chain of events for which we should not care.
Using --shared is therefore helpful in immediately useful ways and IMHO
has no actual downsides; we should use it.
An alternative implementation would be to use worktrees. I've rejected
this since it is essentially the same as shared clones, except adding
additional restrictions on the branch namespace, and could potentially
break existing use cases such as manually handling the SRCDEST in order
to share repositories with normal working copies.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Lookup the existence of matching functions for each protocol, and
fallback on the generic file handler. New source protocols can then be
added via thirdparty libmakepkg drop-ins without requiring modifications
to source.sh
Fixes FS#49076
Signed-off-by: Allan McRae <allan@archlinux.org>
As per https://lists.archlinux.org/pipermail/arch-general/2017-July/043876.html
git doesn't check that the tag name matches what an annotated tag object
*thinks* it should be called. This is a bit of a theoretical attack and
some would argue that we should always use commits since upstream can
legitimately change a tag, but nevertheless this can result in a
downgrade attack if the git download transport was manipulated or the
upstream repository hacked.
So, check the tag blob to make sure the tag actually matches the name we
used for `git checkout`.
This really should be fixed in git itself, rather than forcing all
downstream users of git verify-tag to implement their own checks, but
the git developers disagree, see the discussion surrounding
https://public-inbox.org/git/xmqqk2hzldx8.fsf@gitster.mtv.corp.google.com/
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
A git repository is marked as signed if it contains the query "signed"
as defined by https://tools.ietf.org/html/rfc3986
Adds two utility functions in util/source.sh.in to extract fragments and
queries, and modifies source/git.sh.in to use them.
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
