Previously 'package-transitive-supported-systems' would enter an
infinite loop over the cross-compilation tool chain if
%CURRENT-TARGET-SYSTEM was set.
* guix/packages.scm (package-transitive-supported-systems)[supported-systems-procedure]:
Pass explicit SYSTEM and TARGET parameters.
* tests/packages.scm ("supported-package? vs. %current-target-system"):
New test.
Fixes a regression introduced in
af57d1bf6c whereby tests would end up
building the world.
* guix/gexp.scm (mixed-text-file): Add #:guile parameter and honor it.
* tests/gexp.scm ("mixed-text-file"): Pass #:guile to 'mixed-text-file'.
("file-union"): Pass #:guile to 'file-union'.
("lower-object, computed-file"): Pass #:guile to 'computed-file'.
("lower-object, computed-file + grafts"): Likewise.
* tests/packages.scm ("origin->derivation, single file with snippet"):
Likewise.
* tests/profiles.scm ("profile-derivation, ordering & collisions"):
Likewise.
* guix/tests.scm (test-file): Likewise.
* guix/packages.scm (package-development-inputs): New procedure.
* guix/scripts/environment.scm (package-environment-inputs): Use it.
* tests/packages.scm ("package-development-inputs")
("package-development-inputs, cross-compilation"): New tests.
* doc/guix.texi (package Reference): Document it.
Fixes <https://issues.guix.gnu.org/50676>.
Reported by Mathieu Othacehe <othacehe@gnu.org>.
Commit a779363b6a was partially incorrect:
references passed to #:allowed-references or #:references-graphs *can*
be lowered as references to grafted elements. This is for example the
case when doing:
(computed-file "partition.img" exp
#:options `(#:references-graphs ,inputs))
Here INPUTS must be lowered as a reference to suitably grafted elements.
Failing to do that, the reference graph will not match the actual
INPUTS.
However, when building a package, those references must indeed refer
only to ungrafted packages. This commit preserves that by having build
systems pass #:graft? #f.
* guix/gexp.scm (lower-reference-graphs, lower-references): Remove uses
of 'without-grafting'. This reverts
a779363b6a.
* guix/build-system/cmake.scm (cmake-build, cmake-cross-build):
Pass #:graft? #f.
* guix/build-system/glib-or-gtk.scm (glib-or-gtk-build)
(glib-or-gtk-cross-build): Likewise.
* guix/build-system/gnu.scm (gnu-build, gnu-cross-build): Likewise.
* guix/build-system/meson.scm (meson-build, meson-cross-build): Likewise.
* guix/build-system/trivial.scm (trivial-build, trivial-cross-build):
Likewise.
* tests/gexp.scm ("lower-object, computed-file + grafts"): New test.
* tests/packages.scm ("trivial with #:allowed-references + grafts"): New
test.
Suggested by Maxime Devos <maximedevos@telenet.be>.
* guix/packages.scm (current-definition-location): New syntax parameter.
(define-public*): New macro.
(<package>)[definition-location]: New field.
(package-definition-location): New procedure.
* tests/packages.scm ("package-definition-location"): New test.
These macros are intended to be used in build phases.
More precisely, (assoc-ref %build-inputs "input") can be
replaced by #$(this-package-input "input") or #+(this-package-native-input
"native-input") as appropriate.
* guix/packages.scm
(package-input, package-native-input): New (unexported) procedures.
(this-package-input, this-package-native-input): New macros.
Signed-off-by: Mathieu Othacehe <othacehe@gnu.org>
Fixes a regression introduced in
7d873f194c.
Starting from 7d873f194c, running
guix build -s aarch64-linux sed
on an x86_64-linux machine would return an x86_64-linux machine, whereby
only the top derivation of the graph would be aarch64-linux while all
its dependencies would be x86_64-linux.
* guix/packages.scm (expand-input): Add 'system' parameter and honor it.
(bag->derivation, bag->cross-derivation): Pass SYSTEM to 'expand-input'.
* tests/packages.scm ("package-derivation, different system"): New test.
This reinstates pre-gexp behavior where 'expand-input' would explicitly
pass #:graft? #f in recursive calls, thereby preventing redundant calls
to 'bag-grafts'.
* guix/packages.scm (expand-input): Turn into a monadic procedure.
Lower INPUT when it's a package, passing #:graft? #f.
(bag->derivation, bag->cross-derivation): Adjust accordingly.
* tests/packages.scm ("search paths"): Adjust so BUILD aborts only when
passed the package of interest.
* guix/packages.scm (bag->derivation): Turn into a monadic procedure by
remove 'store' parameter and removing the call to 'store-lower'.
(bag->cross-derivation): Likewise.
(bag->derivation*): New procedure.
(package-derivation, package-cross-derivation): Use it instead of
'bag->derivation'.
* tests/packages.scm ("bag->derivation"): Change to monadic style.
("bag->derivation, cross-compilation"): Likewise.
This test goes from 58s to 10s wall-clock time.
Reported by Maxim Cournoyer <maxim.cournoyer@gmail.com>.
* tests/packages.scm ("fold-available-packages with/without cache"):
Remove 'find-duplicates'. Add 'list->set*' and use it instead of
'find-duplicates', 'delete-duplicates', and 'lset='.
Before this change, only plain directories, tar or zip archives were supported
as the source of a package for the GNU build system; anything else would cause
the unpack phase to fail. Origins relying on snippets would suffer from the
same problem.
This change adds the support to use files of the following extensions: .gz,
.Z, .bz2, .lz, and .xz, even when they are not tarballs. Files of unknown
extensions are treated as uncompressed files and supported as well.
* guix/packages.scm (patch-and-repack): Only add the compressor utility to the
PATH when the file is compressed. Bind more inputs in the mlet, and use them
for decompressing single files. Adjust the decompression and compression
routines.
[decompression-type]: Remove nested variable.
* guix/build/utils.scm (compressor, tarball?): New procedures. Move
%xz-parallel-args to the new 'compression helpers' section.
* tests/packages.scm: Add tests. Add missing copyright year for Jan.
* guix/build/gnu-build-system.scm (first-subdirectory): Return #f when no
sub-directory was found.
(unpack): Support more file types, including uncompressed plain files.
Fixes a bug whereby the presence of propagated inputs could lead to two
non-eq? but actually equal packages in a bag's inputs. The problem
would manifest itself when running, for instance:
guix build inkscape -d --with-graft=glib=glib-networking --no-grafts
The resulting derivation would differ due from that without
'--with-graft'. This was due to the fact that glib propagates libffi;
this instance of libffi was not rewritten even though other instances in
the graph were rewritten. Thus, glib would end up with two non-eq?
libffi instances, which in turn would lead to duplicate entries in its
'%build-inputs' variable.
Fixes <https://bugs.gnu.org/43890>.
* guix/packages.scm (package-mapping)[rewrite]: Remove call to 'cut?'
and call 'replace' unconditionally.
[replace]: Add 'cut?' case.
* tests/guix-build.sh: Add test combining '--no-grafts' and
'--with-graft'.
* tests/packages.scm ("package-input-rewriting/spec, identity")
("package-input-rewriting, identity"): New tests.
This is a followup to 18fa433bf5 and
<https://issues.guix.gnu.org/43508>.
* guix/packages.scm (derivation=?, input=?): New procedures.
(bag->derivation, bag->cross-derivation): Add calls to
'delete-duplicates'.
* tests/packages.scm ("package-derivation, inputs deduplicated"): New
test.
* guix/scripts/package.scm (transaction-upgrade-entry)[upgrade]: Add
'transform' parameter. Pass PKG through it. Use
'manifest-entry-with-transformations'.
Call 'options->transformation' to get the transformation procedure.
* tests/guix-package.sh: Add 'guix package -u' test.
* tests/packages.scm ("transaction-upgrade-entry, transformation options preserved"):
New test.
* doc/guix.texi (Invoking guix package): Mention that transformations
are preserved across upgrades.
(Package Transformation Options): Likewise.
* guix/packages.scm (package-input-rewriting): Add #:deep? and pass it
to 'package-mapping'.
[replacement-property]: New variable.
[rewrite]: Check it.
[cut?]: New procedure.
* tests/packages.scm ("package-input-rewriting"): Pass #:deep? #f and
ensure implicit inputs were not rewritten. Avoid 'eq?' comparisons.
("package-input-rewriting, deep"): New test.
* gnu/packages/guile.scm (package-for-guile-2.0, package-for-guile-3.0):
Pass #:deep? #f.
Previously, something like:
guix build glib --with-graft=glibc=glibc@2.29
would produce a result showing that rewriting rules were not applied to
libx11@1.6.A (a replacement).
* guix/packages.scm (package-mapping): Call REPLACE instead of PROC to
'replacement'.
* tests/packages.scm ("package-input-rewriting/spec, graft"): New test.
With this change, '--with-input', '--with-graft', etc. also apply to
implicit dependencies. Thus, it's now possible to do:
guix build python-itsdangerous --with-input=python-wrapper=python@2
or:
guix build hello --with-graft=glibc=glibc@2.29
Additionally, before, implicit inputs were not rewritten, which could
lead to duplicates in the output of 'bag-transitive-inputs' (packages
that are not 'eq?' but lead to the same derivation). This in turn would
lead to unnecessary rebuilds when using '--with-input' & co. This
change fixes it by ensuring even implicit inputs are rewritten.
Fixes <https://bugs.gnu.org/42156>.
* guix/packages.scm (package-input-rewriting/spec): Add #:deep?
defaulting to #true, and pass it to 'package-mapping'.
[replacement-property]: New variable.
[rewrite]: Check that property and set it on the result of PROC.
[cut?]: New procedure.
* tests/packages.scm ("package-input-rewriting/spec"): Ensure implicit
inputs were unchanged.
("package-input-rewriting/spec, partial match"): Pass #:deep? #f.
("package-input-rewriting/spec, deep")
("package-input-rewriting/spec, no duplicates"): New tests.
(package/inherit): Move before use.
* tests/guix-build.sh: Add tests.
* tests/scripts-build.scm ("options->transformation, with-graft"):
Compare dependencies by package name or derivation file name.
* doc/guix.texi (Defining Packages): Adjust accordingly.
* guix/packages.scm (build-system-with-package-mapping): New procedure.
(package-mapping): Add #:deep? and honor it.
* tests/packages.scm ("package-mapping"): Compare the direct inputs of
the bag of P0 and that of P1.
("package-mapping, deep"): New test.
This should help to debug test failures due to duplicated packages.
* tests/packages ("fold-available-packages with/without cache"): Print
duplicated packages.
Fixes <https://bugs.gnu.org/42327>.
Reported by Jan Nieuwenhuizen <janneke@gnu.org>.
This is a followup to f52fbf7094.
* guix/packages.scm (bag-transitive-inputs, bag-transitive-build-inputs)
(bag-transitive-host-inputs, bag-transitive-target-inputs): Parameterize
%CURRENT-SYSTEM in addition to %CURRENT-TARGET-SYSTEM.
* tests/packages.scm ("package->bag, sensitivity to %current-system"):
New test.
Fixes <https://bugs.gnu.org/41796>.
Reported by Jakub Kądziołka <kuba@kadziolka.net>.
* guix/packages.scm (input-graft): Add 'output' parameter and honor it.
Add OUTPUT to the cache key.
(input-cross-graft): Likewise.
(fold-bag-dependencies): Operate on inputs instead of nodes. Turn
VISITED into a vhash instead of a set. Pass PROC HEAD and OUTPUT
instead of just HEAD.
(bag-grafts): Adjust accordingly.
* tests/packages.scm ("package-grafts, dependency on several outputs"):
New test.
* tests/packages.scm: Use #:prefix instead of #:hide for (gcrypt hash).
This accomodates for 'sha512' syntax literal matches with Guile-Gcrypt 0.3.0,
which exports 'sha512' in addition to 'sha256'.
Fixes a bug whereby a bag's transitive dependencies would depend on the
global '%current-target-system' value.
Partly fixes <https://issues.guix.gnu.org/41182>.
* guix/packages.scm (bag-transitive-inputs)
(bag-transitive-build-inputs, bag-transitive-target-inputs):
Parameterize '%current-target-system'.
* tests/packages.scm ("package->bag, sensitivity to %current-target-system"):
New test.
Fixes <https://bugs.gnu.org/35872>.
Reported by Andy Tai <atai@atai.org>.
* guix/profiles.scm (list=?, manifest-entry=?): New procedures.
* guix/scripts/package.scm (transaction-upgrade-entry): In the '=' case,
use 'manifest-entry=?' to determine whether it's an upgrade.
* tests/packages.scm ("transaction-upgrade-entry, zero upgrades,
propagated inputs"): New test.
Fixes a regression introduced in
131f50cdc9 whereby the install/upgrade
message would not be displayed:
$ guix upgrade -n
2.1 MB would be downloaded:
/gnu/store/…-something-1.2
/gnu/store/…-its-dependency-2.3
This is because we'd directly abort from 'transaction-upgrade-entry' to
the build handler of 'build-notifier'.
* guix/scripts/package.scm (transaction-upgrade-entry): Call 'string=?'
expression in 'with-build-handler'.
* tests/packages.scm ("transaction-upgrade-entry, grafts"): New test.
* guix/scripts/package.scm (transaction-upgrade-entry): Add 'store'
parameter and use it instead of (%store). Export.
* tests/packages.scm ("transaction-upgrade-entry, zero upgrades")
("transaction-upgrade-entry, one upgrade")
("transaction-upgrade-entry, superseded package"): Adjust accordingly.
Previously, (supported-package? coreutils "armhf-linux")
with (%current-system) = "x86_64-linux" would return false. That's
because 'supported-package?' would traverse the x86_64 dependency graph,
which contains 'tcc-boot0', which supports x86 only.
Consequently, 'supported-package?' would match only 53 packages for
"armhf-linux" when running on x86, as is the case during continuous
integration.
* guix/packages.scm (package-transitive-supported-systems): Add an
optional 'system' parameter. Use 'mlambda' instead of 'mlambdaq' for
memoization.
(supported-package?): Pass 'system' to 'package-transitive-supported-systems'.
* tests/packages.scm ("package-transitive-supported-systems, implicit inputs")
("package-transitive-supported-systems: reduced binary seed, implicit inputs"):
Remove calls to 'invalidate-memoization!', which no longer work and were
presumably introduced to work around the bug we're fixing (see commit
0db65c168f).
* tests/packages.scm ("supported-package?"): Rewrite test to use only
existing system name since otherwise 'bootstrap-executable' raises an
exception.
("supported-package? vs. system-dependent graph"): New test.
Until now the caller couldn't tell the different between a spec like
"foo:out" and one like "foo". This change allows users to distinguish
between these two cases.
* gnu/packages.scm (specification->package+output): Disable output
membership test when OUTPUT = #f and SUB-DRV = #f.
* tests/packages.scm ("specification->package+output")
("specification->package+output invalid output")
("specification->package+output no default output")
("specification->package+output invalid output, no default"): New tests.
The switch to the reduced bootstrap broke build time assumptions made by
tests, notably the assumption that GNU-MAKE-BOOT0 was cheap to build.
This commit adjusts this to make these tests cheaper.
* gnu/packages/bootstrap.scm (%bootstrap-inputs-for-tests): New variable.
* guix/tests.scm (gnu-make-for-tests): New variable.
* tests/guix-environment.sh: Use GNU-MAKE-FOR-TESTS instead of
GNU-MAKE-BOOT0. Remove test with FINDUTILS-BOOT0.
* tests/guix-package-net.sh (boot_make): Use GNU-MAKE-FOR-TESTS.
* tests/packages.scm ("GNU Make, bootstrap"): Likewise.
* tests/profiles.scm ("profile-derivation relative symlinks, two entries"):
Likewise.
* tests/union.scm (%bootstrap-inputs): Remove.
("union-build"): Use %BOOTSTRAP-INPUTS-FOR-TESTS instead of
%BOOTSTRAP-INPUTS.
Previously the same package could appear several times if several
variables were bound to it, as is notably the case for "python"
currently. This, in turn, would lead to obnoxious "ambiguous package
specification" messages.
* gnu/packages.scm (generate-package-cache)[expand-cache]: Change RESULT
to RESULT+SEEN and adjust accordingly.
Call 'first' on the result of 'fold-module-public-variables*'.
* tests/packages.scm ("fold-available-packages with/without cache"):
Check for lack of duplicates in FROM-CACHE.
