Currently, if a package has a dependency from another repository and patches it,
generation of the patch cache will fail. Concretization succeeds if a fixed patch
cache is in place.
- [x] don't assume that patched dependencies are in the same repo when indexing
- [x] add some test fixtures to support multi-repo tests.
---------
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
Co-authored-by: Todd Gamblin <tgamblin@llnl.gov>
If a package `foo` doesn't implement `libs`, the default was to search recursively for `libfoo` whenever asking for `spec[foo].libs` (this also happens automatically if a package includes `foo` as a link dependency).
This can lead to some strange behavior:
1. A package that is normally used as a build dependency (e.g. `cmake` at one point) is referenced like
`depends_on(cmake)` which leads to a fully-recursive search for `libcmake` (this can take
"forever" when CMake is registered as an external with a prefix like `/usr`, particularly on NFS mounts).
2. A similar hang can occur if a package is registered as an external with an incorrect prefix
- [x] Update the default library search to stop after a maximum depth (by default, search
the root prefix and each directory in it, but no lower).
- [x]
The following is a list of known changes to `find` compared to `develop`:
1. Matching directories are no longer returned -- `find` consistently only finds non-dirs,
even at `max_depth`
2. Symlinked directories are followed (needed to support max_depth)
3. `find(..., "dir/*.txt")` is allowed, for finding files inside certain dirs. These "complex"
patterns are delegated to `glob`, like they are on `develop`.
4. `root` and `files` arguments both support generic sequences, and `root`
allows both `str` and `path` types. This allows us to specify multiple entry points to `find`.
---------
Co-authored-by: Peter Scheibel <scheibel1@llnl.gov>
Variants can now be propagated from a dependent package to (transitive) dependencies,
even if the source or transitive dependencies have the propagated variants.
For example, here `zlib` doesn't have a `guile` variant, but `gmake` does:
```
$ spack spec zlib++guile
- zlib@1.3%gcc@12.2.0+optimize+pic+shared build_system=makefile arch=linux-rhel8-broadwell
- ^gcc-runtime@12.2.0%gcc@12.2.0 build_system=generic arch=linux-rhel8-broadwell
- ^gmake@4.4.1%gcc@12.2.0+guile build_system=generic arch=linux-rhel8-broadwell
```
Adding this property has some strange ramifications for `satisfies()`. In particular:
* The abstract specs `pkg++variant` and `pkg+variant` do not intersect, because `+variant`
implies that `pkg` *has* the variant, but `++variant` does not.
* This means that `spec.satisfies("++foo")` is `True` if:
* for concrete specs: `spec` and its dependencies all have `foo` set if it exists
* for abstract specs: no dependency of `spec` has `~foo` (i.e. no dependency contradicts `++foo`).
* This also means that `Spec("++foo").satisfies("+foo")` is `False` -- we only know after concretization.
The `satisfies()` semantics may be surprising, but this is the cost of introducing non-subset
semantics (which are more useful than proper subsets here).
- [x] Change checks for variants
- [x] Resolve conflicts
- [x] Add tests
- [x] Add documentation
---------
Co-authored-by: Gregory Becker <becker33@llnl.gov>
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
* `find(..., max_depth=...)` can be used to control how many directories at most to descend into below the starting point
* `find` now enters every unique (symlinked) directory once at the lowest depth
* `find` is now repeatable: it traverses the directory tree in a deterministic order
Remove the constraint for concrete specs and simply take the
max(version) if a version is not given. This should default to the
highest infinity version which is also the logical best guess for
doing development.
* Remove concrete verision constriant for develop, set docs
* Add unit-test
* Update lib/spack/docs/environments.rst
Co-authored-by: kwryankrattiger <80296582+kwryankrattiger@users.noreply.github.com>
* Update lib/spack/spack/cmd/develop.py
Co-authored-by: Greg Becker <becker33@llnl.gov>
* Consolidate env collection in cmd
* Style
---------
Co-authored-by: kwryankrattiger <80296582+kwryankrattiger@users.noreply.github.com>
Co-authored-by: Greg Becker <becker33@llnl.gov>
The current `Spec.splice` model is very limited by the inability to splice specs that
contain multiple nodes with the same name. This is an artifact of the original
algorithm design predating the separate concretization of build dependencies,
which was the first feature to allow multiple specs in a DAG to share a name.
This PR provides a complete reimplementation of `Spec.splice` to avoid that
limitation. At the same time, the new algorithm ensures that build dependencies
for spliced specs are not changed, since the splice by definition cannot change
the build-time information of the spec. This is handled by splitting the dependency
edges and link/run edges into separate dependencies as needed.
Signed-off-by: Gregory Becker <becker33@llnl.gov>
Continuing the work started in #40326, his changes the structure
of Variant metadata on Packages from a single variant definition
per name with a list of `when` specs:
```
name: (Variant, [when_spec, ...])
```
to a Variant definition per `when_spec` per name:
```
when_spec: { name: Variant }
```
With this change, everything on a package *except* versions is
keyed by `when` spec. This:
1. makes things consistent, in that conditional things are (nearly)
all modeled in the same way; and
2. fixes an issue where we would lose information about multiple
variant definitions in a package (see #38302). We can now have,
e.g., different defaults for the same variant in different
versions of a package.
Some notes:
1. This required some pretty deep changes to the solver. Previously,
the solver's job was to select value(s) for a single variant definition
per name per package. Now, the solver needs to:
a. Determine which variant definition should be used for a given node,
which can depend on the node's version, compiler, target, other variants, etc.
b. Select valid value(s) for variants for each node based on the selected
variant definition.
When multiple variant definitions are enabled via their `when=` clause, we will
always prefer the *last* matching definition, by declaration order in packages. This
is implemented by adding a `precedence` to each variant at definition time, and we
ensure they are added to the solver in order of precedence.
This has the effect that variant definitions from derived classes are preferred over
definitions from superclasses, and the last definition within the same class sticks.
This matches python semantics. Some examples:
```python
class ROCmPackage(PackageBase):
variant("amdgpu_target", ..., when="+rocm")
class Hipblas(ROCmPackage):
variant("amdgpu_target", ...)
```
The global variant in `hipblas` will always supersede the `when="+rocm"` variant in
`ROCmPackage`. If `hipblas`'s variant was also conditional on `+rocm` (as it probably
should be), we would again filter out the definition from `ROCmPackage` because it
could never be activated. If you instead have:
```python
class ROCmPackage(PackageBase):
variant("amdgpu_target", ..., when="+rocm")
class Hipblas(ROCmPackage):
variant("amdgpu_target", ..., when="+rocm+foo")
```
The variant on `hipblas` will win for `+rocm+foo` but the one on `ROCmPackage` will
win with `rocm~foo`.
So, *if* we can statically determine if a variant is overridden, we filter it out.
This isn't strictly necessary, as the solver can handle many definitions fine, but
this reduces the complexity of the problem instance presented to `clingo`, and
simplifies output in `spack info` for derived packages. e.g., `spack info hipblas`
now shows only one definition of `amdgpu_target` where before it showed two, one of
which would never be used.
2. Nearly all access to the `variants` dictionary on packages has been refactored to
use the following class methods on `PackageBase`:
* `variant_names(cls) -> List[str]`: get all variant names for a package
* `has_variant(cls, name) -> bool`: whether a package has a variant with a given name
* `variant_definitions(cls, name: str) -> List[Tuple[Spec, Variant]]`: all definitions
of variant `name` that are possible, along with their `when` specs.
* `variant_items() -> `: iterate over `pkg.variants.items()`, with impossible variants
filtered out.
Consolidating to these methods seems to simplify the code a lot.
3. The solver does a lot more validation on variant values at setup time now. In
particular, it checks whether a variant value on a spec is valid given the other
constraints on that spec. This allowed us to remove the crufty logic in
`update_variant_validate`, which was needed because we previously didn't *know* after
a solve which variant definition had been used. Now, variant values from solves are
constructed strictly based on which variant definition was selected -- no more
heuristics.
4. The same prevalidation can now be done in package audits, and you can run:
```
spack audit packages --strict-variants
```
This turns up around 18 different places where a variant specification isn't valid
given the conditions on variant definitions in packages. I haven't fixed those here
but will open a separate PR to iterate on them. I plan to make strict checking the
defaults once all existing package issues are resolved. It's not clear to me that
strict checking should be the default for the prevalidation done at solve time.
There are a few other changes here that might be of interest:
1. The `generator` variant in `CMakePackage` is now only defined when `build_system=cmake`.
2. `spack info` has been updated to support the new metadata layout.
3. split out variant propagation into its own `.lp` file in the `solver` code.
4. Add better typing and clean up code for variant types in `variant.py`.
5. Add tests for new variant behavior.
Change the signature of Spec.from_detection to set the
external prefix, and the external modules, if they are
present.
Delete "spack.package_prefs.spec_externals" since it
is unused.
* Add options for sparse checkout in GitFetcher
Newer versions of git have a beta feature called sparse checkout
that allow users to check out a portion of a large repo.
This feature will be ideal for monolithic repo projects that want to
model their infrastructure via spack. This PR implements an addition
to the GitFetcher that allows users to add a `git_sparse_paths`
attribute to package classes or versions which will then use sparse
checkout on those directories/files for the package.
* Style
* Split git clone into multiple functions
* Add sparse-checkout impl
* Internalize src clone functions
* Docs
* Adding sparse clone test
* Add test for partial clone
* [@spackbot] updating style on behalf of psakievich
* Small fixes
* Restore default branch status
* Fix attributes for package
* Update lib/spack/docs/packaging_guide.rst
Co-authored-by: Matthew Mosby <44072882+mdmosby@users.noreply.github.com>
* Extend unit test to multiple git versions
* style
---------
Co-authored-by: psakievich <psakievich@users.noreply.github.com>
Co-authored-by: Matthew Mosby <44072882+mdmosby@users.noreply.github.com>
Extracted from #45638
When adding the "detectable" tag to a package class that has the
"tag" attribute inherited from a base class, we need to copy it to
avoid modifying the base class.
* Preserve higher weight for CLI git ref versions
Currently the concretizer fails if you reuse a git ref version
that has already been installed but modify the spec at all.
See #38484 for futher diagnosis
The issue here is that since there is no established provenance for
these versions the highest weight they are currently assigned is
that of prior install. Re-use checks then fail because the weight of
the version is identical to the solver.
Ironically, these versions are given the highest weights possible when
specified on the CLI for the first time. They should only appear in a
DAG if they are an exact match or if the user specifies them at the CLI.
Therefore it makes sense to preserve their higher ordering.
Getting this right is critical to moving all branch based versions to a pinned
git-ref in the future.
* [@spackbot] updating style on behalf of psakievich
* Update lib/spack/spack/solver/asp.py
Co-authored-by: Greg Becker <becker33@llnl.gov>
* Add provenance specific to git ref installs
* Sensitvity to name that I could not track down
* Add regression test
* Adjust test
* Add prefer standard unit-test
* Style
* Add required mock
* Format and mark
* Make unit-test case reproduce CLI investigation
* Remove unnecessary mock package
* [@spackbot] updating style on behalf of psakievich
* Use already developed fixture
* Add zlib-ng to mocks again
* Remove accidental adds
* Remove maintainer
* [@spackbot] updating style on behalf of psakievich
* Rename test file
* [@spackbot] updating style on behalf of psakievich
* Remove unused imports
* Update tests
* [@spackbot] updating style on behalf of psakievich
* Style
* Update lib/spack/spack/test/concretize.py
Co-authored-by: Greg Becker <becker33@llnl.gov>
* Update solver rule
* Duplicate installed rules for installed_git_version
* Revert "Duplicate installed rules for installed_git_version"
This reverts commit 17223fc8d1.
---------
Co-authored-by: psakievich <psakievich@users.noreply.github.com>
Co-authored-by: Greg Becker <becker33@llnl.gov>
This commit adds a layer of indirection to improve build isolation with
and without external Python, as well as usability of environment views.
It adds `python-venv` as a dependency to all packages that `extends("python")`,
which has the following advantages:
1. Build isolation: only `PYTHONPATH` is considered in builds, not
user / system packages
2. Stable install layout: fixes the problem on Debian, RHEL and Fedora where
external / system python produces `bin/local` subdirs in Spack install prefixes.
3. Environment views are Python virtual environments (and if you add
`py-pip` things like `pip list` work)
Views work whether they're symlink, hardlink or copy type.
This commit additionally makes `spec["python"].command` return
`spec["python-venv"].command`. The rationale is that packages in repos we do
not own do not pass the underlying python to the build system, which could still
result in incorrectly computed install layouts.
Other attributes like `libs`, `headers` should be on `python` anyways and need no change.
Some packages can't be redistributed in source or binary form. We need an explicit way to say that in a package.
This adds a `redistribute()` directive so that package authors can write, e.g.:
```python
redistribute(source=False, binary=False)
```
You can also do this conditionally with `when=`, as with other directives, e.g.:
```python
# 12.0 and higher are proprietary
redistribute(source=False, binary=False, when="@12.0:")
# can't redistribute when we depend on some proprietary dependency
redistribute(source=False, binary=False, when="^proprietary-dependency")
```
To prevent Spack from adding either their sources or binaries to public mirrors and build caches. You can still unconditionally add things *if* you run either:
* `spack mirror create --private`
* `spack buildcache push --private`
But the default behavior for build caches is not to include non-redistributable packages in either mirrors or build caches. We have previously done this manually for our public buildcache, but with this we can start maintaining redistributability directly in packages.
Caveats: currently the default for `redistribute()` is `True` for both `source` and `binary`, and you can only set either of them to `False` via this directive.
- [x] add `redistribute()` directive
- [x] add `redistribute_source` and `redistribute_binary` class methods to `PackageBase`
- [x] add `--private` option to `spack mirror`
- [x] add `--private` option to `spack buildcache push`
- [x] test exclusion of packages from source mirror (both as a root and as a dependency)
- [x] test exclusion of packages from binary mirror (both as a root and as a dependency)
This commit differentiate linux from other platforms by
using libc compatibility as a criterion for deciding
which buildcaches / binaries can be reused. Other
platforms still use OS compatibility.
On linux a libc is injected by all compilers as an implicit
external, and the compatibility criterion is that a libc is
compatible with all other libcs with the same name and a
version that is lesser or equal.
Some concretization unit tests use libc when run on linux.
The lack of a rule to avoid enforcing requirements on multi-valued variants, when the condition activating the environment was not met, resulted in multiple optimal solutions. The fix is to prevent imposing a requirement if the when= rule activating it is not met.
Fix two separate problems:
1. We want to always visit parents before children while creating views
(when it comes to ignoring conflicts, the first instance generated in
the view is chosen, and we want the parent instance to have precedence).
Our preorder traversal does not guarantee that, but our topological-
order traversal does.
2. For copy style views with packages x depending on y, where
<x-prefix>/foo is a symlink to <y-prefix>/foo, we want to guarantee
that:
* A conflict is not registered
* <y-prefix>/foo is chosen (otherwise, the "foo" symlink would become
self-referential if relocated relative to the view root)
Note that
* This is an exception to [1] (in this case the dependency instance
overrides the dependent)
* Prior to this change, if "foo" was ignored as a conflict, it was
possible to create this self-referential symlink
Add tests for each of these cases
Explicitly requested namespaces are annotated during
the setup phase, and used to retrieve the correct package
class.
An attribute for the namespace has been added for each node.
Currently, a single namespace per package is allowed
during concretization.
Before this PR, variant were not propagated to leaf nodes that could accept
the propagated value, if some intermediate node couldn't accept it.
This PR fixes that issue by marking nodes as "candidate" for propagation
and by setting the variant only if it can be accepted by the node.
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
* Test that setup_run_environment changes to CC/CXX/FC/F77 are dropped in build env
* compilers set in run env shouldn't impact build
Adds `drop` to EnvironmentModifications courtesy of @haampie, and uses
it to clear modifications of CC, CXX, F77 and FC made by
`setup_{,dependent_}run_environment` routines when producing an
environment in BUILD context.
* comment / style
* comment
---------
Co-authored-by: Tom Scogland <scogland1@llnl.gov>
This PR makes it possible to select only a subset of virtual dependencies from a spec that _may_ provide more. To select providers, a syntax to specify edge attributes is introduced:
```
hdf5 ^[virtuals=mpi] mpich
```
With that syntax we can concretize specs like:
```console
$ spack spec strumpack ^[virtuals=mpi] intel-parallel-studio+mkl ^[virtuals=lapack] openblas
```
On `develop` this would currently fail with:
```console
$ spack spec strumpack ^intel-parallel-studio+mkl ^openblas
==> Error: Spec cannot include multiple providers for virtual 'blas'
Requested 'intel-parallel-studio' and 'openblas'
```
In package recipes, virtual specs that are declared in the same `provides` directive need to be provided _together_. This means that e.g. `openblas`, which has:
```python
provides("blas", "lapack")
```
needs to provide both `lapack` and `blas` when requested to provide at least one of them.
## Additional notes
This capability is needed to model compilers. Assuming that languages are treated like virtual dependencies, we might want e.g. to use LLVM to compile C/C++ and Gnu GCC to compile Fortran. This can be accomplished by the following[^1]:
```
hdf5 ^[virtuals=c,cxx] llvm ^[virtuals=fortran] gcc
```
[^1]: We plan to add some syntactic sugar around this syntax, and reuse the `%` sigil to avoid having a lot of boilerplate around compilers.
Modifications:
- [x] Add syntax to interact with edge attributes from spec literals
- [x] Add concretization logic to be able to cherry-pick virtual dependencies
- [x] Extend semantic of the `provides` directive to express when virtuals need to be provided together
- [x] Add unit-tests and documentation
* Add tests to ensure variant propagation syntax can round-trip to/from string
* Add a regression test for the bug in 35298
* Reconstruct the spec constraints in the worker process
Specs do not preserve any information on propagation of variants
when round-tripping to/from JSON (which we use to pickle), but
preserve it when round-tripping to/from strings.
Therefore, we pass a spec literal to the worker and reconstruct
the Spec objects there.
GitLab's .patch URLs only provide abbreviated hashes, while .diff URLs
provide full hashes. There does not seem to be a parameter to force
.patch URLs to also return full hashes, so we should make sure to use
the .diff ones.
This patch adds in a license directive to get the ball rolling on adding in license
information about packages to spack. I'm primarily interested in just adding
license into spack, but this would also help with other efforts that people are
interested in such as adding license information to the ASP solve for
concretization to make sure licenses are compatible.
Usage:
Specifying the specific license that a package is released under in a project's
`package.py` is good practice. To specify a license, find the SPDX identifier for
a project and then add it using the license directive:
```python
license("<SPDX Identifier HERE>")
```
For example, for Apache 2.0, you might write:
```python
license("Apache-2.0")
```
Note that specifying a license without a when clause makes it apply to all
versions and variants of the package, which might not actually be the case.
For example, a project might have switched licenses at some point or have
certain build configurations that include files that are licensed differently.
To account for this, you can specify when licenses should be applied. For
example, to specify that a specific license identifier should only apply
to versionup to and including 1.5, you could write the following directive:
```python
license("MIT", when="@:1.5")
```
* VTK: Add patch for python 3.8 support
* CI: Re-enable VisIt in CI
* Configure spec matrix for stack with VisIt
* Add pugixml dep for 8.2.0
* Make VTK and ParaView consistent on proj dep
* OpenMPI 3: provides MP support by default
* Add details on proj dep in ParaView
* Add python 3.8 to test mock repo
* Patches to get VisIt VTK interface
* CI: Disable VisIt with GUI in DAV
Windows executable paths can have spaces in them, which was leading to
errors when constructing Executable objects: the parser was intended
to handle cases where users could provide an executable along with one
or more space-delimited arguments.
* Executable now assumes that it is constructed with a string argument
that represents the path to the executable, but no additional arguments.
* Invocations of Executable.__init__ that depended on this have been
updated (this includes the core, tests, and one instance of builtin
repository package).
* The error handling for failed invocations of Executable.__call__ now
includes a check for whether the executable name/path contains a
space, to help users debug cases where they (now incorrectly)
concatenate the path and the arguments.
* The module-level skip for tests in `cmd.install` on Windows is removed.
A few classes of errors still persist:
* Cdash tests are not working on Windows
* Tests for failed installs are also not working (this will require
investigating bugs in output redirection)
* Environments are not yet supported on Windows
overall though, this enables testing of most basic uses of "spack install"
* Git repositories cached for version lookups were using a layout that
mimicked the URL as much as possible. This was useful for listing the
cache directory and understanding what was present at a glance, but
the paths were overly long on Windows. On all systems, the layout is
now a single directory based on a hash of the Git URL and is shortened
(which ensures a consistent and acceptable length, and also avoids
special characters).
* In particular, this removes util.url.parse_git_url and its associated
test, which were used exclusively for generating the git cache layout
* Bootstrapping is now enabled for unit tests on Windows
Use curly braces instead of quotes to enclose value or text in Tcl
modulefile. Within curly braces Tcl special characters like [, ] or $
are treated verbatim whereas they are evaluated within quotes.
Curly braces is Tcl recommended way to enclose verbatim content [1].
Note: if curly braces charaters are used within content, they must be
balanced. This point has been checked against current repository and no
unbalanced curly braces has been spotted.
Fixes#24243
[1] https://wiki.tcl-lang.org/page/Tcl+Minimal+Escaping+Style
A LazyReference object is a reference to an attribute of a
lazily evaluated singleton. Its only purpose is to let developers
use shorter names to refer to such attribute.
This class does more harm than good, as it obfuscates the fact
that we are using the attribute of a global object. Also, it can easily
go out of sync with the singleton it refers to if, for instance, the
singleton is updated but the references are not.
This commit removes the LazyReference class entirely, and access
the attributes explicitly passing through the global value to which
they are attached.
