Fixes#43494
Add a set of environment variables SPACK_ALWAYS_CFLAGS (etc.) that
are always applied by the compiler wrapper.
Unlike SPACK_CFLAGS, for example, these will also be applied to
version checks (both SPACK_CFLAGS and SPACK_ALWAYS_CFLAGS will be
applied to the other invocation modes like ccld etc.).
Using this new functionality, the classic Intel and oneAPI compilers
are updated to pass compiler flags that disable warning messages
when newer versions are invoked via their older binary names
(these warnings were also generated for version checks, hence the
need for a new wrapper variable).
---------
Co-authored-by: Peter Josef Scheibel <scheibel1@llnl.gov>
For example: spack install py-pyqt5 or py-pyqt6 would install pylupdate5 pyrcc5 and
pyuic5 into in python-venv's install prefix.
Fix https://github.com/spack/spack/issues/45359
* Allow spec queries by `namespace`
Spack specs have "namespaces" that indicate what package repository they come from, but
there has not been a way to use the spec syntax to match one.
You can say things like this:
```console
spack find builtin.zlib
spack find myrepo.zlib
```
But, because namespaces are written as a dot-separated prefix on the name, you can't say
"find me all specs in namespace myrepo". The syntax doesn't allow it.
This PR allows you to specify anonymous specs with namespaces on the CLI. Specifically
you can do queries like this:
```console
spack find namespace=builtin
spack find namespace=myrepo
```
You can use this anywhere else you use spec syntax, e.g. in a config file to separate
installations based on what repo they came from:
```yaml
spack:
config:
install_tree:
root: $spack/opt/spack
projections:
namespace=myrepo: "myrepo_special_path/{name}-{hash}"
namespace=builtin: "builtin/{name}-{hash}"
```
This PR adds a special `namespace_if_anonymous` attribute to specs, which returns the
`namespace` if the spec has no name, otherwise it returns `None`. This allows us to
print the namespace for anonymous specs but to continue hiding it for most views, as
we've done so far.
This is implemented as a special case, but it's one that already exists, along with
`platform`, `os`, `target`, etc. This also reserves existing special case names for
variants so that users cannot define them in their package files. This change is
potentially breaking, but I do not think it will be common. There are no builtin
packages with a variant called `namespace`, and defining `os`, `target`, or `platform`
as a variant would've likely caused other problems if they were already being used.
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
* format: allow spaces in format specifiers
Key-value pair format specifiers can now contain spaces in the key. This allows us to
add spaces to format strings that are *only* present when the attribute formatted is not
``None``. Instead of writing:
```
{arch=architecture}
```
and special casing `arch=` like a sigil in `Spec.format()`, we can now write:
```
{ arch=architecture}
```
And the space is *only* printed when `architecture` is not `None`. This allows us to
remove the special case in `Spec.format()` for `arch=`.
Previously the only `key=` prefix allowed in format specifiers was `arch=`, but this PR
removes that requirement, and the `key=` part of a key-value specifier can be any name.
It does *not* have to correspond to the formatted attribute.
- [x] modify `SPEC_FORMAT_RE` to allow arbitrary keys in key-value pairs.
- [x] remove special case for `arch=` from `Spec.format()`.
- [x] modify format strings using `{arch=architecture}` to use `{ arch=architecture}`
- [x] add more tests for formatting
This PR saves other more complex attributes like compiler flags and their spacing for later.
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
VC toolset versions 144 and 143 are both associated with the platform
toolset 143; this deviates from prior version choices made by the
MSVC devs; add a special case to report platform toolset version
as 143 when VC toolset version is >= 143 (this will need revisiting
for later releases).
This was missed in #45394 because we don't run unit tests for package PRs, and
`test_all_virtual_packages_have_default_providers`, which would've caught it, is a unit
test, not an audit.
- [x] add a default provider for `wasi-sdk` in `etc/spack/defaults/packages.yaml` (which
we require for all virtuals)
- [x] rework `test_all_virtual_packages_have_default_providers` as an audit called
`_ensure_all_virtual_packages_have_default_providers`
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
The `Executable` class emits a warning when you pass quoted arguments to it:
```
> spack pkg grep '"namespace"'
==> Warning: Quotes in command arguments can confuse scripts like configure.
The following arguments may cause problems when executed:
"namespace"
Quotes aren't needed because spack doesn't use a shell. Consider removing them.
If multiple levels of quotation are required, use `ignore_quotes=True`.
```
This is to warn new package authors who aren't used to calling build commands in python.
It's not useful for `spack pkg grep`, where we really are passing args on the command
line, and if we pass a quoted string, we probably meant to.
- [x] make `ignore_quotes` an instance variable, not just an argument to ``__call__`
- [x] set `ignore_quotes` to `True` in `spack pkg grep`
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
* Several tests can be enabled on Windows with no changes to logic
(either the test logic or logic being tested)
* Test for `spack location` requires modification of the test logic,
but with a minor change can be enabled on Windows
The "use_store" context manager is used to swap the value
of a global variable (spack.store.STORE), while keeping
another global variable consistent (spack.config.CONFIG).
When doing that it tries to evaluate the previous value
of the store, if that was not done already. This is wrong,
since the configuration might be in an "intermediate" state
that was never meant to trigger side effects.
Remove that operation, and add a unit test to
prevent regressions.
Since the the MetaPathFinder now owns a lazily constructed RepoPath object, we need to deserialize environments before the package that needs to be restored. Before we were relying on globals to be inconsistent in a way that let the entire process go.
- Add explicit conflict on ppc64le for 12.5 and newer.
- Update/fix intel compiler conflict to reflect that intel@2021 is compatible
only since 11.4.1 and not since 11.1.1.
- Add intel compiler conflicts to reflect strict support matrix since
12.2.0.
If an external happens to be a provider of anything, the solver would
set its weight to 0, meaning that it is most preferred, even if
packages.yaml config disagrees.
That was done so that `spack external find mpich` would be sufficent to
pick it up as mpi provider.
That may have made sense for mpi specifically, but doesn't make sense
for other virtuals. For example `glibc` provides iconv, and is an
external by design, but it's better to use libiconv as a separate
package as a provider.
Therefore, drop this rule, and instead let users add config:
```
mpi:
require: [mpich]
```
or
```
mpi:
buildable: false
```
which is well-documented.
This change makes `spack external find --not-buildable` mark virtuals
provided by detected packages as non-buildable, so that it's sufficient
for users to let spack detect say mpich and have the concretizer pick it
up as mpi provider even when openmpi is "more preferred".
in case of inheritance the static tags prop may be updated multiple
times, and it turns out builder classes magically inherit from
traditional package classes
While trying to build packages with the OneAPI compiler version 2024.1 I
ran into the following error, indicating that the compiler is unable to
find some necessary libraries:
```
/storage/Software/oneapi/2024.1/compiler/2024.1/bin/sycl-post-link: error
while loading shared libraries: libonnxruntime.1.12.22.721.so: cannot open
shared object file: No such file or directory
icpx: error: unable to execute command: No such file or directory
icpx: error: sycl-post-link command failed due to signal (use -v to see
invocation)
```
Indeed, `libonnxruntime.1.12.22.721.so` does come bundled with the
OneAPI compiler, but it is not available in the build environment by
default. In this commit, I update the custom environment created by
OneAPI to include the `lib/` directory in which these libraries reside
in the `LD_LIBRARY_PATH` environment variable.
Originally if you had `x -> y -> z`, and an env with `x` in its speclist that is concretized but not installed, then `spack find -c y` would not show anything. This was intended: `spack find` has up-until-now only ever listed out installed specs (and `-c` was for adding a preamble section about roots).
This changes `spack find` so:
* `-c` makes it search through all concretized specs in the env (in a sense it is anticipated that a concretized environment would serve as a "speculative" DB and users may want to query it like they query the DB outside of envs)
* Adds a `-i/--install-status` option, equivalent to `-I` from `spack spec`
* Shows install status for either `-c` or `-i`
* As a side effect to prior point, `spack find -i` can now distinguish different installation states (upstream/external)
Examples:
```
$ spack find -r
==> In environment findtest
==> 1 root specs
- raja
==> 6 installed packages (not shown)
==> 12 concretized packages to be installed (not shown)
```
```
$ spack find
==> In environment findtest
==> 1 root specs
- raja
-- darwin-ventura-m1 / apple-clang@14.0.3 -----------------------
berkeley-db@18.1.40 bzip2@1.0.8 diffutils@3.10 gmake@4.4.1 gnuconfig@2022-09-17 libiconv@1.17
==> 6 installed packages
==> 12 concretized packages to be installed (show with `spack find -c`)
```
```
$ spack find -c
==> In environment findtest
==> 1 root specs
- raja
-- darwin-ventura-m1 / apple-clang@14.0.3 -----------------------
[+] berkeley-db@18.1.40 [+] bzip2@1.0.8 - cmake@3.29.4 [+] diffutils@3.10 [+] gmake@4.4.1 [+] libiconv@1.17 - nghttp2@1.62.0 - pkgconf@2.2.0 - readline@8.2
- blt@0.6.2 - camp@2024.02.1 - curl@8.7.1 - gdbm@1.23 [+] gnuconfig@2022-09-17 - ncurses@6.5 - perl@5.38.2 - raja@2024.02.2 - zlib-ng@2.1.6
==> 6 installed packages
==> 12 concretized packages to be installed
```
$ spack -E find
...
==> 82 installed packages
```
Fix a bug where Spec.tree with cover=nodes reduces deptypes from all
in-edges, including from nodes not reachable from the root, which almost
always happens for concrete specs
`Spec.__getitem__` queries dependent edges, which almost always point to
nodes outside the sub-dag considered. It should only ever look at edges
being traversed.
This modifies heuristic to decay to clingo default
over time. The hope is that this helps with specs
that have an optimal solution with a high penalty.
Let target and compiler heuristic decay too, do not
guess compiler
This adds spack.util.spack_yaml.anchorify, which takes a non-cyclic
dict/list structure, and replaces identical values with (back)
references to the first instance, so that yaml serialization will use
anchors.
`repr` is used to identify sub-dags, which in principle is quadratic
complexity in depth of the graph, but in practice the depth is O(1) so
this should not matter.
Then this is used in CI to reduce the size of generated YAML files to
30% of their original size.
In practice people don't care about having their build dependencies reinstalled with say cflags=-O3 if that is what is set at the input spec, so restrict propagation to link/run deps only.
Also simplify the encoding in asp.
Fix the following on Windows:
* `spack load --list` (this printed 0 packages even if packages were
loaded)
* `spack unload <package>` (this said that the package is not loaded
even if it was)
Update unit tests for `spack load` to also run on Windows (specifically
for ".bat"). This involved refactoring a few tests to parameterize
based on whether the unit tests are being run on a Windows system
(and to account for batch syntax).
When we changed how to deal with errors in November,
we didn't realize that for an unconstrained choice
rule it is more important in the heuristic to guess
what is NOT in the answer set, since it will be the
majority of options.
Previously this was following automatically from what
was in the answer set, via `1 { ... } 1` cardinality
constraints.
Here we improve the heuristic and the solve time for specs.
* Core change: logic for extracting RPATHs from modules may return
`None`: filter this out of the set of RPATHs that is auto-generated
* Core change: `CachedCMakePackage` no longer adds ldflags to
`CMAKE_STATIC_LINKER_FLAGS`: generally these flags are not appropriate
for static linking (e.g. invocation of `ar`)
* [mfem] Add version 4.7
* [mfem] Add variant for precision (single/double). Enforce consistency
for precision amongst mfem and hypre/petsc/mumps dependencies
* [mfem] Add cxxstd (and related constraints preventing use of
old cxxstd values for newer versions of some dependencies)
* [hypre] In line with prior point, added support for specifying
precision
* [petsc] Add config option to avoid error when building against
`superlu-dist+rocm`
* [hiop] add proper `raja`/`umpire`/`camp` version constraints for
`hiop` versions 0.3.99-0.4.x; require `+raja` for `+rocm`, and
add dependency on `hiprand` for `+rocm`
* [butterflypack, mfem, strumpack, suite-sparse] Require
`CRAYLIBS_{target-family}` env var to be defined
* [suite-sparse] versions `@7.4:` changed install location of headers:
add symlink from old location to new location
* [zlib-ng] Fix error where shared libs were not successfully built for
`%cce@17` (the build did not fail, but the finished `zlib-ng%cce@17`
install did not have shared libs)
* 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>
#40773 introduced python-venv, which improved build isolation and avoids issues with,
e.g., `ubuntu`'s system python modifying `sysconfig` to include a (very unwanted)
`local` directory within the default install layout.
This addresses a few cases where #40773 removed functionality, without harming the
default cases where we use `python-venv`.
Traditionally, *every* view with `python` in it was essentially a virtual environment,
because we would copy the `python` interpreter and `os.py` into every view when linking.
We now rely on `python-venv` to do that, but only when it's used (i.e. new builds) and
only for packages that have an `extends("python")` directive.
This again makes every view with `python` in it a virtual environment, but only
if we're not already using a package like `python-venv`. This uses a different
mechanism from before -- instead of using the `virtualenv` trick of copying `python`
into the prefix, we instead create a `pyvenv.cfg` like `venv` (the more modern way
to do it).
This fixes two things:
1. If you already had an environment before Spack `v0.22` that worked, it would
stop working without a reconcretize and rebuild in `v0.22`, because we no longer
copy the python interpreter on link. Adding `pyvenv.cfg` fixes this in a more
modern way, so old views will keep working.
2. If you have an env that only includes python packages that use `depends_on("python")`
instead of `extends("python")`, those packages will now be importable as before,
though they won't have the same level of build isolation you'd get with `extends`
and `python-venv`.
* views: avoid making client code deal with link functions
Users of views and ViewDescriptors shouldn't have to deal with link functions -- they
should just say what type of linking they want.
- [x] views take a link_type, not a link function
- [x] views work out the link function from the link type
- [x] view descriptors and commands now just tell the view what they want.
* python: simplify logic for avoiding pyvenv.cfg in copy views
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
In #30882, we made Spack ignore `-Werror` calls so that it could more easily build
projects that inject `-Werror` into their builds. We did this by translating them to
`-Wno-error` in the compiler wrapper. However, some compilers (like `nvhpc`) do not
support `-Wno-error`. We need to exclude them from this feature until they do.
- [x] make a property on `PackageBase` for `keep_werror` that knows not to use it for
`nvhpc`.
- [x] update property so that it keeps only the specific `-Werror=...` args for newer nvhpc's,
which support `-Wno-error` but not `-Wno-error=...`
---------
Co-authored-by: William Mou <william.mou1024@gmail.com>
Co-authored-by: Tom Scogland <scogland1@llnl.gov>
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
When Spack concretizes environments, it prints every (newly concretized) root spec
individually with all of its dependencies. For most reasonably sized environments, this
is too much output. This is true for three commands:
* `spack concretize` when concretizing an environment with newly added specs
* `spack install` when installing an environment with newly added specs
* `spack spec` with no arguments in an environment
The output dates back to before we had unified environments or nicer spec traversal
routines, and we can improve it.
This PR makes environment concretization output analogous to what we do for regular
specs. Just like `spack spec` for a single spec, we show all root specs with no
indentation, so you can easily see the specs you explicitly requested. Dependencies are
shown:
1. With indentation according to their depth in a breadth-first traversal starting at
the roots;
2. Only once if they appear on paths from multiple roots
So, the default is now consistent with `spack spec` for one spec--it's `--cover=nodes`.
i.e., if there are 100 specs in your environment, you'll get 100 lines of output.
If you want to see more details, you can do that with `spack spec` using the arguments
you're already familiar with. For example, if you wanted to see dependency types and
*all* dependencies, you could use `spack spec -l --cover=edges`. Or you could add
deptypes and namespaces with, e.g. `spack spec -ltN`.
With no arguments in an environment, `spack spec` concretizes (if necessary) and shows
the concretized environment. If you run `spack concretize` *first*, inspecting the
environment repeatedly with `spack spec` will be fast, as everything is already in the
`spack.lock` file.
- [x] factor most logic of `Spec.tree()` out of `Spec` class into `spack.spec.tree()`,
which can take multiple specs as roots.
- [x] make `Spec.tree()` call `spack.spec.tree()`
- [x] `spack.environment.display_specs()` now uses `spack.spec.tree()`
- [x] Update `spack concretize`
- [x] Update `spack install`
- [x] Update `spack spec` to call `spack.spec.tree()` for environments.
- [x] Continue to output specs individually for `spack spec` when using
`--yaml` or `--json`
`BuildcacheBootstrapper` uses `Spec.intersects` to match specs needed
for bootstrapping against the binary cache. The specs were not
sufficiently-detailed to prevent matching e.g. cached binaries for
Mac OS on Windows; this commit adds the platform to each requested
bootstrap spec to prevent that.
Remove support for `cray` as a separate platform.
Any platform previously detected as `cray` is now detected as `linux`.
Users who still need platform=cray have to stick to Spack 0.22
Fixes a bug in the concretizer where specs depending on a host
incompatible libc would be used. This bug triggers when nothing is
built.
In the case where everything is reused, there is no libc provider from
the perspective of the solver, there is only compatible_libc. This
commit ensures that we require a host compatible libc on any reused
spec, additionally to requiring compat with the chosen libc provider.
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
Change the installer to take `([pkg], args)` in the constructor instead
of `[(pkg, args)]`. The reason is that certain arguments are global
settings, and the new API ensures that those arguments cannot be
different across different "build requests".
The `explicit` install arg is now a list of hashes, and the installer is
no longer responsible for determining what package is installed
explicitly. This way environment installs can simply pass the list of
environment roots, without them necessarily being explicit build
requests. For example an env with two roots [a, b], where b depends on
a, would not always cause spack install to mark b as explicit.
Notice that `overwrite` already took a list of hashes, this makes
`explicit` consistent.
`package.do_install(explicit=True)` continues to take a boolean.
The windows wrappers for basic functions like `os.symlink`,
`os.readlink` and `os.path.islink` in the `llnl.util.symlink` module
have bugs, and trigger more file system operations on non-windows than
they should.
This commit just binds `llnl.util.symlink.symlink = os.symlink` etc so
built-in functions are used on non-windows
`spack clean <spec>` will now resolve specs based on the active environment if one is active.
If an env is active but no matching spec is found, this will fall back on fully concretizing.
Before this PR, if Spack could see a possibility to reuse a spec that
doesn't match a strong preference, it would do so. After the PR, a
strong preference would take precedence.
avoid calling `spec.target` when None.
When an external compiler package has an `os` set but no `target` set, Spack
currently falls into a codepath that calls `spec.target` (which itself calls
`spec.architecture.target.Microarchitecture`) when `spec.architecture.target`
is None, throwing an error.
e.g.
```
packages:
gcc:
externals:
- spec: gcc@12.3.1 os=rhel7
prefix: /usr
```
---------
Co-authored-by: Todd Gamblin <tgamblin@llnl.gov>
This fixes a bug occurring when two root specs need to select
old versions, and these versions have the same penalty in the
optimization. This sometimes caused an older version to be
preferred to a more recent one.
The issue was the omission of `PackageNode` in the optimization
tuple.
Use correct path separator in get_all_package_diffs for all platforms.
Ensures correct package change computation on Windows when pruning unchanged specs in Gitlab CI
This fixes an issue where ghcr, gitlab and possibly other container registries paginate tags by default, which violates the OCI spec v1.0, but is common practice (the spec was broken itself). After this commit, you can create build cache indices of > 100 specs on ghcr.
Co-authored-by: Harmen Stoppels <me@harmenstoppels.nl>
Add support for Gitlab CI on Windows
This PR adds the config changes required to configure and execute
Gitlab pipelines running Windows builds on Windows runners using
the existing Gitlab CI infrastructure (and newly added Windows
infrastructure).
* Adds support for generating child pipelines dispatched to Windows runners
* Refactors the relevant pre-scripts, scripts, and post scripts to be compatible with Windows
* Adds Windows config section describing Windows jobs
* Adds VTK as Windows build stack (to be expanded later)
* Modifies proj to build on Windows
* Refactors Windows rpath symlinking to avoid system libs and externals
---------
Co-authored-by: Ryan Krattiger <ryan.krattiger@kitware.com>
Co-authored-by: Mike VanDenburgh <michael.vandenburgh@kitware.com>
Co-authored-by: Todd Gamblin <tgamblin@llnl.gov>
Co-authored-by: Scott Wittenburg <scott.wittenburg@kitware.com>
Symlinks on Windows can use longpath prefixes (\\?\); these are fine
in the context of win32 API interactions but break numerous facets of
Spack behavior that rely on string parsing/matching (archiving,
binary distributions, tarball extraction, view regen, etc).
Spack's internal readlink method (llnl.util.symlink.readlink)
gracefully handles this by removing the prefix and otherwise behaving
exactly as os.readlink does, so we should prefer that in all cases.
Apparently urllib can throw a range of different exceptions:
1. HTTPError
2. URLError with e.reason set to the actual exception
3. TimeoutError from getresponse, which is not wrapped
* archive: relative links only
Ensure all links written into tarfiles generated from Spack prefixes do not contain symlinks pointing outside the prefix
* binary_distribution: limit extraction to prefix
Ensure files extracted from spackballs are not links pointing outside of the prefix
* Ensure rpaths are properly set on Windows
* hard error on extraction of absolute links
* refactor for non link-modifying approach
* Restore tarball extraction to original impl
* use custom readlink
* cleanup symlink module
* make lstrip
Add the ability to include any number of (potentially nested) concrete environments, e.g.:
```yaml
spack:
specs: []
concretizer:
unify: true
include_concrete:
- /path/to/environment1
- /path/to/environment2
```
or, from the CLI:
```console
$ spack env create myenv
$ spack -e myenv add python
$ spack -e myenv concretize
$ spack env create --include-concrete myenv included_env
```
The contents of included concrete environments' spack.lock files are
included in the environment's lock file at creation time. Any changes
to included concrete environments are only reflected after the environment
is re-concretized from the re-concretized included environments.
- [x] Concretize included envs
- [x] Save concrete specs in memory by hash
- [x] Add included envs to combined env's lock file
- [x] Add test
- [x] Update documentation
Co-authored-by: Kayla Butler <<butler59@llnl.gov>
Co-authored-by: Tamara Dahlgren <35777542+tldahlgren@users.noreply.github.co
m>
Co-authored-by: Todd Gamblin <tgamblin@llnl.gov>
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
Currently SPACK_COLOR=always is not respected in the build process on
macOS, because the global `_force_color` is re-evaluated in global scope
during module setup, where it is always `None`.
So, move global init bits from main.py to the module itself.
Some specs which were excluded from reuse,
are currently added back to the solve when
we traverse dependencies of other reusable
specs.
This fixes the issue by keeping track of what
we can explicitly reuse.
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.
Currently bootstrapping from source fails because clingo requires gnupg
requires clingo.
This commit stops eager bootstrapping. We don't need `patchelf` nor `gnupg`
generally. They're bootstrapped when needed.
This creates shared infrastructure for compiler packages to implement the
detailed search capabilities from the `spack compiler find` command for the
`spack external find` command.
After this commit, `spack compiler find` can be replaced with
`spack external find --tag compiler`, with the exception of mixed toolchains.
A named env cannot contain `.` and `/`.
So when a user runs `spack env create ./here` do not error but treat it
as `spack env create -d ./here`.
Also fix help string of `spack env create`, which seems to have been
copied from `activate` incorrectly.
Since reuse is the default now, `--reuse-deps` can be confusing, as it
technically does not imply roots are fresh.
So add `--fresh-roots`, which is also easier to discover when running
`spack concretize --fre<tab>`
We recently switched to using the new ReadTheDocs with "addons". That includes its own
analytics, which is nice, but we also want to continue using our GA4 analytics.
Adding GA4 is no longer supported by RTD, so we have to add it manually.
- [x] re-add the gtag to all pages, manually
Signed-off-by: Todd Gamblin <tgamblin@llnl.gov>
Adds a pre-concretization check for the Windows SDK and WGL (Windows
GL) packages as non-buildable externals.
This is a redo of https://github.com/spack/spack/pull/43459, but makes
sure to modify the configuration scope outside of the bootstrap scope:
whichever is highest-precedence in the user's environment at the time
the concretization runs, which should either be an env scope or the
~ scope.
Adds pytest fixture mocking the check for WGL and WSDK as if they were
present.
This PR gives users finer control over which specs are reused during concretization.
The value of the `concretizer:reuse` config option now can take an object with the following properties:
- `roots`: true if reusing roots, false if reusing just dependencies
- `exclude`: list of constraints used to select reusable specs
- `include`: list of constraints used to select reusable specs
- `from`: allows to select the sources of reused specs
### Examples
#### Reuse only specs compiled with GCC
```yaml
concretizer:
reuse:
roots: true
include:
- "%gcc"
```
#### `openmpi` must be used from externals, and it must be the only external used
```yaml
concretizer:
reuse:
roots: true
from:
- type: local
exclude:
- "openmpi"
- type: buildcache
exclude:
- "openmpi"
- type: external
include:
- "openmpi"
```
* PackageStillNeededError: add pkg that needs spec to exception msg
* PackageStillNeededError: f-string with short fmt and hash
* PackageStillNeededError: split long string
The old concretizer creates a cyclic graph when expanding virtuals for
`iconv`, which is a bug. This hack drops glibc and musl as possible
providers for `iconv` in the old concretizer to work around it.
Add debug log for external detection tests. The debug log
is used to print which test is being executed.
Skip version audit on Windows where appropriate
We run `extend spack_flags_list SPACK_LDFLAGS` for `$mode in ld|ccld`.
That's problematic, cause `ccld` needs `-Wl,--flag` whereas `ld` needs
`--flag` directly. Only `-L` and `-l` are common to compiler & linker.
In all build systems `LDFLAGS` is for the compiler not the linker, cause
any linker flag `-x` can be passed as a compiler flag `-Wl,-x`, and there
are many compiler flags that affect the linker invocation, like `-fopenmp`,
`-fuse-ld=`, `-fsanitize=` etc.
So don't pass `LDFLAGS` to the linker directly.
This way users can set `ldflags: -Wl,--allow-shlib-undefined` in compilers.yaml
to work around an issue where the linker tries to resolve the `libcuda.so.1`
stub lib which cannot be located by design in `cuda`.
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)
If there's no compiler we currently don't have any external libc for the solver.
This commit adds a fallback on libc from the current Python process, which works if it is dynamically linked.
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
When looking at where we spend our time in solver setup, I noticed a fair bit of time is spent
in `Spec.format()`, and `Spec.format()` is a pretty old, slow, convoluted method.
This PR does a number of things:
- [x] Consolidate most of what was being done manually with a character loop and several
regexes into a single regex.
- [x] Precompile regexes where we keep them
- [x] Remove the `transform=` argument to `Spec.format()` which was only used in one
place in the code (modules) to uppercase env var names, but added a lot of complexity
- [x] Avoid escaping and colorizing specs unless necessary
- [x] Refactor a lot of the colorization logic to avoid unnecessary object construction
- [x] Add type hints and remove some spots in the code where we were using nonexistent
arguments to `format()`.
- [x] Add trivial cases to `__str__` in `VariantMap` and `VersionList` to avoid sorting
- [x] Avoid calling `isinstance()` in the main loop of `Spec.format()`
- [x] Don't bother constructing a `string` representation for the result of `_prev_version`
as it is only used for comparisons.
In my timings (on all the specs formatted in a solve of `hdf5`), this is over 2.67x faster than the
original `format()`, and it seems to reduce setup time by around a second (for `hdf5`).
The reverse provider lookup may have stale entries for deleted packages, which used to cause errors. It's hard to invalidate those cache entries, so this commit simply drops entries w/o invalidating the cache.
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
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.
Some logic to detect what libc the c / cxx compilers use by default,
based on `-dynamic-linker`.
The function `compiler.default_libc()` returns a `Spec` of the form
`glibc@x.y` or `musl@x.y` with the `external_path` property set.
The idea is this can be injected as a dependency.
If we can't run the dynamic linker directly, fall back to `ldd` relative
to the prefix computed from `ld.so.`
In the future we may transform the database from a single JSON object to
a stream of JSON objects.
This paves the way for constant time writes and constant time rereads
when only O(1) changes are made. Currently both are linear time.
This commit gives just enough forward compat for Spack to produce a
friendly error when we would move to a stream of json objects, and a db
would look like this:
```json
{"database": {"version": "<something newer>"}}
```
* compiler wrapper: prioritize spack managed paths in search order
This commit partitions search paths of -L, -I (and -rpath) into three
groups, from highest priority to lowest:
1. Spack managed directories: these include absolute paths such as
stores and the stage dir, as well as all relative paths since they
are relative to a Spack owned dir
2. Non-system dirs: these are for externals that live in non-system
locations
3. System dirs: your typical `/usr/lib` etc.
It's very easy for Spack to known the prefixes it owns, it's much more
difficult to tell system dirs from non-system dirs. Before this commit
Spack tried to distinguish only system and non-system dirs, and failed
for very trivial cases like `/usr/lib/x/..` which comes up often, since
build systems sometimes copy search paths from `gcc -print-search-dirs`.
Potentially this implementation is even faster than the current state of
things, since a loop over paths is replaced with an eval'ed `case ...`.
* Trigger a pipeline
* Revert "Trigger a pipeline"
This reverts commit 5d7fa863de.
* remove redudant return statement
* Later versions of oneAPI have moved, so update detection to find it
in both old and new location
* Remove reliance on ONEAPI_ROOT env variable when determining Fortran
compiler version for %msvc
* When finding a Fortran compiler for MSVC, there was logic enforcing
a maximum MSVC version for a given oneAPI Fortran version. This
mapping was out of date and excluding valid combinations, so has
been removed (the logic now just picks the latest available
oneAPI Fortran compiler for any given MSVC version).
On Windows, bootstrapping logic now searches for and adds the win-sdk
and wgl packages to the user's top scope as externals if they are not
present.
These packages are generally required to install most packages with
Spack on Windows, and are only available as externals, so it is
assumed that doing this automatically would be useful and avoid
a mandatory manual step for each new Spack instance.
Note this is the first case of bootstrapping logic modifying
configuration other than the bootstrap configuration.
