#8223 replaced regex-based makefile target parsing with an invocation of
"make -q". #8818 discovered that "make -q" can result in an error for some
packages.
Also, the "make -q" strategy relied on interpreting the error code, which only
worked for GNU Make and not BSD Make (which was deemed acceptable at
the time). As an added bonus, this implementation ignores the exit code and
instead parses STDERR for any indications that the target does not exist; this
works for both GNU Make and BSD Make.
#8223 also updated ninja target detection to use "ninja -t targets". This does
not change that behavior but makes it more-explicit with "ninja -t targets all"
This also adds tests for detection of "make" and "ninja" targets.
Fixes#9001#8289 added support for install_tree and copy_tree to merge into an existing
directory structure. However, it did not properly handle relative symlinks and
also removed support for the 'ignore' keyword. Additionally, some of the tests
were overly-strict when checking the permissions on the copied files.
This updates the install_tree/copy_tree methods and their tests:
* copy_tree/install_tree now preserve relative link targets (if the symlink in the
source directory structure is relative, the symlink created in the destination
will be relative)
* Added support for 'ignore' argument back to copy_tree/install_tree (removed
in #8289). It is no longer the object output by shutil.ignore_patterns: you pass a
function that accepts a path relative to the source and returns whether that
path should be copied.
* The openfoam packages (currently the only ones making use of the 'ignore'
argument) are updated for the new API
* When a symlink target is absolute, copy_tree and install_tree now rewrite the
source prefix to be the destination prefix
* copy_tree tests no longer check permissions: copy_tree doesn't enforce
anything about permissions so its tests don't check for that
* install_tree tests no longer check for exact permission matching since it can add
file permissions
- `imp` is deprecated and seems to have started having some weird
issues on certain Linux versions.
- In particular, the file argument to `load_source` is ignored on
arch linux with Python 3.7.
- `imp` is the only way to do imports in 2.6, so we'll keep it around for
now and use it if importlib won't work.
- `importlib` is the new import system, and it allows us to get
lower-level access to the import implementation.
- This consolidates all import logic into `spack.util.imp`, and make it
use `importlib` if it's avialable.
Replace use of `shutil.copytree` with `copy_tree` and `install_tree` functions in `llnl.util.filesystem`.
- `copy_tree` copies without setting permissions. It should be used to copy files around in the build directory.
- `install_tree` copies files and sets permissions. It should be used to copy files into the installation directory.
- `install` and `copy` are analogous single-file functions.
- add more extensive tests for these functions
- update packages to use these functions.
- dependency patching test didn't attempt to apply patches; just to see
whether they were on the spec.
- it applies the patch now and verifies that that patch was applied.
* Branch with the meson build-system
* Fix build_environment for dual loads and add create code
* Add documentation
* Fixed option list
* Update build_system_guess for meson
* Fixed documentation errors
* Added meson to build and configure and updated documentation
* fix typos
- cc cleanup caused a parsing regression in flag handling
- We added proper quoting to array expansions, but flag variables were
never actually converted to arrays. Old code relied on this.
This commit:
- Adds reads to convert flags to arrays.
- Makes the cc test check for improper space handling to prevent future
regressions.
- flags were prepended in reverse order to args, but this makes it hard
to see what order they'll be in on the final command line.
- add them in the order they'll appear to make cc easier to maintain.
- simplify code for assembling the command line
- fix separator used in SPACK_SYSTEM_DIRS test
- This corrects most of the issues found by shellcheck
- This also uses ':' as the delimiter for SPACK_SYSTEM_DIRS, for
consistency with other variables.
- filtering using sed causes most builds to slow down quite a bit, as the
compiler wrapper has to run sed many times, and *it* runs many times
- do the system directory parsing directly in bash
- Add tests to ensure that RPATHs are not added in cc mode, which can
cause some builds to fail.
- Change cc.py to use pytest style
- Instead of writing out all the flags, break the flags down into
variables so that it's easy to read what each test is supposed to
check. This should make cc.py more maintainable in the future.
- Adding -L and -Wl,-rpath to compile-only command lines ("cc mode" or
"-c") causes clang (if not also other compilers) to emit warnings that
confuse configure systems.
- Clang will print warnings about unused command-line arguments.
- This fix ensures that -L and -Wl,-rpath are not added if the compile
line is just building an object file with -c
- This also cleans up the cc script in several places.
Spack currently prepends include paths, library paths, and rpaths to the
compile line. This causes problems when a header or library in the package
has the same name as one exported by one of its dependencies. The
*dependency's* header will be preferred over the package's, which is not
what most builds expect. This also breaks some of our production codes.
This restores the original cc behavior (from *very* early Spack) of parsing
compiler arguments out by type (`-L`, `-I`, `-Wl,-rpath`) and reconstituting
the full command at the end.
`<includes> <other_args> <library dirs> <rpaths>`
This differs from the original behavior in one significant way, though: it
*appends* the library arguments so that dependency libraries do not shadow
those in the build.
This is safe because semantics aren't affected by *interleaving* `-I`, `-L`,
and `-Wl,-rpath` arguments with others, only with each other (so the order of
two `-L` args affects the search path, but we search for all libraries on the
command line using the same search path).
We preserve the following:
1. Any system directory in the paths will be listed last.
2. The root package's include/library/RPATH flags come before flags of the
same type for any dependency.
3. Order will be preserved within flags passed by the build (except system
paths, which are moved to be last)
4. Flags for dependencies will appear between the root flags and the system
flags, and the flags for any dependency will come before those for *its*
dependencies (this is for completeness -- we already guarantee this in
`build_environment.py`)
* Fix performance issue when compiling.
Spack was doing active wait when compiling, spoiling one core.
My fix consists in not setting any timeout for select, instead of
the previous 0 second.
* Fix comments about select.select timeout
- This was a nasty workaround due to the way our compiler wrappers used
to work. We don't want to have to modify our elfutils installation to
install libdwarf.
- Since cd9691de5, we no longer need this because the package will always
come before dependencies in our include order.
Spack currently prepends include paths, library paths, and rpaths to the compile line. This causes problems when a header or library in the package has the same name as one exported by one of its dependencies. The *dependency's* header will be preferred over the package's, which is not what most builds expect. This also breaks some of our production codes.
This restores the original cc behavior (from *very* early Spack) of parsing compiler arguments out by type (`-L`, `-I`, `-Wl,-rpath`) and reconstituting the full command at the end.
`<includes> <other_args> <library dirs> <rpaths>`
This differs from the original behavior in one significant way, though: it *appends* the library arguments so that dependency libraries do not shadow those in the build.
This is safe because semantics aren't affected by *interleaving* `-I`, `-L`, and `-Wl,-rpath` arguments with others, only with each other (so the order fo two `-L` args affects the search path, but we search for all libraries on the command line using the same search path).
We preserve the following:
1. Any system directory in the paths will be listed last.
2. The root package's include/library/RPATH flags come before flags of the same type for any dependency.
3. Order will be preserved within flags passed by the build (except system paths, which are moved to be last)
4. Flags for dependencies will appear between the root flags and the system flags, and the flags for any dependency will come before those for *its* dependencies (this is for completeness -- we already guarantee this in `build_environment.py`)
- previously, output could be confusing when deptypes were only shown for
one dependent when a node had *multiple* dependents
- also fix default coverage of `Spec.tree()`: it previously defaulted to
cover only build and link dependencies, but this is a holdover from
when those were the only types.
- Previously, Spack didn't check the arguments you put in version()
directives.
- So, you could do something like this, where there are arguments for a
URL fetcher AND for a git fetcher:
version('1.0', md5='abc123', git='https://foo.bar', commit='feda2343')
- Now, we check the arguments before constructing a fetcher, to ensure
that each package has *only* arguments for a single type of fetcher.
- Also added `test_package_version_consistency()` to the `package_sanity`
test, so that all builtin packages are required to have valid
`version()` directives.
- packagers can specify two top-level fetch URLs if one is `url`
- e.g., `url` and `git` or `url` and `svn`
- allow only one VCS fetcher so we can differentiate between URL and VCS.
- also clean up fetcher logic and class structure
- Packages can remove the top-level `url` attribute and still work
- These are now legal:
- Packages with *only* version-specific URLs (even with gaps)
- Packages with a top-level git/hg/svn attribute and `version`
directives for that.
- If a package has both a top-level hg/git/svn attribute AND a top-level
url attribute, the url attribute takes precedence.
Some packages do not have a `url` and are instead downloaded via `git`,
`hg`, or `svn`. Some packages like `spectrum-mpi` cannot be downloaded at
all, and are placeholder packages for system installations. Previously,
`__init__()` in `PackageBase` crashed if a package did not have a `url`
attribute defined.
I hacked this section of code out, but I have no idea what the
repercussions of that are.
- This hard-codes the hash lengths rather than computing them on import.
- Also cleans up the code in `spack.util.crypto` to make it easier to
understand.
Fix this output error:
```
$ spack -m module loads mpileaks
==> Error: `spack module loads -m t -m c -m l ...` has been moved. Try this instead:
$ spack module t loads mpileaks
$ spack module c loads mpileaks
$ spack module l loads mpileaks
```
In case a deprecated form of the module command is used, the program
will exit non-zero and print an informative error message suggesting
which command should be used instead.
As requested in the review all the commands meant to manage module
files have been grouped under the `spack module` command.
Unit tests have been refactored to match the new command structure.
fixes#2215fixes#2570fixes#6676fixes#7281closes#3827
This PR reverts the use of `spack module loads` in favor of
`spack module find` when loading module files via Spack. After this PR
`spack load` will accept a single spec at a time, and will be able
to interpret correctly the `--dependencies` option.
fixes#4400
The feature requested in #4400 was already part of the module file
configuration, but it was neither tested nor documented. This
commit takes care of adding a few lines in the documentation and a
regression test.
This just because the fixture has been moved one level above the one
it was originally defined. In this more general context there's more
than one configuration file that could be patched for tests.
'spack module' has been split into multiple commands, each one tied to a
specific module type. This permits the specialization of the new
commands with features that are module type specific (e.g. set the
default module file in lmod when multiple versions of the same package
are installed at the same time).
- repo membership test was broken by the refactor of spack/__init__.py
- refactor singleton so that 'spec in repo' works again for `spack.repo.path`
- fix spec command and add basic tests for `spack spec` and `spack spec --yaml`
- There was a lot of documentation in `PackageBase` dating back to the
very first versions of Spack.
- It was repetitive and out of date, and the docs at spack.readthedocs.io
are better.
- Remove the outdated specifics, and leave the minimal useful set of
developer docs in `package.py`.
- This changes `get_checksums_for_versions` to generate code that uses an
explicit `sha256` argument instead if the bare `md5` hash we used to
generate.
- also use a generic digest parameter for the `version` directive, rather
than a specific `md5` parameter.
- Add command-line scope option to Spack
- Rework structure of main to allow configuration system to raise
errors more naturally
Co-authored-by: Todd Gamblin <tgamblin@llnl.gov>
- Fixes a bug in `llnl.util.lock`
- Locks in the current directory would fail because the parent directory
was the empty string.
- Fix this and return '.' for the parent of locks in the current
directory.
Replace regex-based target detection for Makefiles with a preliminary "make -q"
to check if a target exists. This does not work for NetBSD make; additional work
is required to detect if NetBSD make is present and to use a regex in that case.
The affected makefile target checks are only performed when the "--test" flag is
added to a "spack install" invocation.
Fixes#8036
Before this PR Package.installed was returning True if the spec prefix
existed, without checking the DB. This is wrong for external packages,
whose prefix exists before being registered into the DB. Now the property
checks for both the prefix and a DB entry.
Spack provides a number of classes based on commonly-used build systems
that users can extend when writing packages; the classes provide functionality
to perform the actions relevant to the build system (e.g. running "configure" for
an Autotools-based package). This adds documentation for classes supporting the
following build systems:
* Makefile
* Autotools
* CMake
* QMake
* SCons
* Waf
This includes build systems for managing extensions of the following packages:
* Perl
* Python
* R
* Octave
This also adds documentation on implementing packages that use a custom build
system (e.g. Perl/CMake).
Spack also provides extendable classes which aggregate functionality for related
sets of packages, e.g. those using CUDA. Documentation is added for
CudaPackage.
- The setup-env.sh script currently makes two calls to spack, but it
should only need to make one.
- Add a fast-path shell setup routine in `main.py` to allow the shell
setup to happen in a single, fast call that doesn't load more than it
needs to.
- This simplifies setup code, as it has to eval what Spack prints
- TODO: consider eventually making the whole setup script the output of a
spack command
* update help of `clean --all` to include `-p`
* remove old orphaned `.pyc` removal
* restrict removal or orphaned pyc files to `lib/spack` and `var/spack`
- Clean up error messages for when a lock can't be created, or when an
exclusive (write) lock can't be taken on a file.
- Add a number of subclasses of LockError to distinguish timeouts from
permission issues.
- Add an explicit check to prevent the user from taking a write lock on a
read-only file.
- We had a check for this for when we try to *upgrade* a lock on an RO
file, but not for an initial write lock attempt.
- Add more tests for different lock permission scenarios.
- write locks previously wrote information about the lock holder (host
and pid), and read locks woudl read this in.
- This is really only for debugging, so only enable it then
- add some tests that target debug info, and improve multiproc lock test
output
When a user specifies a URL for a specific version of a package, Spack originally
would use that URL for all newer versions of the package. This behavior has
proven to be generally more harmful than useful, so this PR removes the feature
such that a version-specific URL override affects only that version.
If the user sets "ccache: true" in spack's config.yaml, Spack will use an available
ccache executable when compiling c/c++ code. This feature is disabled by default
(i.e. "ccache: false") and the documentation is updated with how to enable
ccache support
Functional updates:
- `python` now creates a copy of the `python` binaries when it is added
to a view
- Python extensions (packages which subclass `PythonPackage`) rewrite
their shebang lines to refer to python in the view
- Python packages in the same namespace will not generate conflicts if
both have `...lib/site-packages/namespace-example/__init__.py`
- These `__init__` files will also remain when removing any package in
the namespace until the last package in the namespace is removed
Generally (Updated 2/16):
- Any package can define `add_files_to_view` to customize how it is added
to a view (and at the moment custom definitions are included for
`python` and `PythonPackage`)
- Likewise any package can define `remove_files_from_view` to customize
which files are removed (e.g. you don't always want to remove the
namespace `__init__`)
- Any package can define `view_file_conflicts` to customize what it
considers a merge conflict
- Global activations are handled like views (where the view root is the
spec prefix of the extendee)
- Benefit: filesystem-management aspects of activating extensions are
now placed in views (e.g. now one can hardlink a global activation)
- Benefit: overriding `Package.activate` is more straightforward (see
`Python.activate`)
- Complication: extension packages which have special-purpose logic
*only* when activated outside of the extendee prefix must check for
this in their `add_files_to_view` method (see `PythonPackage`)
- `LinkTree` is refactored to have separate methods for copying a
directory structure and for copying files (since it was found that
generally packages may want to alter how files are copied but still
wanted to copy directories in the same way)
TODOs (updated 2/20):
- [x] additional testing (there is some unit testing added at this point
but more would be useful)
- [x] refactor or reorganize `LinkTree` methods: currently there is a
separate set of methods for replicating just the directory structure
without the files, and a set for replicating everything
- [x] Right now external views (i.e. those not used for global
activations) call `view.add_extension`, but global activations do not
to avoid some extra work that goes into maintaining external views. I'm
not sure if addressing that needs to be done here but I'd like to
clarify it in the comments (UPDATE: for now I have added a TODO and in
my opinion this can be merged now and the refactor handled later)
- [x] Several method descriptions (e.g. for `Package.activate`) are out
of date and reference a distinction between global activations and
views, they need to be updated
- [x] Update aspell package activations
- Spack was assuming that a group with gid == current uid would always exist.
- This was breaking the travis build for macos.
- also fix issue with the DB tarball test finding coverage filesx
- pytest was not reporing the correct version from pytest.__version__.
It reported 'unknown'
- this fixes issues on some systems where system-installed pytest plugins
would try to use the version and convert it to an int
The following improvements are made to cxx standard support
(e.g. compiler.cxxNN_flag functions) in compilers:
* Add cxx98_flag property
* Add support for throwing an exception when a flag is not supported (previously
if a flag was not supported the application was terminated with tty.die)
* The name of the flag associated with e.g. c++14 standard support changes for
different compiler versions (e.g. c++1y vs c++14). This makes a few corrections
on what flag to return for which version.
* Added tests to confirm that versions report expected flags for various c++
standards (or raise an exception for versions that don't provide a given cxx
standard)
Note that if a given cxx standard is the default, the associated flag property will
return ""; cxx98 is assumed to be the default standard so this is the behavior for
the associated property in the base compiler class.
Package changes:
* Improvements to the boost spec to take advantage of the improved standard
flag facility.
* Update the clingo spec to catch the new exception rather than look for an
empty flag to indicate non-support (which is not part of the compiler flag API)
Fixes#7885#7193 added the patches_to_apply function to collect patches which are then
applied in Package.do_patch. However this only collects patches that are
associated with the Package object and does not include Spec-related patches
(which are applied by dependents, added in #5476).
Spec.patches already collects patches from the package as well as those applied
by dependents, so the Package.patches_to_apply function isn't necessary. All
uses of Package.patches_to_apply are replaced with Package.spec.patches.
This also updates Package.content_hash to require the associated spec to be
concrete: Spec.patches is only set after concretization. Before this PR, it was
possible for Package.content_hash to be valid before concretizing the associated
Spec if all patches were associated with the Package (vs. being applied by
dependents). This behavior was unreliable though so the change is unlikely to
be disruptive.
Fixes#8345
Spack environment modifications are applied before modules are loaded; this
includes settings to CC, FC, F77, and CXX, which point to the Spack compiler
wrappers. If the loaded modules set CC, this overrides the Spack compiler
wrappers. This PR adds a context manager to preserve the values of CC etc. that
are set by Spack: any effects on the CC, FC, F77, and CXX variables from modules
are undone and their original values are restored.
* pybind11: test support
Add a test functionality to pybind11.
* CMake: test also on "make check"
Some projects use non-CTest manual targets for tests.
* extend Prefix class with join() member to support dynamic directories
* add more tests for Prefix.join()
* more tests for Prefix.join()
* add docstring
* add example to docstring of Prefix class
* cleanup Prefix.join() tests
* use Prefix.join() in Packaging Guide
Fixes#8217
Trying to relocate a distribution when the new and old paths are
equal leads to failure, because the test that ensures that no
unrelocated bits are left over always fails. As an example, this
occurs if a user installs a package, generates a binary with it
using 'spack buildcache', uninstalls it, and then attempts to
reinstall into the same spack installation using the generated
binary package.
This updates the relocation check to accept the presence of the
old prefix in binaries if the package is being reinstalled into
its original location.
* allow user to constrain dependencies that are added conditionally
* remove check for not-visited deps from normalize, move it to concretize. The check now runs after the concretization loop completes (so an error is only reported if the user-mentioned spec doesnt appear anywhere in the dag)
* remove separate full_spec_deps variable; rename spec_deps to all_spec_deps to clarify that it merges user-specified dependencies with derived dependencies
* add unit test to confirm new functionality
- `spack config blame` is similar to `spack config get`, but it prints
out the config file and line number that each line of the merged
configuration came from.
- This is a debugging tool for understanding where Spack config settings
come from.
- add tests for config blame
Fixes: #8258#8090 altered import behavior so that import spack no longer
provides access to many other Spack modules. This addresses
a case which depended on the prior behavior and was not
updated as part of #8090. This particular import error only
came up when users were setting compiler flags on specs.
See also: #8194
- there were some leftover spack.* names being used after we removed
globals and moved everything in the top-level namespace to spack.pkgkit
- point those references to their new homes
- remove most `import spack` statements, except for files that need
`spack_version`
- import spack is no longer sufficient to use submodules
(e.g. spack.directives).
- these submodules must be imported directly. Update references
accordingly.
- Spack packages were originally expected to call `from spack import *`
themselves, but it has become difficult to manage imports in the
Spack core.
- the top-level namespace polluted by package symbols, and it's not
possible to avoid circular dependencies and unnecessary module loads in
the core, given all the stuff the packages need.
- This makes the top-level `spack` package essentially empty, save for a
version tuple and a version string, and `from spack import *` is now
essentially a no-op.
- The common routines and directives that packages need are now in
`spack.pkgkit`, and the import system forces packages to automatically
include this so that old packages that call `from spack import *`
will continue to work without modification.
- Since `from spack import *` is no longer required, we could consider
removing ``from spack import *`` from packages in the future and
shifting to ``from spack.pkgkit import *``, but we can wait a while to
do this.
- spack.util.lock behaves the same as llnl.util.lock, but Lock._lock and
Lock._unlock do nothing.
- can be disabled with a control variable.
- configuration options can enable/disable locking:
- `locks` option in spack configuration controls whether Spack will use filesystem locks or not.
- `-l` and `-L` command-line options can force-disable or force-enable locking.
- Spack will check for group- and world-writability before disabling
locks, and it will not allow a group- or world-writable instance to
have locks disabled.
- update documentation
- Spack core has long used llnl.util.filesystem.join_path, but
os.path.join is pretty much the same thing, and is more efficient.
- Use os.path.join in the core Spack code from now on.
- simplify the singleton pattern across the codebase
- reduce lines of code needed for crufty initialization
- reduce functions that need to mess with a global
- Singletons whose semantics changed:
- spack.store.store() -> spack.store
- spack.repo.path() -> spack.repo.path
- spack.config.config() -> spack.config.config
- spack.caches.fetch_cache() -> spack.caches.fetch_cache
- spack.caches.misc_cache() -> spack.caches.misc_cache
- `spack.cmd.all_commands` does a directory listing on
`lib/spack/spack/cmd`, regardless of whether it is needed
- make this lazy so that the directory listing won't happen unless it's
necessary.
- It turns out that jsonschema is one of the more expensive imports.
- move imports of jsonschema into functions to avoid the performance hits
for calls that don't need config.
- spack.store was previously initialized at the spack.store module level,
but this means the store has to be initialized on every spack call.
- this moves the state in spack.store to a singleton so that the store is
only initialized when needed.
- spack.repository module is now spack.repo
- `spack.repo` is now `spack.repo.path()` and loaded lazily
- Added `spack.repo.get()` and `spack.repo.all_package_names()` as
convenience functions to simplify the new lazy interface.
- updated tests and code
- no longer require `spack_version` to be a Version (it isn't used that
way anyway)
- use a simple tuple `spack_version_info` with major, minor, patch
versions
- generate `spack_version` from the tuple
- replace `spack.config.get_configuration()` with `spack.config.config()`
- replace `get_config`/`update_config` with `get`, `set`
- add a path syntax that can be used to refer to specific config options
without firt getting the entire configuration dict
- update usages of `get_config` and `update_config` to use `get` and `set`
- Current configuration code forces the config system to be initialized
at module scope, so configs are parsed on every Spack run, essentially
before anything else.
- We need more control over configuration init order, so move the config
scopes into a class and reduce global state in config.py
