Whenever the rpath string actually _grows_, it falls back to patchelf,
when it stays the same length or gets shorter, we update it in-place,
padded with null bytes.
This PR only deals with absolute -> absolute rpath replacement. We don't
use `_build_tarball(relative=True)` in our CI. If `relative` then it falls
back to the old replacement code.
With this PR, relocation time goes down significantly, likely because patchelf
does some odd things with mmap, causing lots of overhead. Example:
- `binutils`: 700MB installed, goes from `1.91s` to `0.57s`, or `3.4x` faster.
Relocation time: 27% -> 10% of total install time
- `llvm`: 6.8GB installed, goes from `28.56s` to `5.38`, or `5.3x` faster.
Relocation time: 44% -> 13% of total install time
The bottleneck is now decompression.
Note: I'm somewhat confused about the "relative rpath" code paths. Right
now this PR only deals with absolute -> absolute replacement. As far as
I understand, if you embrace relative rpaths when uploading to the
buildcache, the whole point is you _don't_ want to patch rpaths on
install? So it seems fine to not expand `$ORIGIN` again imho.
When a package asks for non-parallel make, we need to force `make -j1` because just doing `make` will run in parallel under jobserver (e.g. `spack env depfile`).
We now always add `-j1` when asked for a non-parallel execution (even if there is no jobserver).
And each `MakeExecutable` can now ask for jobserver support or not. For example: the default `ninja` does not support jobserver so spack applies the default `-j`, but `ninja@kitware` or `ninja-fortran` does, so spack doesn't add `-j`.
Tips: you can run `SPACK_INSTALL_FLAGS=-j1 make -f spack-env-depfile.make -j8` to avoid massive job-spawning because of build tools that don't support jobserver (ninja).
We try to avoid non-default variant values in the concretizer, but this doesn't make
sense for variants forced to take some non-default value by variant propagation.
Counting this as a penalty effectively biases the concretizer for small specs dependency
graphs -- something we try very hard to avoid elsewhere because it can lead to very
strange decisions.
Example: with the penalty, `spack spec hdf5` will choose the default `openmpi` as its
`mpi` provider, but `spack spec hdf5 ~~shared` will choose `mpich` because it has to set
fewer non-default variant values because `mpich`'s DAG is smaller. That's not a good
reason to prefer a non-default virtual provider.
To fix this, if the user explicitly requests a non-default value to be propagated, there
shouldn't be a penalty. Variant values set on the CLI already don't count as default; we
just need to extend that to propagated values.
Adds another post install hook that loops over the install prefix, looking for shared libraries type of ELF files, and sets the soname to their own absolute paths.
The idea being, whenever somebody links against those libraries, the linker copies the soname (which is the absolute path to the library) as a "needed" library, so that at runtime the dynamic loader realizes the needed library is a path which should be loaded directly without searching.
As a result:
1. rpaths are not used for the fixed/static list of needed libraries in the dynamic section (only for _actually_ dynamically loaded libraries through `dlopen`), which largely solves the issue that Spack's rpaths are a heuristic (`<prefix>/lib` and `<prefix>/lib64` might not be where libraries really are...)
2. improved startup times (no library search required)
Untouched spec pruning was added to reduce the number of specs
developers see getting rebuilt in their PR pipelines that they
don't understand. Because the state of the develop mirror lags
quite far behind the tip of the develop branch, PRs often find
they need to rebuild things untouched by their PR.
Untouched spec pruning was previously implemented by finding all
specs in the environment with names of packages touched by the PR,
traversing in both directions the DAGS of those specs, and adding
all dependencies as well as dependents to a list of concrete specs
that should not be considered for pruning.
We found that this heuristic results in too many pruned specs, and
that dependents of touched specs must have all their dependencies
added to the list of specs that should not be considered for pruning.
This issue was introduced in #29761:
```
==> Installing ncurses-6.3-22hz6q6cvo3ep2uhrs3erpp2kogxncbn
==> No binary for ncurses-6.3-22hz6q6cvo3ep2uhrs3erpp2kogxncbn found: installing from source
==> Using cached archive: /spack/var/spack/cache/_source-cache/archive/97/97fc51ac2b085d4cde31ef4d2c3122c21abc217e9090a43a30fc5ec21684e059.tar.gz
==> No patches needed for ncurses
==> ncurses: Executing phase: 'autoreconf'
==> ncurses: Executing phase: 'configure'
==> ncurses: Executing phase: 'build'
==> ncurses: Executing phase: 'install'
==> Error: AttributeError: 'str' object has no attribute 'propagate'
The 'ncurses' package cannot find an attribute while trying to build from sources. This might be due to a change in Spack's package format to support multiple build-systems for a single package. You can fix this by updating the build recipe, and you can also report the issue as a bug. More information at https://spack.readthedocs.io/en/latest/packaging_guide.html#installation-procedure
/spack/lib/spack/spack/build_environment.py:1075, in _setup_pkg_and_run:
1072 tb_string = traceback.format_exc()
1073
1074 # build up some context from the offending package so we can
>> 1075 # show that, too.
1076 package_context = get_package_context(tb)
1077
1078 logfile = None
```
It turns out this was caused by a bug that had been around much longer, in which the flags were passed by reference to the flag_handler, and the flag_handler was modifying the spec object, not just the flags given to the build system. The scope of this bug was limited by the forking model in Spack, which is how it went under the radar for so long.
PR includes regression test.
* remove deptype_query remnants
* deptypes -> deptype
These arguments haven't existed since 2017, but `traverse` now fails on unknown **kwargs, so they have finally popped up.
This updates the propagation logic used in `concretize.lp` to avoid rules with `path()`
in the body and instead base propagation around `depends_on()`.
Currently, compiler flags and variants are inconsistent: compiler flags set for a
package are inherited by its dependencies, while variants are not. We should have these
be consistent by allowing for inheritance to be enabled or disabled for both variants
and compiler flags.
- [x] Make new (spec language) operators
- [x] Apply operators to variants and compiler flags
- [x] Conflicts currently result in an unsatisfiable spec
(i.e., you can't propagate two conflicting values)
What I propose is using two of the currently used sigils to symbolized that the variant
or compiler flag will be inherited:
Example syntax:
- `package ++variant`
enabled variant that will be propagated to dependencies
- `package +variant`
enabled variant that will NOT be propagated to dependencies
- `package ~~variant`
disabled variant that will be propagated to dependencies
- `package ~variant`
disabled variant that will NOT be propagated to dependencies
- `package cflags==True`
`cflags` will be propagated to dependencies
- `package cflags=True`
`cflags` will NOT be propagated to dependencies
Syntax for string-valued variants is similar to compiler flags.
Fixes an issue on the RHEL8 UBI container where this test would fail because `gr_mem`
was empty for every entry in the `grp` DB.
You have to check *both* the `pwd` database (which has primary groups) and `grp` (which
has other gorups) to do this correctly.
- [x] update `llnl.util.filesystem.group_ids()` to do this
- [x] use it in the `sbang` test
This PR introduces breadth-first traversal, and moves depth-first traversal
logic out of Spec's member functions, into `traverse.py`.
It introduces a high-level API with three main methods:
```python
spack.traverse.traverse_edges(specs, kwargs...)
spack.traverse.traverse_nodes(specs, kwags...)
spack.traverse.traverse_tree(specs, kwargs...)
```
with the usual `root`, `order`, `cover`, `direction`, `deptype`, `depth`, `key`,
`visited` kwargs for the first two.
What's new is that `order="breadth"` is added for breadth-first traversal.
The lower level API is not exported, but is certainly useful for advanced use
cases. The lower level API includes visitor classes for direction reversal and
edge pruning, which can be used to create more advanced traversal methods,
especially useful when the `deptype` is not constant but depends on the node
or depth.
---
There's a couple nice use-cases for breadth-first traversal:
- Sometimes roots have to be handled differently (e.g. follow build edges of
roots but not of deps). BFS ensures that root nodes are always discovered at
depth 0, instead of at any depth > 1 as a dep of another root.
- When printing a tree, it would be nice to reduce indent levels so it fits in the
terminal, and ensure that e.g. `zlib` is not printed at indent level 10 as a
dependency of a build dep of a build dep -- rather if it's a direct dep of my
package, I wanna see it at depth 1. This basically requires one breadth-first
traversal to construct a tree, which can then be printed with depth-first traversal.
- In environments in general, it's sometimes inconvenient to have a double
loop: first over the roots then over each root's deps, and maintain your own
`visited` set outside. With BFS, you can simply init the queue with the
environment root specs and it Just Works. [Example here](3ec7304699/lib/spack/spack/environment/environment.py (L1815-L1816))
Currently, many tests hardcode to older versions of gcc for comparisons of
concretization among compiler versions. Those versions are too old to concretize for
`aarch64`-family targets, which leads to failing tests on `aarch64`.
This PR fixes those tests by updating the compiler versions used for testing.
Currently, many tests hardcode the expected architecture result in concretization to the
`x86_64` family of architectures.
This PR generalizes the tests that can be generalized, to cover multiple architecture
families. For those that test specific relationships among `x86_64`-family targets, it
ensures that concretization uses the `x86_64`-family targets in those cases.
Currently, many tests rely on the fact that `AutotoolsPackage` imposes no dependencies
on the inheriting package. That is not true on `aarch64`-family architectures.
This PR ensures that the fact `AutotoolsPackage` on `aarch64` pulls in a dependency on
`gnuconfig` is ignored when testing for the appropriate relationships among dependencies
Additionally, 5 tests currently prompt the user for input when `gpg` is available in the
user's path. This PR fixes that issue. And 7 tests fail currently when the user has a
yubikey available. This PR fixes the incorrect gpg argument causing those issues.
The `spack info <package>` command does not show the `Virtual Packages:` output unless the `--virtuals` command option is passed. Before this changes, the information that the command is supposed to be illustrating is not shown in the example and is confusing.
Changes to improve locating shared libraries on Windows, which in
turn enables the use of Clingo. This PR attempts to establish a
proper distinction between linking on Windows vs. Linux/Mac: on
Windows, linking is always done with .lib files (never .dll files).
This somewhat complicates the model since the Spec.lib method could
return libraries that were used for both linking and loading, but
since these are not always the same on Windows, it was decided to
treat Spec.libs as being for link-time libraries. Additional functions
are added to help dependents locate run-time libraries.
* Clingo is now the default concretizer on Windows
* Clingo is now the concretizer used for unit tests on Windows
* Fix a permissions issue that can occur while moving Git files during
fetching/staging
* Packages can now implement "win_add_library_dependent" to register
files/directories that include libraries that would need to link
to dependency dlls
* Packages can now implement "win_add_rpath" to register the locations
of dlls that dependents would want to load
* "Spec.libs" on Windows is updated to return link-time libraries
(i.e. .lib files, rather than .dll files)
* PackageBase.rpath on Windows is now updated to return the most-likely
locations where .dlls will be found (which is generally in the bin/
directory)
Currently there's a slow sequential step in binary relocation where all
strings of a binary are collected, with rpaths removed, and then
filtered for the old install root.
This is completely unnecessary, and also incorrect, since we replace
more than just the old install root in the prefix to prefix mapping. And
in fact the prefix to prefix mapping is parallel, and a single pass. So
even as an optimization, this filter makes no sense anymore.
Therefor we remove it
- single pass over the binary data matching all prefixes
- collect offsets and replacement strings
- do in-place updates with `fseek` / `fwrite`, since typically our
replacement touch O(few bytes) while the file is O(many megabytes)
- be nice: leave the file untouched if some string can't be
replaced
* Add patches for building clingo with MSVC
* Help python find clingo DLL
* If an executable is located in "C:\Program Files", Executable was
running into issues with the extra space. This quotes the exe
to ensure that it is treated as a single value.
Signed-off-by: Kiruya Momochi <65301509+KiruyaMomochi@users.noreply.github.com>
This commit extends the DSL that can be used in packages
to allow declaring that a package uses different build-systems
under different conditions.
It requires each spec to have a `build_system` single valued
variant. The variant can be used in many context to query, manipulate
or select the build system associated with a concrete spec.
The knowledge to build a package has been moved out of the
PackageBase hierarchy, into a new Builder hierarchy. Customization
of the default behavior for a given builder can be obtained by
coding a new derived builder in package.py.
The "run_after" and "run_before" decorators are now applied to
methods on the builder. They can also incorporate a "when="
argument to specify that a method is run only when certain
conditions apply.
For packages that do not define their own builder, forwarding logic
is added between the builder and package (methods not found in one
will be retrieved from the other); this PR is expected to be fully
backwards compatible with unmodified packages that use a single
build system.
Instead of looping over multiple regexes and the entire text file
contents, create a giant regex with all literal prefixes and do a single
pass over files to detect prefixes. Not only is a single pass faster,
it's also likely that the regex is compiled better, given that most
prefixes share a common ... prefix.
In the dfs code, flip edges so that `parent` means `from` and
`spec` means `to` in the direction of traversal. This makes it slightly
easier to write generic/composable code. For example when using visitors
where one visitor reverses direction, and another only cares about
accepting particular edges or not depending on whether the target node
is seen before, it would be good if the second visitor didn't have to
know whether the order was changed or not.
Use the same compression level as `gzip` (6) instead of what Python uses
(9).
The LLVM tarball takes 4m instead of 12m to create, and is <1% larger.
That's not worth the wait...
#32137 added an option to update() a BinaryCacheIndex with a
cooldown: repeated attempts within this cooldown would not
actually retry. However, the cooldown was not properly
tracked for failures (which is common when the mirror
does not store any binaries and therefore has no index.json).
This commit ensures that update(..., with_cooldown=True) will
also skip the update even if a failure has occurred within the
cooldown period.
Due to reuse concretization, we may generate DAGs with two occurrences
of the same package corresponding to distinct specs. This happens when
build dependencies are reused, since their dependencies are ignored in
concretization.
This caused a regression, for example: `spec['openssl']` would take the
'openssl' of the build dep `cmake`, instead of the direct `openssl`
dependency, simply because the edge to `cmake` was traversed first and
we do depth first traversal.
One solution that was discussed is to limit `spec[name]` to just direct
deps, or direct deps + transitive link deps, but this is too breaking.
Instead, this PR simply prioritizes transitive link and direct
build/run/test deps, and then falls back to a full DAG traversal. So,
it's just about order of iteration.
Scan the text files for relocatable prefixes *before* creating a tarball,
to reduce the amount of work to be done during install from binary
cache.
Co-authored-by: Harmen Stoppels <harmenstoppels@gmail.com>
Instead of showing
```
==> Error: Timed out waiting for a write lock.
```
show
```
==> Error: Timed out waiting for a write lock after 1.200ms and 4 attempts on file: /some/file
```
s.t. we actually get to see where acquiring a lock failed even when not
running in debug mode.
And use pretty time units everywhere, so we don't get 1.45e-9 seconds
but 1.450ns etc.
* backtraces without --debug
Currently `--debug` is too verbose and not-`--debug` gives to little
context about where exceptions are coming from.
So, instead, it'd be nice to have `spack --backtrace` and
`SPACK_BACKTRACE=1` as methods to get something inbetween: no verbose
debug messages, but always a full backtrace.
This is useful for CI, where we don't want to drown in debug messages
when installing deps, but we do want to get details where something goes
wrong if it goes wrong.
* completion
Currently `relocate_text` and `relocate_text_bin` are unsafe in the
sense that they run in parallel, and lead to races when modifying
different items pointing to the same inode.
This leads to the issue observed in #33453.
This PR:
1. Renames those functions to `unsafe_*` so people are aware
2. Adds logic to deal with hardlinks in current binary packages
3. Adds logic to deal with hardlinks when creating new binary tarballs,
so the install side doesn't have to de-dupe hardlinks.
4. Adds a test for 3
The assumption is that all our relocation logic preserves inodes. That
is, we should never copy a file, modify it, and then move it back. I
quickly verified, and its seems like this is true for (binary) text
relocation, as well as rpath patching in patchelf (even when the file
grows) and mach-o fixes.
* gitlab: Do not use root_spec['pkg_name'] anymore
For a long time it was fine to index a concrete root spec with the name
of a dependency in order to access the concrete dependency spec. Since
pipelines started using `--use-buildcache dependencies:only,package:never`
though, it has exposed a scheduling issue in how pipelines are
generated. If a concrete root spec depends on two different hashes of
`openssl` for example, indexing that root with just the package name
is ambiguous, so we should no longer depend on that approach when
scheduling jobs.
* env: make sure exactly one spec in env matches hash
When installing some/all specs from a buildcache, build edges are pruned
from those specs. This can result in a much smaller effective DAG. Until
now, `spack env depfile` would always generate a full DAG.
Ths PR adds the `spack env depfile --use-buildcache` flag that was
introduced for `spack install` before. This way, not only can we drop
build edges, but also we can automatically set the right buildcache
related flags on the specific specs that are gonna get installed.
This way we get parallel installs of binary deps without redundancy,
which is useful for Gitlab CI.
When downloading from binary cache not only replace RPATHs to dependencies, but
also text references to dependencies.
Example:
`autoconf@2.69` contains a text reference to the executable of its dependency
`perl`:
```
$ grep perl-5 /shared/spack/opt/spack/linux-amzn2-x86_64_v3/gcc-7.3.1/autoconf-2.69-q3lo/bin/autoreconf
eval 'case $# in 0) exec /shared/spack/opt/spack/linux-amzn2-x86_64_v3/gcc-7.3.1/perl-5.34.1-yphg/bin/perl -S "$0";; *) exec /shared/spack/opt/spack/linux-amzn2-x86_64_v3/gcc-7.3.1/perl-5.34.1-yphg/bin/perl -S "$0" "$@";; esac'
```
These references need to be replace or any package using `autoreconf` will fail
as it cannot find the installed `perl`.
Co-authored-by: Stephen Sachs <stesachs@amazon.com>
"spack install" will not update the binary index if given a concrete
spec, which causes it to fall back to direct fetches when a simple
index update would have helped. For S3 buckets in particular, this
significantly and needlessly slows down the install process.
This commit alters the logic so that the binary index is updated
whenever a by-hash lookup fails. The lookup is attempted again with
the updated index before falling back to direct fetches. To avoid
updating too frequently (potentially once for each spec being
installed), BinaryCacheIndex.update now includes a "cooldown"
option, and when this option is enabled it will not update more
than once in a cooldown window (set in config.yaml).
Co-authored-by: Tamara Dahlgren <35777542+tldahlgren@users.noreply.github.com>
* Add libpressio and dependencies; some of these packages are
maintained as forks of the original repositories and in those
cases the docstring mentions this.
* Add optional dependency in adios2 on libpressio
* cub package: set CUB_DIR environment variable for dependent
installations
* Clear R_HOME/R_ENVIRON before Spack installation (avoid sources
outside of Spack from affecting the installation in Spack)
* Rename dlib to dorian3d-dlib and update dependents; add new dlib
implementation. Pending an official policy on how to handle
packages with short names, reviewer unilaterally decided that
the rename was acceptable given that the new Spack dlib package
is referenced more widely (by orders of magnitude) than the
original
Co-authored-by: Samuel Li <shaomeng@users.noreply.github.com>
When installing an individual spec `spack --only=package --cache-only /xyz`
from a buildcache, Spack currently issues tons of warnings about
missing build deps (and their deps) in the database.
This PR disables these warnings, since it's fine to have a spec without
its build deps in the db (they are just "missing").
Currently `traverse_dependencies` fixes deptypes to traverse once and
for all in the recursion, but this is incorrect, since deptypes depend
on the node (e.g. if it's a dependency and cache-only, don't follow
build type edges, even if the parent is build from sources and needs
build deps.)
Support spackbot rebuilding all specs from source when asked (with "rebuild everything")
- Allow overriding --prune-dag cli opt with env var
- Use job variable to optionally prevent rebuild jobs early exit behavior
- ci rebuild: Use new install argument to insist deps are always installed from binary, but
package is only installed from source.
- gitlab: fix bug w/ untouched pruning
- ci rebuild: install from hash rather than json file
- When doing a "rebuild everything" pipeline, make sure that each install job only consumes
binary dependencies from the mirror being populated by the current pipeline. This avoids
using, e.g. binaries from develop, when rebuilding everything on a PR.
- When running a pipeline to rebuild everything, do not die because we generated a hash on
the broken specs list. Instead only warn in that case.
- bugfix: Replace broken no-args tty.die() with sys.exit(1)
Print a message of the form
```
Fetch mm:ss. Build: mm:ss. Total: mm:ss
```
when installing from buildcache.
Previously this only happened for source builds.
Currently "spack ci generate" chooses the first matching entry in
gitlab-ci:mappings to fill attributes for a generated build-job,
requiring that the entire configuration matrix is listed out
explicitly. This unfortunately causes significant problems in
environments with large configuration spaces, for example the
environment in #31598 (spack.yaml) supports 5 operating systems,
3 architectures and 130 packages with explicit size requirements,
resulting in 1300 lines of configuration YAML.
This patch adds a configuraiton option to the gitlab-ci schema called
"match_behavior"; when it is set to "merge", all matching entries
are applied in order to the final build-job, allowing a few entries
to cover an entire matrix of configurations.
The default for "match_behavior" is "first", which behaves as before
this commit (only the runner attributes of the first match are used).
In addition, match entries may now include a "remove-attributes"
configuration, which allows matches to remove tags that have been
aggregated by prior matches. This only makes sense to use with
"match_behavior:merge". You can combine "runner-attributes" with
"remove-attributes" to effectively override prior tags.
When a pipeline generation job is automatically failed because it
generated jobs for specs known to be broken on develop, print better
information about the broken specs that were encountered. Include
at a minimum the hash and the url of the job whose failure caused it
to be put on the broken specs list in the first place.
* env depfile: allow deps only install
- Refactor `spack env depfile` to use a Jinja template, making it a bit
easier to follow as a human being.
- Add a layer of indirection in the generated Makefile through an
`<prefix>/.install-deps/<hash>` target, which allows one to specify
different options when installing dependencies. For example, only
verbose/debug mode on when installing some particular spec:
```
$ spack -e my_env env depfile -o Makefile --make-target-prefix example
$ make example/.install-deps/<hash> -j16
$ make example/.install/<hash> SPACK="spack -d" SPACK_INSTALL_FLAGS=--verbose -j16
```
This could be used to speed up `spack ci rebuild`:
- Parallel install of dependencies from buildcache
- Better readability of logs, e.g. reducing verbosity when installing
dependencies, and splitting logs into deps.log and current_spec.log
* Silence please!
* spack.compiler.Compiler: introduce prefix property
We currently don't really have something that gives the GCC install
path, which is used by many LLVM-based compilers (llvm, llvm-amdgpu,
nvhpc, ...) to fix the GCC toolchain once and for all.
This `prefix` property is dynamic in the sense that it queries the
compiler itself. This is necessary because it's not easy to deduce the
install path from the `cc` property (might be a symlink, might be a
filename like `gcc` which works by having the compiler load a module
that sets the PATH variable, might be a generic compiler wrapper based
on environment variables like on cray...).
With this property introduced, we can clean up some recipes that have
the logic repeated for GCC.
* intel-oneapi-compilers: set --gcc-sysroot to %gcc prefix
Caches used by repositories don't reference the global spack.repo.path instance
anymore, but get the repository they refer to during initialization.
Spec.virtual now use the index, and computation done to compute the index
use Repository.is_virtual_safe.
Code to construct mock packages and mock repository has been factored into
a unique MockRepositoryBuilder that is used throughout the codebase.
Add debug print for pushing and popping config scopes.
Changed spack.repo.use_repositories so that it can override or not previous repos
spack.repo.use_repositories updates spack.config.config according to the modifications done
Removed a peculiar behavior from spack.config.Configuration where push would always
bubble-up a scope named command_line if it existed
Resolves#31782
With this change, if a spec is concrete after parsing (e.g. spec.yaml
or /hash-based), then it is not disambiguated (a process which requires
(a) that the spec be installed and (b) that it be part of the
currently-active environment).
This commit allows you to:
* Diff specs from an environment regardless of whether they have
been installed (more useful for projection/matrix-based envs)
* Diff specs read from .yaml files which may or may not be entirely
different installations of Spack
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
* Docs: Getting Started Dependencies
Finally document what one needs to install to use Spack on
Linux and Mac :-)
With <3 for minimal container users and my colleagues with
their fancy Macs.
* Debian Update Packages: GCC, Python
- build-essential: includes gcc, g++ (thx Cory)
- Python: add python3-venv, python3-distutils (thx Pradyun)
* Add RHEL8 Dependencies
* filter_file: introduce argument 'start_at'
* autotools: extend patching of the libtool script
* autotools: refactor _patch_usr_bin_file
* autotools: improve readability of the filtering
* autotools: keep the modification time of the configure scripts
* autotools: do not try to patch directories
* autotools: explain libtool patching for posterity
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
Co-authored-by: Harmen Stoppels <harmenstoppels@gmail.com>
Remove `module-info mode load` condition that prevents auto-unloading when autoloading is enabled. It looks like this condition was added to work around an issue in environment-modules that is no longer necessary.
Add quotes to make is-loaded happy
When concurrent misc_cache provider index rebuilds happen, try to
rebuild it only once, so we don't exceed misc_cache lock timeout.
For example, when using `spack env depfile`, with no previous
misc_cache, running `make -f depfile -j8` could run at most 8 concurrent
`spack install` locking on misc_cache to rebuild the provider index. If
one rebuild takes 30s, before this fix, the "worst" lock could wait up
to 30s * 7, easily exceeding misc_cache lock timeout. Now, the "worst"
lock would take 30s * 1 + ~1s * 6.
Currently, module changes from `setup_dependent_package` are applied only to the module of the package class, but not to any parent classes' modules between the package class module and `spack.package_base`.
In this PR, we create a custom class to accumulate module changes, and apply those changes to each class that requires it. This design allows us to code for a single module, while applying the changes to multiple modules as needed under the hood, without requiring the user to reason about package inheritance.
* find/list: display package counts last
We have over 6,600 packages now, and `spack list` still displays the number of packages
before it lists them all. This is useless for large sets of results (e.g., with no args)
as the number has scrolled way off the screen before you can see it. The same is true
for `spack find` with large installations.
This PR changes `spack find` and `spack list` so that they display the package count
last.
* add some quick testing
Co-authored-by: Danny McClanahan <1305167+cosmicexplorer@users.noreply.github.com>
Allow environment variables and spack-specific path substitution variables (e.g. `$spack`) to be
used in the paths associated with develop specs, while maintaining the ability to keep those
paths relative to the environment rather than the working directory.
Install: Add use-buildcache option to install
* Allow differentiating between top level packages and dependencies when
determining whether to install from the cache or not.
* Add unit test for --use-buildcache
* Use metavar to display use-buildcache options.
* Update spack-completion
Make it possible to install the Clingo package on Windows; this
also provides a means to use Clingo with Spack on Windows.
This includes
* A new "winbison" package: Windows has a port of bison and flex where
the two packages are grouped together. Clingo dependencies have been
updated to use winbison on Windows and bison elsewhere (this avoids
complicating the existin bison/flex packages until we can add support
for implied virtuals).
* The CMake build system was incorrectly converting CMAKE_INSTALL_PREFIX
to POSIX format.
* The re2c package has been modified to use CMake on Windows; for now
this is done by overloading the configure/build/install methods to
perform CMake-appropriate operations; the package should be refactored
once support for multiple build systems in one Package is available.
This commit fixes#27027.
The root cause of the issue is that the `SPACK_OLD_PROMPT` variable
was evaluated in string interpolation regardless of whether the
guard condition above evaluates to true or false. This commit uses
the `eval` keyword to defer evaluation until the command is executed.
Co-authored-by: Alexander Hornburg <alexande@xilinx.com>
Spack currently depends on parsing filenames of downloaded files to
determine what type of archive they are and how to decompress them.
This commit adds a preliminary check based on magic numbers to
determine archive type (but falls back on name parsing if the
extension type cannot be determined).
As part of this work, this commit also enables decompression of
.tar.xz-compressed archives on Windows.
Include exception info related to url retrieval in debug messages
which otherwise would be swallowed. This is intended to be useful
for detecting if CA configuration interferes with downloads from
HTTPS links.
This fixes a bug where two installations that differ only by package hash will not show
up in `spack find`.
The bug arose because `_cmp_node` on `Spec` didn't include the package hash in its
yielded fields. So, any two `Spec` objects that were only different by package hash
would appear to be equal and would overwrite each other when inserted into the same
`dict`. Note that we could still *install* specs with different package hashes, and they
would appear in the database, but we code that needed to put them into data structures
that use `__hash__` would have issues.
This PR makes `Spec.__hash__` and `Spec.__eq__` include the `process_hash()`, and it
makes `Spec._cmp_node` include the package hash. All of these *should* include all
information in a spec so that we don't end up in a situation where we are blind to
particular field differences.
Eventually, we should unify the `_cmp_*` methods with `to_node_dict` so there aren't two
sources of truth, but this needs some thought, since the `_cmp_*` methods exist for
speed. We should benchmark whether it's really worth having two types of hashing now
that we use `json` instead of `yaml` for spec hashing.
- [x] Add `package_hash` to `Spec._cmp_node`
- [x] Add `package_hash` to `spack.solve.asp.spec_clauses` so that the `package_hash`
will show up in `spack diff`.
- [x] Add `package_hash` to the `process_hash` (which doesn't affect abstract specs
but will make concrete specs correct)
- [x] Make `_cmp_iter` report the dag_hash so that no two specs with different
process hashes will be considered equal.
53a7b49 created a reference error which broke `.libs` (and
`find_libraries`) for many packages. This fixes the reference
error and improves the testing for `find_libraries` by actually
checking the extension types of libraries that are retrieved by
the function.
* catch json schema errors and reraise as property of SpackError
* no need to catch subclass of given error
* Builtin json library for Python 2 uses more generic type
* Correct instantiation of SpackError (requires a string rather than an exception)
* Use exception chaining (where possible)
Add a post-install step which runs (only) on Windows to modify an
install prefix, adding symlinks to all dependency libraries.
Windows does not have the same concept of RPATHs as Linux, but when
resolving symbols will check the local directory for dependency
libraries; by placing a symlink to each dependency library in the
directory with the library that needs it, the package can then
use all Spack-built dependencies.
Note:
* This collects dependency libraries based on Package.rpath, which
includes only direct link dependencies
* There is no examination of libraries to check what dependencies
they require, so all libraries of dependencies are symlinked
into any directory of the package which contains libraries
* ci: restore coverage computation
* Mark "test_foreground_background" as xfail
* Mark "test_foreground_background_output" as xfail
* Make number of processes explicit, remove verbosity on linux
* Run coverage on just 3 Python jobs for linux
* Run coverage on just 3 Python jobs for linux
* Run coverage on just 2 Python jobs for linux
* Add back verbose, since before we didn't encounter the xdist internal error
* Reduce the workers to 2
* Try to use command line
* Fix a version cmp bug in asp.py
* Fix submodule bug for git refs
* Add branch in logic for submodules
* Fix git version comparisons
main does not satisfy git.foo=main
git.foo=main does satisfy main
* Add two no-op jobs named "all-prechecks" and "all"
These are a suggestion from @tgamblin, they are stable named markers we
can use from gitlab and possibly for required checks to make CI more
resilient to refactors changing the names of specific checks.
* Enable parallel testing using xdist for unit testing in CI
* Normalize tmp paths to deal with macos
* add -u flag compatibility to spack python
As of now, it is accepted and ignored. The usage with xdist, where it
is invoked specifically by `python -u spack python` which is then passed
`-u` by xdist is the entire reason for doing this. It should never be
used without explicitly passing -u to the executing python interpreter.
* use spack python in xdist to support python 2
When running on python2, spack has many import cycles unless started
through main. To allow that, this uses `spack python` as the
interpreter, leveraging the `-u` support so xdist doesn't error out when
it unconditionally requests unbuffered binary IO.
* Use shutil.move to account for tmpdir being in a separate filesystem sometimes
This patchset refactors our GitHub actions into a single top-level ci workflow that
invokes a series of reusable actions. The main goal of this is to be able to easily
control which tests run and in what order based on the success or failure of top-level
prechecks. Our previous workflows ran in three sets:
* nix tests: style and verification first, then linux and macos tests if successful
* windows tests: style and verification first, then linux and macos tests if successful
* bootstrap tests
As a result, the bootstrap tests ran even if the style failed, and style and verification
had to run on two different platforms despite running identical checks. I'm relatively
sure that's because of the limitation on dependencies between steps in the jobs.
Reusable workflows allow us to run the style, verification and now audit checks once,
then depending on the results, and the files changed, run the appropriate nix, windows
and bootstrap tests. While it saves only a few minutes by itself, this makes it easier to
refactor checks to subset tests without having to replicate tests or other workflow
components in the future.
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
Move the copying of the buildcache to a root job that runs after all the child
pipelines have finished, so that the operation can be coordinated across all
child pipelines to remove the possibility of race conditions during potentially
simlutandous copies. This lets us ensure the .spec.json.sig and .spack files
for any spec in the root mirror always come from the same child pipeline
mirror (though which pipeline is arbitrary). It also allows us to avoid copying
of duplicates, which we now do.
If you have an environment like
```
$ cat spack.yaml
spack:
specs: [openmpi@4.1.0+cuda]
```
this PR provides a new command `spack change` that you can use to adjust environment specs from the command line:
```
$ spack change openmpi~cuda
$ cat spack.yaml
spack:
specs: [openmpi@4.1.0~cuda]
```
in other words, this allows you to tweak the details of environment specs from the command line.
Notes:
* This is only allowed for environments that do not define matrices
* This is possible but not anticipated to be needed immediately
* If this were done, it should probably only be done for "named"/not-anonymous specs (i.e. we can change `openmpi+cuda` but not spec like `+cuda` or `@4.0.1~cuda`)
fixes#31484
Before this change if anything was matching an external
condition, it was considered "external" and thus something
to be "built".
This was happening in particular to external packages
that were re-read from the DB, which then couldn't be
reused, causing the problems shown in #31484.
This PR fixes the issue by excluding specs with a
"hash" from being considered "external"
* Test that users have a way to select a virtual
This ought to be solved by extending the "require"
attribute to virtual packages, so that one can:
```yaml
mpi:
require: 'multi-provider-mpi'
```
* Prevent conflicts to be enforced on specs that can be reused.
* Rename the "external_only" fact to "buildable_false", to better reflect its origin
* Preliminary support for include URLs in spack.yaml (environment) files
This commit adds support in environments for external configuration files obtained from a URL with a preference for grabbing raw text from GitHub and gitlab for efficient downloads of the relevant files. The URL can also be a link to a directory that contains multiple configuration files.
Remote configuration files are retrieved and cached for the environment. Configuration files with the same name will not be overwritten once cached.
Extend the semantics of package requirements to
allow using them also under a virtual package
attribute in packages.yaml
These requirements are enforced whenever that
virtual spec is present in the DAG.
Allow users to express default requirements in packages.yaml.
These requirements are overridden if more specific requirements
are present for a given package.
This support requires adding the '--tests' option to 'spack ci rebuild'.
Packages whose stand-alone tests are broken (in the CI environment) can
be configured in gitlab-ci to be skipped by adding them to
broken-tests-packages.
Highlights include:
- Restructured 'spack ci' help to provide better subcommand summaries;
- Ensured only one InstallError (i.e., installer's) rather than allowing
build_environment to have its own; and
- Refactored CI and CDash reporting to keep CDash-related properties and
behavior in a separate class.
This allows stand-alone tests from `spack ci` to run when the `--tests`
option is used. With `--tests`, stand-alone tests are run **after** a
**successful** (re)build of the package. Test results are collected
and report(able) using CDash.
This PR adds the following features:
- Adds `-t` and `--tests` to `spack ci rebuild` to run stand-alone tests;
- Adds `--fail-fast` to stop stand-alone tests after the first failure;
- Ensures a *single* `InstallError` across packages
(i.e., removes second class from build environment);
- Captures skipping tests for externals and uninstalled packages
(for CDash reporting);
- Copies test logs and outputs to the CI artifacts directory to facilitate
debugging;
- Parses stand-alone test results to report outputs from each `run_test` as
separate test parts (CDash reporting);
- Logs a test completion message to allow capture of timing of the last
`run_test` part;
- Adds the runner description to the CDash site to better distinguish entries
in CDash tables;
- Adds `gitlab-ci` `broken-tests-packages` to CI configuration to skip
stand-alone testing for packages with known issues;
- Changes `spack ci --help` so description of each subcommand is a single line;
- Changes `spack ci <subcommand> --help` to provide the full description of
each command (versus no description); and
- Ensures `junit` test log file ends in an `.xml` extension (versus default where
it does not).
Tasks:
- [x] Include the equivalent of the architecture information, or at least the host target, in the CDash output
- [x] Upload stand-alone test results files as `test` artifacts
- [x] Confirm tests are run in GitLab
- [x] Ensure CDash results are uploaded as artifacts
- [x] Resolve issues with CDash build-and test results appearing on same row of the table
- [x] Add unit tests as needed
- [x] Investigate why some (dependency) packages don't have test results (e.g., related from other pipelines)
- [x] Ensure proper parsing and reporting of skipped tests (as `not run`) .. post- #28701 merge
- [x] Restore the proper CDash URLand or mirror ONCE out-of-band testing completed
* modified list.py and added functionality for --tag
* Removed long and very long, shifted rest of code above return statement
* removed results variable
* added import statement at top
* added the line accidentally deleted
* added line accidentally deleted
* changed p.name to p, added line inside if statement
* line order switched
* [@spackbot] updating style on behalf of sparkyniner
* ran update completion command
* add tests
* Update lib/spack/spack/test/cmd/list.py
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* [@spackbot] updating style on behalf of sparkyniner
* changed argument to mock_packages and moved code under filter by tag
* removed bad rebase code and added additional test
* [@spackbot] updating style on behalf of sparkyniner
* added line removed earlier
* added line removed earlier
* replaced function
* added more recommended changes
Co-authored-by: sairaj <sairaj@sairajs-MacBook-Pro.local>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Assertions without messages if/when hit create a blank error message for users.
This PR adds error messages to all assertions in asp.py even
if it seems unlikely they will ever be needed.
The argument is very likely a typo, and was meant to
be given to the fixture decorator. Since the value
being passed is the default, let's just remove it.
Ensure that build tools with module-level commands in spack use
the version built as part of their build graph if one exists.
This is now also required for mesa, scons, cmake and ctest, out
of graph versions of these tools in path will not be found unless
added as an external.
This bug appeared because a new version of rocprim needs cmake
3.16, while I have 3.14 in my path I had added an external for
cmake 3.20 to the dag, but 3.14 was still used to configure
rocprim causing it to fail. As far as I can tell, all the build
tools added in build_environment.py had this problem, despite the
fact that they should have been resolving these tools by name
with a path search and find the one in the dag that way. I'm
still investigating why the path searching and Executable logic
didn't do it, but this makes three of the build systems much more
explicit, and leaves only gmake and ninja as dependencies from
out in the system while ensuring the version in the dag is used
if there is one.
The additional sqlite version is to perturb the hash of python to
work around a relocation bug which will be fixed in a subsequent
PR.
* filesystem: use lstat in recursive mtime
When a `develop` path contains a dead symlink, the `os.stat` in the recursive `mtime` determination trips up over it.
Closes#32165.
`requirement_policy/3` is generated and may not be in Spack's inputs to Clingo.
Currently this is causing warnings like:
```
$ spack spec zlib
/global/u2/t/tgamblin/src/spack/lib/spack/spack/solver/concretize.lp:510:3-43: info: atom does not occur in any rule head:
requirement_policy(Package,X,"one_of")
/global/u2/t/tgamblin/src/spack/lib/spack/spack/solver/concretize.lp:517:3-43: info: atom does not occur in any rule head:
requirement_policy(Package,X,"one_of")
/global/u2/t/tgamblin/src/spack/lib/spack/spack/solver/concretize.lp:523:3-43: info: atom does not occur in any rule head:
requirement_policy(Package,X,"any_of")
/global/u2/t/tgamblin/src/spack/lib/spack/spack/solver/concretize.lp:534:3-43: info: atom does not occur in any rule head:
requirement_policy(Package,X,"any_of")
Input spec
--------------------------------
zlib
Concretized
--------------------------------
zlib@1.2.11%gcc@7.5.0+optimize+pic+shared arch=cray-sles15-haswell
```
- [x] Silence warning with `#defined requirement_policy/3`
Spack doesn't have an easy way to say something like "If I build
package X, then I *need* version Y":
* If you specify something on the command line, then you ensure
that the constraints are applied, but the package is always built
* Likewise if you `spack add X...`` to your environment, the
constraints are guaranteed to hold, but the environment always
builds the package
* You can add preferences to packages.yaml, but these are not
guaranteed to hold (Spack can choose other settings)
This commit adds a 'require' subsection to packages.yaml: the
specs added there are guaranteed to hold. The commit includes
documentation for the feature.
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
All PRs are failing the docs build on account of an error with
pygments. These errors coincide with a new release of pygments
(2.13.0) and restricting to < 2.13 allows the doc tests to pass,
so this commit enforces that constraint for the docs build.
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
A few attribute in packages are meant to be reserved for
Spack internal use. This audit checks packages to ensure none
of these attributes are overridden.
- [x] add additional audit check
This PR fixes the performance regression reported in #31985 and a few
other issues found while refactoring the spack mirror create command.
Modifications:
* (Primary) Do not require concretization for
`spack mirror create --all`
* Forbid using --versions-per-spec together with --all
* Fixed a few issues when reading specs from input file (specs were
not concretized, command would fail when trying to mirror
dependencies)
* Fix issue with default directory for spack mirror create not being
canonicalized
* Add more unit tests to poke spack mirror create
* Skip externals also when mirroring environments
* Changed slightly the wording for reporting (it was mentioning
"Successfully created" even in presence of errors)
* Fix issue with colify (was not called properly during error
reporting)
`LD_LIBRARY_PATH` can break system executables (e.g., when an enviornment is loaded) and isn't necessary thanks to `RPATH`s. Packages that require `LD_LIBRARY_PATH` can set this in `setup_run_environment`.
- [x] Prefix inspections no longer set `LD_LIBRARY_PATH` by default
- [x] Document changes and workarounds for people who want `LD_LIBRARY_PATH`
These changes make many packages build on nixos where nearly nothing
comes from /bin or /usr/bin (the only things in "system locations" are
/bin/sh and /usr/bin/env, all the rest is found through PATH).
Many configuration scripts hardcode /usr/bin/file instead of using the
one from PATH. This patches them to use file from PATH.
fixes#31736
Catch errors when concretizing specs and report them as
debug messages. The corresponding spec is skipped.
Co-authored-by: Greg Becker <becker33@llnl.gov>
* Extracted two functions in cmd/install.py
* Extracted a function to perform installation from the active environment
* Rename a few functions, remove args from their arguments
* Rework conditional in install_from_active_environment to reduce nesting in the function
* Extract functions to parsespecs from cli and files
* Extract functions to getuser confirmation for overwrite
* Extract functions to install specs inside and outside environments
* Rename a couple of functions
* Fix outdated comment
* Add missing imports
* Split conditional to dedent one level
* Invert check and exit early to dedent one level when requiring user confirmation
The current use of git ref's as a version requires a search algorithm to pick the right matching version based on the tags in the git history of the package.
This is less than ideal for the use case where users already know the specific version they want the git ref to be associated with. This PR makes a new version syntax [package]@[ref]=[version] to allow the users to specify the exact hash they wish to use.
On some systems the shell in login mode wipes important parts of the
environment, such as PATH. This causes the build to fail since it can't
find `spack`.
For better robustness, don't use a login shell.
In a full CI job the final spack install is run in an environment formed by scripts running in this order:
export AWS_SECRET=... # 1. Load environment from GitLab project variables
source spack/share/spack/setup-env.sh # 2. Load Spack into the environment (PATH)
spack env activate -V concrete_env # 3. Activate the concrete environment
source /etc/profile # 4. Bash login shell (from -l)
spack install ...
Whereas when a user launches their own container with (docker|podman) run -it, they end up running spack install in an environment formed in this order:
source /etc/bash.bashrc # (not 4). Bash interactive shell (default with TTY)
export AWS_SECRET=... #~1. Manually load environment from GitLab project variables
source spack/share/spack/setup-env.sh # 2. Load Spack into the environment (PATH)
spack env activate -V concrete_env # 3. Activate the concrete environment
spack install ...
The big problem being that (4) has a completely different position and content (on Leap 15 and possibly other containers).
So in context, this PR removes (4) from the CI job case, leaving us with the simpler:
export AWS_SECRET=... # 1. Load environment from GitLab project variables
source spack/share/spack/setup-env.sh # 2. Load Spack into the environment (PATH)
spack env activate -V concrete_env # 3. Activate the concrete environment
spack install ...
* database: don't sort on return from query_local
* ASP-based solver: don't build the hash-lookup dictionary twice
Building this dictionary twice and traversing all the specs
might be time-consuming for large buildcaches.
This test relied on an old version of the `flake8_package` fixture that modified
the spack repository, but it doesn't do that anymore. There are other tests for
`changed_files()` that do a better job of mocking up a git repository with
changes, so we can just delete this one.
`spack style` tests were annoyingly brittle because we could not easily be
specific about which tools to run (we had to use `--no-black`, `--no-isort`,
`--no-flake8`, and `--no-mypy`). We should be able to specify what to run OR
what to skip.
Now you can run, e.g.:
spack style --tool black,flake8
or:
spack style --skip black,isort
- [x] Remove `--no-black`, `--no-isort`, `--no-flake8`, and `--no-mypy` args.
- [x] Add `--tool TOOL` argument.
- [x] Add `--skip TOOL` argument.
- [x] Allow either `--tool black --tool flake8` or `--tool black,flake8` syntax.
- [x] remove alignment spaces from tempaltes
- [x] replace single with double quotes
- [x] Makefile template now generates parsable code
(function body is `pass` instead of just a comment)
- [x] template checks now run black to check output
Previously we'd accept any version for bootstrapping black, but we need <= 21.
- [x] modify bootstrapping code to check black version before accepting an
executable from `PATH`.
- [x] add `.git-blame-ignore-revs` to ignore black reformatting
- [x] make `spack blame` respect `.git-blame-ignore-revs`
(even if the user hasn't configured git to do so)
Some of our tests rely on single vs. double quotes, and others rely on specific
line numbers in the source. These needed fixing after the switch to Black.
Black will automatically fix a lot of the exceptions we previously allowed for
directives, so we don't need them in our custom `flake8_formatter` anymore.
- [x] remove `E501` (long line) exceptions for directives from `flake8_formatter`,
as they won't help us now.
- [x] Refine exceptions for long URLs in the `flake8_formatter`.
- [x] Adjust the mock `flake8-package` to exhibit the exceptions we still allow.
- [x] Update style tests for new `flake8-package`.
- [x] Blacken style test.
Many noqa's in the code are no longer necessary now that the column limit is 99
characters. Others can easily be eliminated, and still more can just be made more
specific if they do not have to do with line length.
The only E501's still in the code are in the tests for `spack.util.path` and the tests
for `spack style`.
This adds necessary configuration for flake8 and black to work together.
This also sets the line length to 99, per the data here:
* https://github.com/spack/spack/pull/24718#issuecomment-876933636
Given the data and the spirit of black's 88-character limit, we set the limit to 99
characters for all of Spack, because:
* 99 is one less than 100, a nice round number, and all lines will fit in a
100-character wide terminal (even when the text editor puts a \ at EOL).
* 99 is just past the knee the file size curve for packages, and it means that packages
remain readable and not significantly longer than they are now.
* It doesn't seem to hurt core -- files in core might change length by a few percent but
seem like they'll be mostly the same as before -- just a bit more roomy.
- [x] set line length to 99
- [x] remove most exceptions from `.flake8` and add the ones black cares about
- [x] add `[tool.black]` to `pyproject.toml`
- [x] make `black` run if available in `spack style --fix`
Co-Authored-By: Tom Scogland <tscogland@llnl.gov>
In #31618 the idea was to determine the file extension heuristically by dropping query params etc from a url and then consider it as a file path. That broke for URLs that only have query params like http://example.com/?patch=x as it would result in empty string as basename. This PR reverts to the old behavior of saving files as ?patch=x in that case.
Co-authored-by: Stephen Sachs <stesachs@amazon.com>
* llvm: Use variant when clauses for many of the expressed conflicts
* llvm: Remove the shared variant as it wasn't really used
* llvm: Remove unnecessary deps and make explicit the ones that are
* llvm: Cleanup patch conditions
* pocl: Update for llvm cleanup
* unit-test: update unparse package hash with the updated llvm package
* llvm: Fix ppc long double patching and add clarifying comments
`self.archive_file` is (among others) a symlink to a tarball. `extension()` on a
symlink will result in no extension. This patch fixes the behavior introduced in
https://github.com/spack/spack/pull/31618.
Co-authored-by: Stephen Sachs <stesachs@amazon.com>
