Currently Spack keeps track of the origin in the code of any
modification to the environment variables. This is very slow
and enabled unconditionally even in code paths where the
origin of the modification is never queried.
The only place where we inspect the origins of environment
modifications is before we start a build, If there's an override
of the type `e.set(...)` after incremental changes like
`e.append_path(..)`, which is a "suspicious" change.
This is very rare though.
If an override like this ever happens, it might mean a package is
broken. If that leads to build errors, we can just ask the user to run
`spack -d install ...` and check the warnings issued by Spack to find
the origins of the problem.
It can be frustrating to successfully run `spack test run --alias <name>` only to find you cannot get the results because you already use `<name>` in some previous stand-alone test execution. This PR prevents that from happening.
Using the Spec.constrain method doesn't work since it might
trigger a repository lookup which could break our directives
and triggers a circular import error.
To fix that we introduce a function to merge abstract anonymous
specs, based only on package names, which does not perform any
lookup in the repository.
The buildcache is now extracted in a temporary folder within the current store,
moved to its final place and relocated.
"spack clean -s" has been extended to also clean the temporary extraction directory.
Add hardlinks with absolute paths for libraries in the corge, garply and quux packages
to detect incorrect handling of hardlinks in tests.
The `find` command was missing for the examples forcing colorized output. Without this (or another suitable) command, spack produces output that is not using any color. Thus, without the `find` command one does not see any difference between forced colorized and non-colorized output.
when deployed on kubernetes, the server sends back permanent redirect responses.
This is elegantly handled by the requests library, but not urllib that we have
to use here, so I have to manually handle it by parsing the exception to
get the Location header, and then retrying the request there.
Signed-off-by: vsoch <vsoch@users.noreply.github.com>
The ASP-based solver maximizes the number of values in multi-valued
variants (if other higher order constraints are met), to avoid cases
where only a subset of the values that have been specified on the
command line or imposed by another constraint are selected.
Here we swap the priority of this optimization target with the
selection of the default providers, to avoid unexpected results
like the one in #26598
Seems like https://bugs.python.org/issue29699 is relevant. Better to
just ignore errors when removing them tmpdir. The OS will remove it
anyways.
Errors are happening randomly from tests that are using this fixture.
TL;DR: there are matching groups trying to match 1 or more occurrences of
something. We don't use the matching group. Therefore it's sufficient to test
for 1 occurrence. This reduce quadratic complexity to linear time.
---
When parsing logs of an mpich build, I'm getting a 4 minute (!!) wait
with 16 threads for regexes to run:
```
In [1]: %time p.parse("mpich.log")
Wall time: 4min 14s
```
That's really unacceptably slow...
After some digging, it seems a few regexes tend to have `O(n^2)` scaling
where `n` is the string / log line length. I don't think they *necessarily*
should scale like that, but it seems that way. The common pattern is this
```
([^:]+): error
```
which matches `: error` literally, and then one or more non-colons before that. So
for a log line like this:
```
abcdefghijklmnopqrstuvwxyz: error etc etc
```
Any of these are potential group matches when using `search` in Python:
```
abcdefghijklmnopqrstuvwxyz
bcdefghijklmnopqrstuvwxyz
cdefghijklmnopqrstuvwxyz
⋮
yz
z
```
but clearly the capture group should return the longest match.
My hypothesis is that Python has a very bad implementation of `search`
that somehow considers all of these, even though it can be implemented
in linear time by scanning for `: error` first, and then greedily expanding
the longest possible `[^:]+` match to the left. If Python indeed considers
all possible matches, then with `n` matches of length `1 .. n` you
see the `O(n^2)` slowness (i verified this by replacing + with {1,k}
and doubling `k`, it doubles the execution time indeed).
This PR fixes this by removing the `+`, so effectively changing the
O(n^2) into a O(n) worst case.
The reason we are fine with dropping `+` is that we don't use the
capture group anywhere, so, we just ensure `:: error` is not a match
but `x: error` is.
After going from O(n^2) to O(n), the 15MB mpich build log is parsed
in `1.288s`, so about 200x faster.
Just to be sure I've also updated `^CMake Error.*:` to `^CMake Error`,
so that it does not match with all the possible `:`'s in the line.
Another option is to use `.*?` there to make it quit scanning as soon as
possible, but what line that starts with `CMake Error` that does not have
a colon is really a false positive...
Installing packages with a lot of dependencies does not have an easy way
of judging the current progress (apart from running `spack spec -I pkg`
in another terminal). This change allows Spack to update the terminal's
title with status information, including its current progress as well as
information about the current and total number of packages.
- Do not store the full list of environment variables in
<prefix>/.spack/spack-build-env.txt because it may contain user secrets.
- Only store environment variable modifications upon installation.
- Variables like PATH may still contain user and system paths to make
spack-build-env.txt sourceable. Variables containing paths are
modified through prepending/appending, and if we don't apply these
to the current environment variable, we end up with statements like
`export PATH=/path/to/spack/bin` with system paths missing, meaning
no system binaries are in the path, which is a bad user experience.
- Do write the full environment to spack-build-env.txt in the staging dir,
but ensure it is readonly for the current user, to make it a bit safer
on shared systems.
Creates an environment in a temporary directory and activates it, which
is useful for a quick ephemeral environment:
```
$ spack env activate -p --temp
[spack-1a203lyg] $ spack add zlib
==> Adding zlib to environment /tmp/spack-1a203lyg
==> Updating view at /tmp/spack-1a203lyg/.spack-env/view
```
The DB should be what is trusted for certain operations.
If it is not present when read we should assume the
corresponding store is empty, rather than trying a
write operation during a read.
* Add a unit test
* Document what needs to be there in tests
When a symlink to a license file exists but is broken, the license symlink post-install hook fails
because os.path.exists() checks the existence of the target not the symlink itself.
os.path.lexists() is the proper function to use.
Environments push/pop scopes upon activation. If some lazily
evaluated value depending on the current configuration was
computed and cached before the scopes are pushed / popped
there will be an inconsistency in the current state.
This PR fixes the issue for stores, but it would be better
to move away from global state.
The `spack.architecture` module contains an `Arch` class that is very similar to `spack.spec.ArchSpec` but points to platform, operating system and target objects rather than "names". There's a TODO in the class since 2016:
abb0f6e27c/lib/spack/spack/architecture.py (L70-L75)
and this PR basically addresses that. Since there are just a few places where the `Arch` class was used, here we query the relevant platform objects where they are needed directly from `spack.platforms`. This permits to clean the code from vestigial logic.
Modifications:
- [x] Remove the `spack.architecture` module and replace its use by `spack.platforms`
- [x] Remove unneeded tests
* Use gnuconfig package for config file replacement for RISC-V.
This extends the changes in #26035 to handle RISC-V. Before this change,
many packages fail to configure on riscv64 due to config.guess being too
old to know about RISC-V. This is seen out of the box when clingo fails
to build from source due to pkgconfig failing to configure, throwing
error: "configure: error: cannot guess build type; you must specify one".
* Add riscv64 architecture
* Update vendored archspec from upstream project.
These archspec updates include changes needed to support riscv64.
* Update archspec's __init__.py to reflect the commit hash of archspec being used.
Cherry-picked from #25564 so this is standalone.
With this PR we can activate an environment in Spack itself, without computing changes to environment variables only necessary for "shell aware" env activation.
1. Activating an environment:
```python
spack.environment.activate(Environment(xyz)) -> None
```
this basically just sets `_active_environment` and modifies some config scopes.
2. Activating an environment **and** getting environment variable modifications for the shell:
```python
spack.environment.shell.activate(Environment(xyz)) -> EnvironmentModifications
```
This should make it easier/faster to do unit tests and scripting with spack, without the cli interface.
* Isolate bootstrap configuration from user configuration
* Search for build dependencies automatically if bootstrapping from sources
The bootstrapping logic will search for build dependencies
automatically if bootstrapping anything form sources. Any
external spec, if found, is written in a scope that is specific
to bootstrapping.
* Don't clean the bootstrap store with "spack clean -a"
* Copy bootstrap.yaml and config.yaml in the bootstrap area
- [x] Our wrapper error messages are sometimes hard to differentiate from other build
output, so prefix all errors from `die()` with '[spack cc] ERROR:'
- [x] The error we raise when running, say, `fc` without a Fortran compiler was not
clear enough. Clarify the message and the comment.
This converts everything in cc to POSIX sh, except for the parts currently
handled with bash arrays. Tests are still passing.
This version tries to be as straightforward as possible. Specifically, most conversions
are kept simple -- convert ifs to ifs, handle indirect expansion the way we do in
`setup-env.sh`, only mess with the logic in `cc`, and don't mess with the python code at
all.
The big refactor is for arrays. We can't rely on bash's nice arrays and be ignorant of
separators anymore. So:
1. To avoid complicated separator logic, there are three types of lists. They are:
* `$lsep`-separated lists, which end with `_list`. `lsep` is customizable, but we
picked `^G` (alarm bell) for `$lsep` because it's ASCII and it's unlikely that it
would actually appear in any arguments. If we need to get fancier (and I will lose
faith in the world if we do) then we could consider XON or XOFF.
* `:`-separated directory lists, which end with `_dirs`, `_DIRS`, `PATH`, or `PATHS`
* Whitespace-separated lists (like flags), which can have any other name.
Whitespace and colon-separated lists come with the territory with PATHs from env
vars and lists of flags. `^G` separated lists are what we use for most internal
variables, b/c it's more likely to work.
2. To avoid subshells, use a bunch of functions that do dirty `eval` stuff instead. This
adds 3 functions to deal with lists:
* `append LISTNAME ELEMENT [SEP]` will put `ELEMENT` at the end of the list called
`LISTNAME`. You can optionally say what separator you expect to use. Note that we
are taking advantage of everything being global and passing lists by name.
* `prepend LISTNAME ELEMENT [SEP]` like append, but puts `ELEMENT` at the start of
`LISTNAME`
* `extend LISTNAME1 LISTNAME2 [PREFIX]` appends everything in LISTNAME2 to
LISTNAME1, and optionally prepends `PREFIX` to every element (this is useful for
things like `-I`, `-isystem `, etc.
* `preextend LISTNAME1 LISTNAME2 [PREFIX]` prepends everything in LISTNAME2 to
LISTNAME1 in order, and optionally prepends `PREFIX` to every element.
The routines determine the separator for each argument by its name, so we don't have to
pass around separators everywhere. Amazingly, as long as you do not expand variables'
values within an `eval` environment, you can do all this and still preserve quoting.
When iterating over lists, the user of this API still has to set and unset `IFS`
properly.
We ended up having to ignore shellcheck SC2034 (unused variable), because using evals
all over the place means that shellcheck doesn't notice that our list variables are
actually used.
So far this is looking pretty good. I took the most complex unit test I could find
(which runs a sample link line) and ran the same command line 200 times in a shell
script. Times are roughly as follows:
For this invocation:
```console
$ bash -c 'time (for i in `seq 1 200`; do ~/test_cc.sh > /dev/null; done)'
```
I get the following performance numbers (the listed shells are what I put in `cc`'s
shebang):
**Original**
* Old version of `cc` with arrays and `bash v3.2.57` (macOS builtin): `4.462s` (`.022s` / call)
* Old version of `cc` with arrays and `bash v5.1.8` (Homebrew): `3.267s` (`.016s` / call)
**Using many subshells (#26408)**
* with `bash v3.2.57`: `25.302s` (`.127s` / call)
* with `bash v5.1.8`: `27.801s` (`.139s` / call)
* with `dash`: `15.302s` (`.077s` / call)
This version didn't seem to work with zsh.
**This PR (no subshells)**
* with `bash v3.2.57`: `4.973s` (`.025s` / call)
* with `bash v5.1.8`: `4.984s` (`.025s` / call)
* with `zsh`: `2.995s` (`.015s` / call)
* with `dash`: `1.890s` (`.0095s` / call)
Dash, with the new posix design, is easily the winner.
So there are several interesting things to note here:
1. Running the posix version in `bash` is slower than using `bash` arrays. That is to be
expected because it's doing a bunch of string processing where it likely did not have
to before, at least in `bash`.
2. `zsh`, at least on macOS, is significantly faster than the ancient `bash` they ship
with the system. Using `zsh` with the new version also makes the posix wrappers
faster than `develop`. So it's worth preferring `zsh` if we have it. I suppose we
should also try this with newer `bash` on Linux.
3. `bash v5.1.8` seems to be significantly faster than the old system `bash v3.2.57` for
arrays. For straight POSIX stuff, it's a little slower. It did not seem to matter
whether `--posix` was used.
4. `dash` is way faster than `bash` or `zsh`, so the real payoff just comes from being
able to use it. I am not sure if that is mostly startup time, but it's significant.
`dash` is ~2.4x faster than the original `bash` with arrays.
So, doing a lot of string stuff is slower than arrays, but converting to posix seems
worth it to be able to exploit `dash`.
- [x] Convert all but array-related portions to sh
- [x] Fix basic shellcheck issues.
- [x] Convert arrays to use a few convenience functions: `append` and `extend`
- [x] Get `cc` tests passing.
- [x] Add `cc` tests where needed passing.
- [x] Benchmarking.
Co-authored-by: Tom Scogland <scogland1@llnl.gov>
Co-authored-by: Danny McClanahan <1305167+cosmicexplorer@users.noreply.github.com>
When using modules for compiler (and/or external package), if a
package's `setup_[dependent_]build_environment` sets `PYTHONHOME`, it
can influence the python subprocess executed to gather module
information.
The error seen was:
```
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)
```
But the actual hidden error happened in the `python -c 'import
json...'` subprocess, which made it return an empty string as json:
```
ModuleNotFoundError: No module named 'encodings'
```
This fix uses `python -E` to ignore `PYTHONHOME` and
`PYTHONPATH`. Should be safe here because the python subprocess code
only use packages built-in python.
The python subprocess in `environment.py` was also patched to be safe
and consistent.
* Remove redundant preserve environment code in build environment
* Remove fix for a bug in a module
See https://github.com/spack/spack/issues/3153#issuecomment-280460041,
this shouldn't be part of core spack.
* Don't module unload cray-libsci on all platforms
Spack has logic to preserve an installation prefix when it is being
overwritten: if the new install fails, the old files are restored.
This PR adds error handling for when this backup restoration fails
(i.e. the new install fails, and then some unexpected error prevents
restoration from the backup).
