It is common to run some form of linters over the code base. It is desirable to also use our build tool for this to have the benefit of parallel (or even remote) build and sound caching. Additionally, this also allows the lint tools to see the file layout as it occurs in the actual compile action, including generated files. Remember that even for source files this layout does not have to coincide with the layout of files in the source repository.
Conveniently, our build rules have support for collecting the
relevant information needed for linting built in. If a target is
built with the configuration variable LINT
set to a true value, lint
information is provided for the transitive sources;
as third-party dependencies are typically
exported without LINT
among the flexible variables, that naturally
forms a boundary of the "own" code (to be linted, as opposed to
third-party code). So, continuing the
third-party tutorial, we can obtain
abstract nodes for our sources (main.cpp
, greet/greet.hpp
,
greet/greet.cpp
).
$ just-mr analyse -D '{"LINT": true}' --dump-nodes -
INFO: Performing repositories setup
INFO: Found 5 repositories to set up
INFO: Setup finished, exec ["just","analyse","-C","...","-D","{\"LINT\": true}","--dump-nodes","-"]
INFO: Requested target is [["@","tutorial","","helloworld"],{"LINT":true}]
INFO: Analysed target [["@","tutorial","","helloworld"],{"LINT":true}]
INFO: Export targets found: 0 cached, 1 uncached, 0 not eligible for caching
INFO: Result of target [["@","tutorial","","helloworld"],{"LINT":true}]: {
"artifacts": {
"helloworld": {"data":{"id":"edf0113a4dff26d1d2453947fe0c7ae11a6cabb125a5ddb2f15e10106e01781c","path":"work/helloworld"},"type":"ACTION"}
},
"provides": {
"debug-hdrs": {
},
"debug-srcs": {
},
"lint": [
{"id":"28df1af04041db0c150bbdef440fc3265a57e1163258fd15a4373b7279e4b91a","type":"NODE"},
{"id":"f76b7acba64fc35a5c67f81d52a4aa47e0b0c8966eaf19c1f51477a4e0b8fc79","type":"NODE"},
{"id":"bd8ee55d88fade7ebc8121ab7e230aed3888b27f9e87841482b2b08ecf47acb0","type":"NODE"}
],
"package": {
"to_bin": true
},
"run-libs": {
}
},
"runfiles": {
}
}
INFO: Target nodes of target [["@","tutorial","","helloworld"],{"LINT":true}]:
{
"28df1af04041db0c150bbdef440fc3265a57e1163258fd15a4373b7279e4b91a": {
"result": {
"artifact_stage": {
"include": {
"data": {
"id": "ac340b9e4bcdf82d972ff9286bbda4cd7219d6d3487867875418aeb2b03012b5"
},
"type": "TREE"
},
"work/main.cpp": {
"data": {
"path": "main.cpp",
"repository": "tutorial"
},
"type": "LOCAL"
}
},
"provides": {
"cmd": ["c++","-O2","-Wall","-I","work","-isystem","include","-c","work/main.cpp","-o","work/main.o"],
"src": "work/main.cpp"
},
"runfiles": {
}
},
"type": "VALUE_NODE"
},
"bd8ee55d88fade7ebc8121ab7e230aed3888b27f9e87841482b2b08ecf47acb0": {
"result": {
"artifact_stage": {
"include": {
"data": {
"id": "124bb6d1afd5839463acf1f602109c4229ea303dc5dbfc63d2d4ce21fa590d24"
},
"type": "TREE"
},
"work/greet/greet.cpp": {
"data": {
"path": "greet/greet.cpp",
"repository": "tutorial"
},
"type": "LOCAL"
},
"work/greet/greet.hpp": {
"data": {
"path": "greet/greet.hpp",
"repository": "tutorial"
},
"type": "LOCAL"
}
},
"provides": {
"cmd": ["c++","-O2","-Wall","-I","work","-isystem","include","-c","work/greet/greet.cpp","-o","work/greet/greet.o"],
"src": "work/greet/greet.cpp"
},
"runfiles": {
}
},
"type": "VALUE_NODE"
},
"f76b7acba64fc35a5c67f81d52a4aa47e0b0c8966eaf19c1f51477a4e0b8fc79": {
"result": {
"artifact_stage": {
"include": {
"data": {
"id": "124bb6d1afd5839463acf1f602109c4229ea303dc5dbfc63d2d4ce21fa590d24"
},
"type": "TREE"
},
"work/greet/greet.hpp": {
"data": {
"path": "greet/greet.hpp",
"repository": "tutorial"
},
"type": "LOCAL"
}
},
"provides": {
"cmd": ["c++","-O2","-Wall","-I","work","-isystem","include","-E","work/greet/greet.hpp"],
"src": "work/greet/greet.hpp"
},
"runfiles": {
}
},
"type": "VALUE_NODE"
}
}
We find the sources in correct staging, together with the respective compile command (or preprocessing, in case of headers) provided. The latter is important, to find the correct include files and to know the correct defines to be used.
Of course, those abstract nodes are just an implementation detail and there is a rule to define linting for the collected sources. It takes two programs (targets consisting of a single artifact),
- the
linter
for running the lint task on a single file, and - the
summarizer
for summarizing the lint results; additionally, arbitraryconfig
data can be given to have config files available, but also to use a linter built from source.
As for every rule, the details can be obtained with the describe
subcommand.
$ just-mr --main rules-cc describe --rule lint targets
INFO: Performing repositories setup
INFO: Found 2 repositories to set up
INFO: Setup finished, exec ["just","describe","-C","...","--rule","lint","targets"]
| Run a given linter on the lint information provided by the given targets.
Target fields
- "linter"
| Single artifact running the lint checks.
|
| This program is invoked with
...
Let's go through these programs we have to provide one by one. The
first one is supposed to call the actual linter; as many linters,
including clang-tidy
which we use as an example, prefer to obtain
the command information through a
compilation database
there is actually some work to do, especially as the directory entry
has to be an absolute path. We also move the configuration file
.clang-tidy
from the configuration directory (located in a directory given
to us through the environment variable CONFIG
) to the position
expected by clang-tidy
.
#!/usr/bin/env python3
import json
import os
import shutil
import subprocess
import sys
def dump_meta(src, cmd):
OUT = os.environ.get("OUT")
with open(os.path.join(OUT, "config.json"), "w") as f:
json.dump({"src": src, "cmd": cmd}, f)
def run_lint(src, cmd):
dump_meta(src, cmd)
config = os.environ.get("CONFIG")
shutil.copyfile(os.path.join(config, ".clang-tidy"),
".clang-tidy")
db = [ {"directory": os.getcwd(),
"arguments": cmd,
"file": src}]
with open("compile_commands.json", "w") as f:
json.dump(db,f)
new_cmd = [ "clang-tidy", src ]
return subprocess.run(new_cmd).returncode
if __name__ == "__main__":
sys.exit(run_lint(sys.argv[1], sys.argv[2:]))
The actual information on success or failure is provided through
the exit code and information on the problems discovered (if any)
is reported on stdout or stderr. Additionally, our launcher also
writes the meta data in a file config.json
in the directory for
additional (usually machine-readable) diagnose output; the location
of this directory is given to us by the environment variable OUT
.
We use a pretty simple .clang-tidy
for demonstration purpose.
Checks: 'clang-analyzer-*,misc-*,-misc-include-*'
WarningsAsErrors: 'clang-analyzer-*,misc-*,-misc-include-*'
Computing a summary of the individual lint results (given to the summarizer as subdirectories of the current working directory) is straight forward: the overall linting passed if all individual checks passed and for the failed tests we format stdout and stderr in some easy-to-read way; additionally, we also provide a machine-readable summary of the failures.
#!/usr/bin/env python3
import json
import os
import sys
FAILED = {}
for lint in sorted(os.listdir()):
if os.path.isdir(lint):
with open(os.path.join(lint, "result")) as f:
result = f.read().strip()
if result != "PASS":
record = {}
with open(os.path.join(lint, "out/config.json")) as f:
record["config"] = json.load(f)
with open(os.path.join(lint, "stdout")) as f:
log = f.read()
with open(os.path.join(lint, "stderr")) as f:
log += f.read()
record["log"] = log
FAILED[lint] = record
with open(os.path.join(os.environ.get("OUT"), "failures.json"), "w") as f:
json.dump(FAILED, f)
failures = list(FAILED.keys())
for f in failures:
src = FAILED[f]["config"]["src"]
log = FAILED[f]["log"]
print("%s %s" % (f, src))
print("".join([" " + line + "\n"
for line in log.splitlines()]))
if failures:
sys.exit(1)
Of course, our launcher and summarizer have to be executable
$ chmod 755 run_clang_tidy.py summary.py
Now we can define our lint target.
...
, "lint":
{ "type": ["@", "rules", "lint", "targets"]
, "targets": ["helloworld"]
, "linter": ["run_clang_tidy.py"]
, "summarizer": ["summary.py"]
, "config": [".clang-tidy"]
}
...
As most rules, the lint rules also have a "defaults"
target,
which allows to set PATH
appropriately for all lint actions.
This can be useful if the linters are installed in a non-standard
directory.
$ mkdir -p tutorial-defaults/lint
$ echo '{"defaults": {"type": "defaults", "PATH": ["'"${TOOLCHAIN_PATH}"'"]}}' > tutorial-defaults/lint/TARGETS
$ git add tutorial-defaults
$ git commit -m 'add lint defaults'
We now can build our lint report in the same way as any test report.
$ just-mr build lint -P report
INFO: Performing repositories setup
INFO: Found 5 repositories to set up
INFO: Setup finished, exec ["just","build","-C","...","lint","-P","report"]
INFO: Requested target is [["@","tutorial","","lint"],{}]
INFO: Analysed target [["@","tutorial","","lint"],{}]
INFO: Export targets found: 0 cached, 1 uncached, 0 not eligible for caching
INFO: Target tainted ["lint"].
INFO: Discovered 11 actions, 7 trees, 0 blobs
INFO: Building [["@","tutorial","","lint"],{}].
INFO: Processed 7 actions, 3 cache hits.
INFO: Artifacts built, logical paths are:
out [a90a9e3a8ac23526eb31ae46c80434cfd5810ed5:41:t]
report [e69de29bb2d1d6434b8b29ae775ad8c2e48c5391:0:f]
result [7ef22e9a431ad0272713b71fdc8794016c8ef12f:5:f]
work [52b9cfc07b53c59fb066bc95329f4ca6457e7338:111:t]
INFO: Backing up artifacts of 1 export targets
INFO: Target tainted ["lint"].
To see that some real linting is going on, let's modify one of our source files. Say, we'll make the greeting independent of the recipient.
#include "greet.hpp"
#include <fmt/format.h>
void greet(std::string const& s) {
fmt::print("Hello!\n");
}
Building succeeds without any warning.
$ just-mr build helloworld
INFO: Performing repositories setup
INFO: Found 5 repositories to set up
INFO: Setup finished, exec ["just","build","-C","...","helloworld"]
INFO: Requested target is [["@","tutorial","","helloworld"],{}]
INFO: Analysed target [["@","tutorial","","helloworld"],{}]
INFO: Export targets found: 1 cached, 0 uncached, 0 not eligible for caching
INFO: Discovered 4 actions, 2 trees, 0 blobs
INFO: Building [["@","tutorial","","helloworld"],{}].
INFO: Processed 4 actions, 1 cache hits.
INFO: Artifacts built, logical paths are:
helloworld [2cb87c743e9fd3d18543732945df3ef9ca084be6:132736:x]
However, the linter reports it.
$ just-mr build lint -P report || :
INFO: Performing repositories setup
INFO: Found 5 repositories to set up
INFO: Setup finished, exec ["just","build","-C","...","lint","-P","report"]
INFO: Requested target is [["@","tutorial","","lint"],{}]
INFO: Analysed target [["@","tutorial","","lint"],{}]
INFO: Export targets found: 1 cached, 0 uncached, 0 not eligible for caching
INFO: Target tainted ["lint"].
INFO: Discovered 8 actions, 6 trees, 0 blobs
INFO: Building [["@","tutorial","","lint"],{}].
WARN (action:b9abc2d5c9766644da1f9db5ec6586f6ced35f36670046b14f73ad532ce12ba4): lint failed for work/greet/greet.cpp (exit code 1)
INFO: Processed 4 actions, 2 cache hits.
INFO: Artifacts built, logical paths are:
out [c298959107421711f8d87a2b96e95858c065b9b9:41:t] FAILED
report [0b0ab9eb90c28ece0f14a13a6ae5c97da4a32170:531:f] FAILED
result [94e1707e853c36f514de3876408c09a0e0ca6fc4:5:f] FAILED
work [007eec6bad8b691c067dd2c54165ac2912711474:111:t] FAILED
INFO: Failed artifacts:
out [c298959107421711f8d87a2b96e95858c065b9b9:41:t] FAILED
report [0b0ab9eb90c28ece0f14a13a6ae5c97da4a32170:531:f] FAILED
result [94e1707e853c36f514de3876408c09a0e0ca6fc4:5:f] FAILED
work [007eec6bad8b691c067dd2c54165ac2912711474:111:t] FAILED
0000000002 work/greet/greet.cpp
work/greet/greet.cpp:4:31: error: parameter 's' is unused [misc-unused-parameters,-warnings-as-errors]
4 | void greet(std::string const& s) {
| ^
| /*s*/
287 warnings generated.
Suppressed 286 warnings (286 in non-user code).
Use -header-filter=.* to display errors from all non-system headers. Use -system-headers to display errors from system headers as well.
1 warning treated as error
INFO: Target tainted ["lint"].
WARN: Build result contains failed artifacts.