This is the command-line usage
The full command line is
AltCover [/i[nputDirectory]=VALUE] [/o[utputDirectory]=VALUE] [/y|symbolDirectory=VALUE] [/d[ependency]=VALUE] [/k[ey]=VALUE] [/sn|strongNameKey=VALUE] [/r[eport]=VALUE] [/f[ileFilter]=VALUE] [/p[athFilter]=VALUE] [/s|assemblyFilter=VALUE] [/e|assemblyExcludeFilter=VALUE] [/t[ypeFilter]=VALUE] [/m[ethodFilter]=VALUE] [/a[ttributeFilter]=VALUE] [/attributetoplevel=VALUE] [/typetoplevel=VALUE] [/methodtoplevel=VALUE] [--l[ocalSource]] [/c[allContext]=VALUE] [/reportFormat=VALUE] [--inplace] [--save] [--zipfile] [--methodpoint] [--all] [--linecover] [--branchcover] [--dropReturnCode] [--sourcelink] [--eager] [--v[isibleBranches]] [/showstatic[=VALUE]] [--showGenerated] [--trivia] [--portable] [-q] [--verbose] [--?|help|h] [-- ] [...]
or
AltCover Runner [/r[ecorderDirectory]=VALUE] [/w[orkingDirectory]=VALUE] [/x|executable=VALUE] [--collect] [/l[covReport]=VALUE] [/t[hreshold]=VALUE] [/c[obertura]=VALUE] [/p[ackage]=VALUE] [/o[utputFile]=VALUE] [--dropReturnCode] [/summary|teamcity[=VALUE]] [-q] [--verbose] [--?|help|h] [-- ] [...]
or
AltCover ImportModule
or
AltCover Version
or, for the global tool only
AltCover TargetsPath
In detail
-i, --inputDirectory=VALUE Optional, multiple: A folder containing assemblies
to instrument (default: current directory)
-o, --outputDirectory=VALUE
Optional, multiple: A folder to receive the
instrumented assemblies and their companions (
default: sub-folder '__Instrumented' of the
current directory; or '__Saved' if '--inplace'
is set).
See also '--inplace'
-y, --symbolDirectory=VALUE
Optional, multiple: Additional directory to search
for matching symbols for the assemblies in the
input directory
-d, --dependency=VALUE Optional, multiple: assembly path to resolve
missing reference.
-k, --key=VALUE Optional, multiple: any other strong-name key to
use
--sn, --strongNameKey=VALUE
Optional: The default strong naming key to apply
to instrumented assemblies (default: None)
-r, --report=VALUE Optional: The output report template file (default:
'coverage.xml' or 'coverage.json' in the
current directory)
-f, --fileFilter=VALUE Optional, multiple: source file name to exclude
from instrumentation
-p, --pathFilter=VALUE Optional, multiple: source file path to exclude
from instrumentation
-s, --assemblyFilter=VALUE Optional, multiple: assembly name to exclude from
instrumentation
-e, --assemblyExcludeFilter=VALUE
Optional, multiple: assembly which links other
instrumented assemblies but for which internal
details may be excluded
-t, --typeFilter=VALUE Optional, multiple: type name to exclude from
instrumentation
-m, --methodFilter=VALUE Optional, multiple: method name to exclude from
instrumentation
-a, --attributeFilter=VALUE
Optional, multiple: attribute name to exclude from
instrumentation
--attributetoplevel=VALUE
Optional, multiple: Types marked with an attribute
of a type that matches the regex are considered
top-level, and are not excluded from coverage on
the basis of any type which textually encloses
them.
--typetoplevel=VALUE Optional, multiple: Types with a name that matches
the regex are considered top-level, and are not
excluded from coverage on the basis of any type
which textually encloses them.
--methodtoplevel=VALUE Optional, multiple: Methods with a name that
matches the regex are considered top-level, and
are not excluded from coverage on the basis of
any method which textually encloses them.
-l, --localSource Optional: Don't instrument code for which the
source file is not present.
-c, --callContext=VALUE Optional, multiple: Tracking either times of
visits in ticks or designated method calls
leading to the visits.
A single digit 0-7 gives the number of
decimal places of seconds to report; everything
else is at the mercy of the system clock
information available through DateTime.UtcNow
A string in brackets "[]" is interpreted as
an attribute type name (the trailing "Attribute"
is optional), so [Test] or [TestAttribute] will
match; if the name contains one or more ".",
then it will be matched against the full name of
the attribute type.
Other strings are interpreted as method
names (fully qualified if the string contains
any "." characters).
--reportFormat=VALUE Optional: Generate the report in the specified
format (Json, NCover or the default OpenCover)
--inplace Optional: Instrument the inputDirectory, rather
than the outputDirectory (e.g. for dotnet test)
--save Optional: Write raw coverage data to file for
later processing
--zipfile Optional: Emit the coverage report inside a zip
archive.
--methodpoint Optional: record only whether a method has been
visited or not. Overrides the --linecover and --
branchcover options.
--all Optional: record every hit at any location, and
for each context if --callContext is operating.
--linecover Optional: Do not record branch coverage. Ignored
for the --reportFormat=NCover option.
Incompatible with --branchcover.
--branchcover Optional: Do not record line coverage. Ignored
for the --reportFormat=NCover option.
Incompatible with --linecover.
--dropReturnCode Optional: Do not report any non-zero return code
from a launched process.
--sourcelink Optional: Display sourcelink URLs rather than file
paths if present.
--eager Optional, writes runner-mode coverage data
immediately.
-v, --visibleBranches Optional: Hide complex internal IL branching
implementation details in switch/match
constructs, and just show what the source level
logic implies.
--showstatic[=VALUE] Optional: Instrument and show code that is by
default skipped as trivial. --showstatic:- is
equivalent to omitting the parameter; --
showstatic or --showstatic:+ sets the unvisited
count to a negative value interpreted by the
visualizer (but treated as zero by
ReportGenerator) ; --showstatic:++ sets the
unvisited count to zero.
--showGenerated Optional: Mark generated code with a visit count
of -2 (Automatic) for the Visualizer if unvisited
--trivia Optional: Omit trivial sequence points
--portable Optional: portable operations - coverage report,
control file and recorded data co-located with
the recorder assembly
-q Optional, multiple: Quiet mode -- once to switch
off informational messages, twice to switch off
warnings as well, three (or more) times to
switch off all messages.
--verbose Optional, multiple: Verbose mode -- once to switch
on detailed messages (or undo one level of -q if
any have been set)
-?, --help, -h Prints out the options.
-- ... Anything on the command line after a free-standing "--" is considered a separate command line to be executed after the instrumentation has been done.
or
Runner
-r, --recorderDirectory=VALUE
The folder containing the instrumented code to
monitor (including the AltCover.Recorder.g.dll
generated by previous a use of the .net core
AltCover).
-w, --workingDirectory=VALUE
Optional: The working directory for the
application launch
-x, --executable=VALUE The executable to run e.g. dotnet
--collect Optional: Process previously saved raw coverage
data, rather than launching a process.
-l, --lcovReport=VALUE Optional: File for lcov format version of the
collected data
-t, --threshold=VALUE Optional: one or more of minimum acceptable
statement (S), branch (B) or method (M/AM)
coverage percentage (integer, 1 to 100) or
maximum acceptable CRAP score (C/AC followed by
integer, 1 to 255) e.g. M80C40B50. If the value
starts with a number, a leading S is assumed. If
any threshold is specified more than once, the
last instance is assumed -- so 25S50 counts as
S50.
Zero/absent values are ignored. If a coverage
result is below threshold, or the CRAP score is
above threshold, the return code of the process
is the largest abs(threshold - actual) rounded
up to the nearest integer. AM, AC include
methods with no source; the default is to ignore
such methods.
-c, --cobertura=VALUE Optional: File for Cobertura format version of the
collected data
-p, --package=VALUE Optional, multiple: package root path for
Cobertura reports
-o, --outputFile=VALUE Optional: write the recorded coverage to this file
rather than overwriting the original report file.
--dropReturnCode Optional: Do not report any non-zero return code
from a launched process.
--summary, --teamcity[=VALUE]
Optional: Defines the format of any summary; an
optional string value consisting of the tokens :
'N' = No summary at all, regardless of other
values
'B' = TeamCity format, with branches as Block
items
'R' = TeamCity format, with branches as bRanch
items
'O' = Classic OpenCover format
'C' = Change Risk Anti-Pattern score
'+' is the same as "OC" (special case - "+" by
itself is the same as 'BOC')
If the option is present, but has no value, this
is the same as 'B'
If the option is not present, then the default
is 'OC'.
-q Optional, multiple: Quiet mode -- once to switch
off informational messages, twice to switch off
warnings as well, three (or more) times to
switch off all messages.
--verbose Optional, multiple: Verbose mode -- once to switch
on detailed messages (or undo one level of -q if
any have been set)
-?, --help, -h Prints out the options.
-- ... Anything on the command line after a free-standing "--" is considered arguments for the executable to run.
or
ImportModule Prints out the PowerShell script to import the associated PowerShell module
or
Version Prints out the AltCover build version
or, for the global tool only
TargetsPath Prints out the path to the 'altcover.global.targets' file
(as the tool cannot be 'dotnet add'ed to the project).
The 'altcover.global.props' file is present in the same directory
If the --localSource argument is given, the process checks the first source file path that it finds in the debug symbols for each assembly, and if that file is not found on the current machine, the assembly is treated as if the symbols file did not exist.
The --dependency argument will expand environment variables in the paths from release 4.0.653; %USERPROFILE%, %ProgramFiles% and %NUGET_PACKAGES% are likely to be the most useful here, e.g. %ProgramFiles%/dotnet/shared/Microsoft.AspNetCore.App/2.1.5/Microsoft.AspNetCore.Cryptography.KeyDerivation.dll or similar to pick up ASP.Net Core assemblies needed for type references
The --callContext argument has no effect with NCover report format. Tracked methods are recorded at instrumentation time, but call context information is only recorded in runner mode, including runner --collect, which allows for heavier processing after the process terminates than the ProcessExit handler.
In runner mode, exactly one of a command to be executed (-x) or the --collect option is required. If --collect is supplied then anything after a free-standing -- is ignored.
In the runner -x case, before the executable is launched AltCover creates a file in the same folder as the previously nominated XML report, and while that file is present, the raw coverage data are written there, rather than being held in memory. After the executable terminates, the data are read and processed into the XML report. This file is the same one written by altcover --save
Before v6.0 the strong-name key arguments (-k, --sn) were not available in the .net core build. At v6.0, AltCover includes a work-around for what was a netstandard1.3 limitation in Mono.Cecil (suitable APIs weren’t there), in anticipation of this being included formally in a future Mono.Cecil 0.11 release with netstandard2.0 support.
Filter values are semi-colon (;) separated regular expressions, applied by type in the order as they are defined in the command line; any item whose name matches the expression will be excluded from the coverage reporting. In the simplest case, with no special regex items, this means that a name containing the filter item as a sub-string will be excluded. In v6.0.700 or later, should the need ever arise to have a semi-colon in a regex, then escape it in by doubling (;;); if a triplet ;;; or longer is present, doubling gets grouped from the left.
Except where being driven by AltCover in “runner” mode, coverage statistics are written to the file nominated by the r|report= parameter as instrumented assemblies are unloaded from an executing AppDomain, even if this is days or weeks later. In practice the instrumented assemblies should be deleted after the relevant testing has been run, and the report file will thus be freed up.
valid arguments for --teamcity are B, R, +B, +R or nothing at all (same as B). The letter indicates which symbol to use in the TeamCity format for branch coverage (B is for Block, which by experiment did show in the build report, and R is for bRanch which is documented, but did not show when I tried it), the optional + indicates that the OpenCover summary should also be emitted.
ImportModule will print a string on Mono/non-Windows, but Windows PowerShell won’t be there to make use of it.
-e vs -s : what gives?In the case where you have Unit Tests => Code Under Test => Libraries, and the coverage of the unit tests is not of interest, then exclude those assemblies with -e, so their dependencies get rewritten, but their contents are not instrumented or added to the coverage file. The stable libraries consumed by the code under test (and stable libraries for the unit test framework) should be marked as -s to be left untouched.
-p vs -f : what gives?The distinction is that a single -p can exclude a folder $(SolutionRoot)Generated (though it has to be specified with manual expansion of the MSBuild variable) and everything in it, while a single -f can be used to exclude all .g.cs files wherever they are.
Runner mode : what gives?There are three stages to the coverage operation – instrumenting the code, exercising it, then presenting the results. Classic altcover does the first, and as part of the process, injects the result presentation code into the system under test, to be run when the test process exits (as an exit handler).
This essentially doesn’t work with dotnet test in .net core (the exit handler doesn’t have enough time to work), so the runner mode was introduced to wrap the test processing, catching coverage data as it was generated, through a named pipe connection, and then doing the presentation work after the tests completed (release 1.5). The named pipe implementation proved unreliable as a cross-platform mechanism, so a file-based mechanism was adopted instead (release 1.6) that does work reliably in all places.
With the data being buffered to disk, rather than being stored entirely in memory in one or other process, the --save and runner --collect options were introduced (release 3.0) to do, respectively, the signalling to the system under test that it should dump live to an intermediate file, and the translation of the intermediate information into the final report, with the testing to be done at some point between, rather than having to run the tests through altcover runner.
In release 5.3, the writing the collected data has been offloaded to an in-process data handler, which removes this constraint; the data collector is automatically hooked into the MSBuild process.