hackager: Hackage testing tool

[ bsd3, compiler, ghc, program, testing ] [ Propose Tags ]

Hackager is a program for compiling the entirety of Hackage as a way of testing a Haskell compiler.

[Skip to Readme]


Note: This package has metadata revisions in the cabal description newer than included in the tarball. To unpack the package including the revisions, use 'cabal get'.

Maintainer's Corner

Package maintainers

For package maintainers and hackage trustees


  • No Candidates
Versions [RSS], 1.0.0, 1.0.1,,,,
Change log CHANGELOG.md
Dependencies base (>=2 && <5), Cabal (<2), containers, directory, filepath, mtl (<2.3), process, regex-tdfa [details]
License BSD-3-Clause
Author The GHC Team, David Terei <code@davidterei.com>
Maintainer David Terei <code@davidterei.com>
Revised Revision 1 made by Bodigrim at 2023-06-26T20:42:12Z
Category Compiler, GHC, Testing
Home page https://github.com/dterei/Hackager
Bug tracker https://github.com/dterei/Hackager/issues
Source repo head: git clone git://github.com/dterei/Hackager.git
Uploaded by DavidTerei at 2015-11-20T22:40:02Z
Reverse Dependencies 1 direct, 0 indirect [details]
Executables hackager
Downloads 4903 total (16 in the last 30 days)
Rating (no votes yet) [estimated by Bayesian average]
Your Rating
  • λ
  • λ
  • λ
Status Docs not available [build log]
Last success reported on 2016-11-30 [all 4 reports]

Readme for hackager-

[back to package description]

Hackager Hackage version Build Status

Hackager is a tool to compile all of the Haskell Hackage package repository. This is useful for testing Haskell compilers.


Hackager consists of one tool that supports multiple commands.

''hackager'' is invoked with the following options:

$ usage: hackager [--version] [--help] <command> [<args>]

The valid hackager commands are:
    record    Try building all of hackage and record results
    report    Compare two 'record' runs and display results

See 'hackager help <command>' for more information on a specific command

''hackager record'' has the following options:

usage: hackager record -o NAME [-c CABAL] [-g GHC] [-p GHC-PKG] [-d DEP-FLAGS]
                      [-f PKG-FLAGS] [-n THREADS] [PKGS...]

    NAME         A name by which the results of this hackager run will
                 be referred, e.g. \"ghc-7.6.1\"
    CABAL        The path to the cabal program to use
    GHC          The path to the ghc program to use
    GHC-PKG      The path to the ghc-pkg program to use
    DEP-FLAGS    The flags to use when compiling dependencies of a package
                 e.g. \"--ghc-option=-XFoo\"
    PKG-FLAGS    The flags to use when compiling a package
                 e.g. \"--ghc-option=-XBar\"
    THREADS      Number of threads to use to build in parallel
    PKGS         A list of packages to build. If not specified all of
                 hackage is built

Executing a run of Hackager

Here is a run with GHC, no special options and using 4 threads (note that this generally takes a long time, i.e. a few days):

$ hackager record -o normal -n 4

This run has two parts. First, the 'stats' part, where Hackager checks which packages of the ones requested it believes it can build. Packages that can't be built are ones that we can't satisfy the dependencies for, usually due to the package itself or one of its dependencies not being compatible with the version of GHC in use. This produces files of the form stats.* in the output directory and should only take a few minutes.

The second part consists of attempting to build every package (in isolation) that Hackager reported it could attempt to build from the first part. This takes hours to days (for all of Hackage), and stores results in files of the form build.*. Log files for the build results of each package are also saved under folders (with alphabetical grouping to make browsing easier).

Comparing Results of Two Runs

After the first fun, execute a second run with the delta you wish. For example, this time using ''-XAlternativeLayoutRule'' to compile each package (but not the dependencies of the package):

$ hackager record -o altern -f "--ghc-option=-XAlternativeLayoutRule" -n 4

Once done, you can compare the results of the two runs:

$ hackager report normal altern

                     Built, Failed, Deps Failed, Not Tried
altern Built           628       0            0          0
       Failed           73     215            0          0
       Deps Failed       0       0          170          0
       Not Tried         0       0            0          0

These results mean that 73 packages became unbuildable when the alternative layout rule is used.

File Output

When looking at the files created by a single run of Hackager, the important one is stats.summary, which cotains the following fields:

  • ''Num packages'': Number of packges we are testing.
  • ''Installable'': Packages we believe we can install.
  • ''Uninstallable'': Packages we can't build (i.e., wrong GHC version)
  • ''Errored '': Packages Cabal claims can be built but we don't understand Cabal's output.
  • ''Installations'': Total number of packages builds we will perform during the run.

Hackager also produces a reverse dependency list for each package and a histogram of the reverse dependency count for pacakges, storing them in files of the form stats.*.


Hackager can cause arbitrary code to run on your machine. For example:

  • TemplateHaskell is run at compile time and can execute arbitrary code
  • Package configure scripts will be run
  • Custom Setup.hs programs will be run

Get involved!

We are happy to receive bug reports, fixes, documentation enhancements, and other improvements.

Please report bugs via the github issue tracker.

Master git repository:

  • git clone git://github.com/dterei/Hackager.git


This library is BSD-licensed.