The new cmsMake!
(only compilation using system packages work)
There are may ways to install task-maker on your system, you can compile it yourself (instructions below) or, if you are on Ubuntu or Archlinux, there are easier methods.
If you are using Arch Linux you may want to install task-maker from the AUR: task-maker or task-maker-git.
At releases you
can find prebuilt .deb files. To install them you can issue:
sudo dpkg -i that_file.deb
and then sudo apt --fix-broken install
.
After installing task-maker, run task-maker
in the task folder to compile
and run everything. Specifying no option all the caches are active, the next
executions will be very fast, actually doing only what's needed.
If you really want to repeat the execution of something provide the --cache
option:
task-maker --cache=nothing
Possible values of --cache
are: all
, reevaluate
(do not regenerate
inputs/outputs if not needed), nothing
(disable all the cache).
Sometimes you only want to test only some solutions, speeding up the compilation and cleaning a bit the output:
task-maker sol1.cpp sol2.py
Note that you may or may not specify the folder of the solution (sol/ or solution/). You can also specify only the prefix of the name of the solutions you want to check.
By default the task in the current directory is executed, if you want to change
the task without cd
-ing away:
task-maker --task-dir ~/tasks/poldo
All the compiled files are kept in an internal folder but if you want to
use them, for example to debug a solution, passing --copy-exe
all the
useful files are copied to the bin/
folder inside the task directory.
task-maker --copy-exe
Before a contest it's good practise to check that your checker does not
behave weirdly with malformed outputs from the contestants. task-maker
provides a simple fuzzer using radamsa
(which has to be installed where
the workers are running). It takes an input file and the corresponding
output file and will try to mutate the output file in order to crash the
checker. Note that task-maker
cannot determine if the checker behave
correctly (i.e. gives the correct score), it only checks if the checker
crashes or not, and if it prints a valid score (from 0.0 to 1.0).
The results are stored in fuzz_checker_task_name
in the current working
directory. Remember to clean it before each run!
task-maker --fuzz-checker in_file out_file
If you want to clean everything, for example after the contest, simply run:
task-maker --clean
You may want to build all the tasks of a contest, to do so just issue
contest-maker
in the contest directory. You can also specify
--contest-dir
to build all the tasks inside that folder or you can select
a specific contest.yaml
file with --contest-yaml
.
One of the best feature of task-maker is the ability to execute a task remotely. The setup is really simple, you need to start some programs: a server and a group of workers. The server will accept connection from workers and clients, a worker is the program that executes a command, a client is you!
First start a server:
task-maker --run-server --server-address=0.0.0.0
Then start a worker in each machine, specifying the server to connect to:
task-maker --run-worker --worker-address=<ip of the server>
To run the execution remotely just pass:
task-maker --server server_ip:7070
Note that the TCP port 7070 is used, the connection has to be available, stable and reliable. A local network is suggested but it should work also via the Internet.
If something went wrong and you want to kill task-maker you have also to kill
the background processes that may have been spawned.
You may find the pid of those processes looking at /tmp/task-maker-*-*.pid
If you want to compile tak-maker yourself you need the following dependencies:
g++ make cmake
You also need a compiler capable of compiling C++14.
To start the compilation simply run:
mkdir -p build
cd build
cmake ..
make
This will pull all the dependencies and compile everything. If you want to
speedup the compilation you may want to add -j X
(with X the number of cores)
to make
.
If you want to enable the optimization remember to put
-DCMAKE_BUILD_TYPE=Release
to the cmake
command.
Note that the build system uses Hunter for managing the dependencies, in some
cases you may find a new directory in your home called .hunter
, to prevent
this you can add -DHUNTER_ROOT=/some/path
to the cmake command.
Once the build finish you may want to install it, in build/python
you will
find a Python3 package. Create a virtualenv with python3 as default version
(virtualenv -p python3 /path
), install the python dependencies using
pip -r requirements.txt
and then install task-maker with ./setup.py install
.
We support build without Hunter (ie using system packages) only on Arch Linux, there are some dependencies to be installed:
core/elfutils
extra/cmake
community/gmock
community/gtest
community/capnproto
community/pybind11
community/python-ruamel-yaml
community/python-traits
community/python-pytoml
To compile the project issue:
mkdir -p build
cd build
cmake .. -DHUNTER_ENABLED=OFF
make