faysalahmed / hackzor Goto Github PK

Manual Testing is HARD. Find someway to automate the process

There is currently no facility in the interface to control the start of the
contest. There is also a need for an interface that will set the
avaliability of links. For example the "submit" link should be visible only
during contest hours.

Support the following languages:
1. Common Lisp
2. Scheme
3. C#
4. Obj-C
5. Pascal
6. Fortran

Other GNU supported languages

Find a preferably free web host to host a beta instance on

Ease automated testing by loading fixtures into the data base.

First use only happy cases, and then move onto edge cases.

Implement logging of all events like evaluation of each submit, its
details, etc. This will remove the requirement of printing to stdout. A
cleaner and more professional interface :-)

POSSIBLE SOLUTION: 
use the python logging module. This will generate an excellently formatted
and standard output.

Notify the user of messages such as warnings, accepts, WA's ,etc.

This feature is already present in Mooshak.

Look into the possible use of an AJAX interface for this

Every time the judge starts it downloads the test cases and evaluators. Can
we do something like just download the problem names and the file names of
the corresponding test cases and evaluators initially, check whether there
is already a local copy, and if so, use the local copy instead of re
downloading the entire thing?

In the Kurukshetra online contest, the problems that were queued were
randomly chosen to be evaluated. Some problems were also never evaluated.

When forking child process to exec the executable of the submited code.
Limit the no of open files to 3(stdin, stdout, stderr)

The Current Form-Manipulator system in Django is being replaced. 
Port to newforms as soon as the next version of Django is released

Current ranking sorts only based on scores. This is obviously inefficient
as there is no way to resolve ties in this method. Implement a better
ranking algorithm changing the models if necessary

Current System assumes only single user participation. Move to team-based
authentication

Submitting wrong code results in submission being accepted.

GPG is a burden which we never got working properly anyway. So we'll remove
it and find a way to encrypt communication later

I was wondering if there is way we can package Hackzor along with its
dependencies into a custom package that can directly be installed on a
linux machine? Because right now, deployment of Hackzor is very difficult
and would stop anyone from directly trying it and contributing to development. 

Currently working dir, where the evaluator store the submissions and judge
code is a part of the code directories. Need to move it out to a place that
is preferably under /tmp or /shm

Currently, when a http error occurs, the judge crashes and I think the
solution is lost. Must confirm. In either case, the judge has to restart,
which apparently takes time if the test cases are large.

The current version of the trunk is broken. Fix it ASAP

Can we check the file size on the main server itself, and not on the judge?
It seems like a waste of resources to download a huge file to the judge and
_then_ discard it because its too big.

Compilation time in evaluator should be limited to prevent certain DoS attacks.

Making the UserProfile pointing to solved Attempts make it easier to get
details on the user's submits and will also make ranking easier

Some processes are not being killed properly, while others are. This
happens if a solution overshoots the time limit