cms-dev/cms

Contest Management System

liangfeng-source

@liangfeng-source

I do not understand how to pass the right file names to the contestant's program and how the checker can find the right file name

liangfeng-source

@liangfeng-source

after I defined input and output file for the contestants' program, I found that the platform does not pass the input and output filepaths to contestant's program. Instead, the contestant can directly use the string inside its program to locate them.

and the names are the same as the names defined by the admin in the platform "I/O (blank for stdin/stdout)" input and output .

Nekosyndrome

@Nekosyndrome

Hi, I have question about the scoreboard. I can connect to my RWS server but I can only see empty results :(

Some more details: Almost everything seems good, judging and scoring works, I can see judge results in admin ranking page, RWS works and API requests return 200OK without any error message. the only issue is that every API returns an empty object {} withtout any result... (e.g. /contests, /teams, /scores ...)

Any hint for debugging the issue? thanks in advance.

RezwanArefin01

@RezwanArefin01

@Nekosyndrome Start the RWS first then cmsResourceService -a.

Nekosyndrome

@Nekosyndrome

@RezwanArefin01 tried to start RWS first, but still getting empty scoreboard.

Ranking — Mozilla Firefox 2020_12_01 23_13_18.png

I found my resourceService is continuing restarting proxyservice every few seconds, is this an expected behaviour?

2020-12-01 14:15:34,920 - INFO [Resource,0] Restarting (ProxyService, 0)...
2020-12-01 14:15:39,957 - INFO [Resource,0] Restarting (ProxyService, 0)...
2020-12-01 14:15:44,963 - INFO [Resource,0] Restarting (ProxyService, 0)...
2020-12-01 14:15:49,977 - INFO [Resource,0] Restarting (ProxyService, 0)...
...

liangfeng-source

@liangfeng-source

hi, everyone. why does CMS generate "Execution killed (could be triggered by violating memory limits)" for some testcases, while the same code and testdata can run without problem at the command line ?

the task is to read big files over 2GB.

liangfeng-source

@liangfeng-source

we decided to shrink the side to be 1 GB to make it easier to run.

RezwanArefin01

@RezwanArefin01

@Nekosyndrome No that isn't expected behavior. Try running cmsProxyService separately. See what is causing the error. Probably some address is wrong.

Laci Nikhazy

@lacitoo

Hi all, I have a question regarding the number of contestants. What was the greatest number of contestants participating in a competition using CMS that you know of? (I am from Hungary, we used it for the CEOI this year, and we consider using it for national contests.)

RezwanArefin01

@RezwanArefin01

ApIO generally have 600+ contestants. It was hosted in CMS in some years.

CMS recommends 100 participants per CWS (but actually you can go much higher, 200 doesn't make much problem).

Nekosyndrome

@Nekosyndrome

@Nekosyndrome No that isn't expected behavior. Try running cmsProxyService separately. See what is causing the error. Probably some address is wrong.

Thanks for your help, finally found the error message in proxy log.

  File "/usr/local/lib/python3.6/dist-packages/cms-1.5.dev0-py3.6.egg/cms/grading/scoretypes/abc.py", line 350, in retrieve_target_testcases
    "No testcase matches against the regexp '%s'" % t)
ValueError: No testcase matches against the regexp '0_.*'

The root cause is because I imported experimental tasks and scores without importing the testdata file, after removing those tasks, everything works!

Yen-Thanh Le

@yenthanh-t7_gitlab

Hi all, I have a question about scoring logic. In this problem we don't have a maximum score, we want the contestant that has the maximum score will have 100 point and other contestants' score is calculated by ratio with the maximum score. I want to know is this possible and what is the best way to do it? Thank you.

Andrey Vihrov

@andreyv

@yenthanh-t7_gitlab Not sure if this is possible (CMS doesn't recalculate scores dynamically), but you could have a similar setup where "100 points" is the jury best answer or a theoretical maximum, and the checker uses that as a reference.

wil93

@wil93:matrix.org

[m]

@lacitoo: the main limitation is the performance of the physical machine where the DB is running on, if that is powerful enough (and has enough RAM) then you can theoretically spin up lots of different machines that run the ContestWebServer (the user-facing part of CMS) and Workers (to evaluate submissions) and it should scale

actually the EvaluationService is another bottleneck (it's the component that sends work to Workers and collects their results) since it's single threaded

usually, EvaluationService is run in the same (powerful) machine that runs the DB

Andrey Vihrov

@andreyv

@melongist @sabinM1 CMS master is now updated to run on Ubuntu 20.04 (see installation docs)

Maxim Sabin

@sabinM1

Thanks for the mention, I'll check it out later today!

Laci Nikhazy

@lacitoo

@RezwanArefin01 @wil93:matrix.org Thanks for the replies!

Maxim Sabin

@sabinM1

Hey, is there any way to set the maximum user test size for a batch task?

Andrey Vihrov

@andreyv

@sabinM1 If you use Nginx, you can limit all uploads with client_max_body_size

liangfeng-opensource

@liangfeng-opensource

@andreyv Hi, can anyone tell me how to terminate the tasks in the status of evaluating and compiling? we have some tasks stuck with the platform and occupying the resources. We do not know how to terminate them, could anyone help us on this?

liangfeng-opensource

@liangfeng-opensource

Hi, all. We encountered "disabling and shutting down worker 0 because of no reponse in 0:11:19". It seems some task is forcing the shut down of workers. How can we identify these tasks and turn them off or terminate them?

Andrey Vihrov

@andreyv

@liangfeng-opensource You can see what's executing in AWS status page and in worker logs. You are probably hitting WORKER_TIMEOUT (600 seconds), it can happen if one testcase is being evaluated for too long.

Grace Hawkins

@Grackins

Hi, I was thinking is there any specific effort to support compatibility with the newer python versions?
for example, the latest version of PyCrypto uses time.clock which is deprecated in python3.8, or the gevent and psycopg with the versions in requirements.txt have some problems during installation in python3.8.

Andrey Vihrov

@andreyv

@Grackins Try CMS master, it should have these issues fixed

Armando La Placa

@a-lp

Hello everyone! Is there anybody here that tried to run CMS in a raspberry pi 4, with ubuntu server 20.04 LTS? I've been trying to execute some compilations, but it returns an error: "cms.grading.Sandbox.SandboxInterfaceException: Failed to initialize sandbox with command: isolate --cg --box-id=30 --init (error 2)". I also tested the CMS in a VM with same OS hosted in my computer, and it works fine. I cloned last version of CMS in both machines. Thank you all

Andrey Vihrov

@andreyv

@a-lp You might be in uncharted waters there :) Try running the same isolate command from the command line, under the same user, and see if it prints any errors.

Armando La Placa

@a-lp

It works with the logged user, which is not cmsuser, but it says that the box-id is already allocated and I should free it to use (with a command I don't remember now).

Since I already figured out how to make it works in another VM, I think i'll avoid the raspberry, also because it is not as performant as my self-hosted VM

Armando La Placa

@a-lp

Thank you anyway!

Grace Hawkins

@Grackins

@a-lp thanks! thousands of problems solved at once by switching to master! :D

Armando La Placa

@a-lp

Greetings again! I've been searching some discussions regarding the conversion of LaTeX statements into PDF, in order to be uploaded in the Task. Is there a way to import a Task using command cmsImportTask, making it compile the tex statement? If not, what should be the workflow to do so?

Ranald Lam

@ranaldmiao

Hey, in recent IOIs (as far as I can remember, 2017-2020), there has a been a "Notice" task type that was patched on top of CMS master. It is used to provide information to the contestants which are not task-specific (Eg: instructions related to a backup submission system/IDE).

Under the hood, it is implemented as a task type that only holds a problem statement, and cannot be submitted to.

I understand the solution isn't ideal, but would the general CMS community benefit from this patch? I can work on submitting a PR with it!

Andrey Vihrov

@andreyv

@ranaldmiao Indeed, the solution doesn't fit the CMS task model well. But I also understand that it is useful to have a "Notice" task in the IOI translation system.

Maybe the concept can be generalized in some way to make a better fit. For example, the task type could be named StatementOnly.

The main problem seems to be excluding such a task from scoring results. I'll check the code to get a better understanding.

Armando La Placa

@a-lp

Greetings everyone! Are there any plans to update the frontend? An idea would be to provide some APIs, then people would design their own frontend.

Andrey Vihrov

@andreyv

@a-lp Can you please clarify the question? Is it about Tornado?

Armando La Placa

@a-lp

I'm sorry, I wasn't very clear. So, I'm asking if you'll provide some APIs in order to permits user to build their own frontend, instead of using the one written in Tornado. I find the backend and core part glorious, but I would like to easily customize the frontend, since it's kinda intricate without any documentation.

Andrey Vihrov

@andreyv

@a-lp The frontend is a separate service, ContestWebServer, and communicates with other parts through RPC calls and the database. This could be considered the "API". In theory you can replace the service with another one.

Armando La Placa

@a-lp

I see. Well, I will try to figure it out. Thank you :D

Andrey Vihrov

@andreyv

@a-lp This probably isn't exactly what you wanted, but it is what it is ☺

Armando La Placa

@a-lp

This is an awesome project, the frontend is an extra, I'm not complaining XD

MacTin

@MacTin

Please , Where Can T set 2 workes for evaluation?

Andrey Vihrov

@andreyv

@MacTin You can add workers in cms.conf, see the example configuration and the docs

Andrey Vihrov

@andreyv

⚠️⚠️⚠️ Codecov security advisory: https://about.codecov.io/security-update/ ⚠️⚠️⚠️

This script is run by CMS testsuite if the non-default --coverage flag is passed. This is the case when running tests on Travis CI.

It doesn't look like the Travis CI environment has access to any secrets, but still worth to note.

Where communities thrive

People

Repo info

Activity