Comments
On dinhhuubaonam123 → Regarding cheating, 3 weeks ago
+41

Thanks for the report. I did my own investigation and confirm that BabyOrange5925, c12ste_vietrongtht26 and QuangThang2012 are cheaters. All of these accounts and their linked accounts have been blocked.

Your account has been restored.

In Round 1129 we run checks during the contest for unusually fast solves of hard problems. Your profile looked like a typical case: one rated contest, rating 666, A and then F within 28 minutes while skipping B–E. Most accounts with this pattern in that round were indeed cheating, so you were caught by the same check. We need to act quickly in such cases, otherwise cheaters would flood the contests.

After a detailed review of your submission history, your code and your AtCoder results, we found no evidence of cheating, so the ban has been lifted. Sorry for the inconvenience.

I checked this personally. Cheating is confirmed and the block stays.

Hi! There is no filter based on whether you might be able to solve the problems. Contest announcements are filtered only by the division's rating limit. Div. 3 announcements go to users rated below 1600 and to unrated users, and Div. 2 announcements go to users rated below 2100. Rounds open to everyone, such as Div. 1 + Div. 2 and Educational rounds, go to all subscribed users. Your account qualifies for all of them, and your "Contest email notification" setting is on (it's on by default).

We checked our mail server (Postfix) logs for your address. There are eight deliveries: September 21, 24 and 25, and October 6, 9, 11, 13 and 15. That is every contest announcement sent since September 20, including Codeforces Round 1125 (Div. 3) and WorldQuant Round 1. Every one of them is logged as status=sent (250 2.0.0 OK) from gmail-smtp-in.l.google.com, so Google's mail server accepted them all. Nothing for you is waiting in our queue now. Logs older than about September 20 have already been rotated, so we can't check the August announcements. Our mailing system keeps per-recipient records only for the two latest campaigns, so for older ones the server logs are the only source.

Gmail accepting a message doesn't guarantee it reached your Inbox. Most likely the announcements are in your Promotions or Spam folder. If you find one there, mark it as "Not spam" or move it to Primary so future ones arrive normally. Some other providers, such as Microsoft and Apple, are currently rejecting part of our mail because of a blocklist issue, but that doesn't affect Gmail: it accepted all of your messages.

News straight from the source: https://t.me/durov/548

Thank you. You can rest assured that the organizers will carefully review your information.

Update: CF correctly detected a cheater. The only false positive here was the one your comment invented.

Thanks for the reminder. I've blocked both you and your second account t_jakhar_15.

On jeroenodb → Polygon Server Issues, 3 months ago
+82

Thank you for reaching out. I can confirm the issue with invocations, and I am already working on fixing it. Sorry for the inconvenience!

On Keshi → I have finally made it!, 5 months ago
+132

This email is a scam/phishing attempt. I do not send such emails, and this message was not sent by me or the Codeforces team.

It was quickly fixed

Hi. This change was made by me, so let me explain what happened.

Originally, contest.standings worked for regular public contests with quite a lot of functionality. Unfortunately, this method is much heavier than it may look from outside. It is not always just “read standings from a table and return JSON”. Depending on the request, it may involve current virtual participants of the requester, practice submissions, filtering/paging, and then return a pretty large response.

Recently this endpoint started to be used much more heavily by automated tools/scripts. I also suspect that some of this traffic came from AI/research-related usage. As a result, the whole site could become noticeably slower, because Codeforces had to calculate and return too much data too often.

A few weeks ago I reacted by heavily restricting this method. In particular, I removed access to regular public contest standings through this API for ordinary users. This did break tools like Carrot, CF-Predictor, and similar extensions. I understand that it was unpleasant for users who relied on them.

Today I made the restriction softer and changed the behavior to a compromise.

For gym and mashup contests, contest.standings is available only through an authenticated API request from a user who can view the contest. Optional parameters may still be used there.

For regular contests, ordinary and anonymous users can request standings only for public contests, and only by an anonymous GET request with exactly one query parameter: contestId.

So for public regular contests, please use:

https://codeforces.me/api/contest.standings?contestId=<id>

The response for such requests is cached for a short time, usually around 1–3 minutes. This should be much safer for the site and, I hope, should be enough for tools like Carrot, CF-Predictor, and similar extensions, as long as they only need public standings by contest id.

If some tool currently sends an authenticated/signed request or passes extra parameters for regular contests, it will probably need to be adjusted to use the simple anonymous URL above.

I don't currently have a technical way to keep the old fully-functional API behavior without risking performance degradation of the whole site. So this is not a full rollback to the old behavior, but an attempt to return the most useful part in a safe form.

Sorry for the breakage. I'll watch how this compromise works in practice.

Banned

You cheated in at least four contests. Under our policy, we do not grant coach mode to such accounts.

I have unbanned hananabuzenib and Nassar_Al_Barahmeh. Unfortunately, it is very difficult to keep bots under control, and their accounts were mistakenly flagged as bots due to unusual behavior patterns.

You can add it to the Catalog!

On ibrm → Bots?, 13 months ago
+4
On ClimbingToGrandmaster → ?, 13 months ago
+1
On sleepyAdarsh → WTF is happening on CF , 13 months ago
+46

fixed now

One of the servers (called Floyd) just rebooted. It looks like it has some hardware issues :-( This is the first time it has happened.

It makes absolutely no sense to draw any conclusions about cheaters before we process the results and remove them. During the round, we only catch the most obvious and blatant cases (and their number is obviously limited). Let's wait until the cheaters are processed, and only then start making any conclusions.

On BledDest → Kotlin Heroes 13 Announcement, 14 months ago
+31

It's not OK. We've updated the announcement to include:

The use of AI-based tools during the main contest is strictly limited.** To find out the guidelines for proper AI use, what is allowed and what is prohibited, please read the post "Rule Restricting the use of AI". Improper use of AI tools during the main contest may result in disqualification. Using automatic translation tools (including but not limited to neural networks and AI) to convert code written in other programming languages to Kotlin goes against the spirit of the competition, so it is also strictly prohibited.

See you tomorrow!

It is fixed now. Please, check again.

+94

The issue has been fixed, and animated GIFs are now supported as well.

On cadmiumky → Inserting GIFs in statements, 19 months ago
+104

Animated GIFs are now supported.

I can't reproduce it on Brave. It works exactly as on Chrome for me.

Hi. Please, read the response: https://codeforces.me/blog/entry/141490.

+116

Negative ratings will disappear after we process all cheaters and roll back the ratings. Please just wait a little.

Sorry, I lied a little. I didn’t go to eat cake. My brothers came from far away, and we went to the gym. The three of us in the gym for the first time!

+15

Here are the statistics on how you used magic this year. Shortly before the magic rollback, I gathered the following numbers.

By ranks:

Rank Count
legendary grandmaster 18582
newbie 11836
candidate master 7899
expert 6577
specialist 5723
master 5705
pupil 5205
grandmaster 4245
international grandmaster 3379
international master 3328

By colors:

Color Count
legendary 18582
gray 11836
orange 9033
violet 7899
red 7624
blue 6577
cyan 5723
green 5205

Thanks. We've implemented the "exclude problems created before" option.

On MikeMirzayanov → MaraTON Challenge 1, 23 months ago
+13

Here is the link to the table with the current progress on rewards for daily leaders: https://codeforces.me/r/maraton-1-leadership-awards

+48

I'll take a look on it.

Done. Thanks for the reminder.

They all used secondary accounts to participate in today's round. Such behavior is unacceptable and violates the rules. I think this could be a good lesson. I spent the entire two hours of the round investigating and analyzing the situation because of these "jokers".

The testing protocol contains Testset 'tests' is empty for .... Ask problem writers to fix it.

Hi. I've implemented some fix. Please, try again. But it seems the reason is just unusual archive (it contains extra files or something like it).

Add "codeforces" as a contest developer to the contest in Polygon.

UPD: It seems we don't have a large number of issues with this. But thanks to everyone who reported: I've applied one more reconfiguration and hope to see even fewer (or almost none) reports now.

Thank you for the initiative in developing the community!

On yummy → OpenAI o1 IOI submissions, 2 years ago
+82

Thanks for the posting such details. You guys do so interesting things!

Thanks!

Unfortunately, I like Hat (Alias) only in Russian.

Sounds great! I'll be happy if he joins!

I look forward to seeing him on Wednesday!

Challenge accepted!

Planning to hit the gym at the Astana Hilton this evening. Want to join? The first workout is free!

Maybe it is a good idea for my Instagram :-)

Please, try again.

We are also unable to determine whether you are solving the problems on your own or generating all the ideas with a large team. Indeed, in the context of online competitions, we cannot enforce all the rules strictly, but I believe we have always done what is within our power and moved forward by continuing to hold contests. I hope that the Codeforces community remains the same as before – with a significant number of participants who follow the rules and do not constantly look for loopholes to break them.

Thanks, added to the list

Hi! We're making progress on this issue. Don't go too far, and you'll see changes soon.

Right now this rule still works.

Please, try again. Thanks for testing!

Actually, I don't see any contest proposals from you. Probably, you didn't create a contest proposal (just separate problem proposals). I will consider the possibility of strengthening the warnings, but it seems that you are the first to encounter such a misunderstanding. After all, it has been repeatedly mentioned specifically about contest proposals.

There are some quotes from the sidebar:

Codeforces coordinators do not usually review single problems. They review complete contests.

To propose a contest, you should do the following: create a contest, add problems to it (at least 5), then publish the contest for Codeforces coordinators review.

Also in this post, there is:

Those who meet the requirements will find the special link “propose a contest/problems” in the personal menu in the right sidebar. Click on it and you will have the opportunity to offer a contest, add problems to it. Read all instructions carefully. Carefully prepare and submit your proposal for a contest and for problems. I think that user interface is simple and intuitive.

Fixed. Please, try again.

I can't find such Ray ID in Cloudflare logs. Are you sure in it?

Hello. I've reconfigured some settings on the Cloudflare side, but in such cases, please report to me as described here: https://codeforces.me/blog/entry/133113

Thanks. mingw-w64-clang-x86_64-toolchain looks good. Another issue: %Lf doesn't work with long double for scanf/printf (libc++ specifics?). But maybe we can live without it.

Please read my update. I completely switched the GCC-14 distribution to another one, and I haven't seen any indication that you've tried it yet.

Please, check C++23 (GCC 14-64, msys2).

Please, check C++23 (GCC 14-64, msys2).

It is fixed now. Please, try again.

Please, check C++23 (GCC 14-64, msys2).

Some time ago, there was news here that we supported C++23 by installing GCC 14 from the Winlibs project. It turned out that with this method of supporting C++23, some compatibility issues lead to significant performance drops in certain submissions. So, I tried another distribution for GCC 14: the MSYS2 project. It seems that this problem does not occur with it, and now there are no unexpected performance drops. Let's test this C++23 support option together!

Restart cmd session to reload environment variables. Also check your PATH variable for another gcc before this.

I fixed the print issue, but performance is still poor in some cases performance is fixed too.

Thank you for bringing the issue to my attention. I have implemented a fix. Please check it now.

Hello! Actually, you wrote all these complaints before we conducted our regular cheater search. All cheaters were found using the usual method, and the submissions mentioned in the reports provided only minimal assistance. For example, dozens of cheaters were reported, but we found hundreds. I think it's not worth getting worried until the cheater search procedure is completed. In most cases, your complaints will be addressed simply through the standard process. We are making significant efforts to identify as many cases of plagiarism as possible. We are trying very hard!

That moment when you suddenly want to implement reactions to comments.

For the role of a researcher, I nominate this guy rock-star: https://codeforces.me/profile/Darooha https://en.wikipedia.org/wiki/Daniel_Sleator

We are carefully looking into the situation and will inform you when we make a decision.

On jakao → Ghost scoreboard upload fail, 2 years ago
+12

Please, send me a link to the exact dat-file you are trying to upload. I'll take a look.

+36

In this round, there were three problems based on my ideas: 1974B - Symmetric Encoding, 1974C - Beautiful Triple Pairs, 1974F - Cutting Game.

Give me some time to figure all this out.

Hi, I've fixed it. The rating will be recalculated once the cheaters have been processed.

On Tutis → Golang version, 3 years ago
+35

Here's the update! I've upgraded Go to version 1.22.2 and PyPy3-64 to version 3.10/7.3.15. Enjoy using them!

P.S. After my return from Egypt, you can expect more updates!

+34

Thanks! I'll look into it after WF in Egypt.

I sent you all the details via private message, including the transaction link. In short, at the time of the data upload, a different wallet was specified by you, which you changed later. The TONs were sent to that wallet.

On orz → C++20 is back, 3 years ago
+37

The plan is to begin the migration on Monday morning.

+142

Sorry, but I'm moving the round 10 minutes forward. The training session was very cool and long. I'm running to you from the gym!

On orz → C++20 is back, 3 years ago
+148

TLDR: Very soon, I will migrate the testing servers to Windows Server 2022. I have spent a great deal of effort in recent days to accomplish this.

Good news: most of these issues are not reproducible there (more precisely, we do not know how to reproduce them), and it will be possible to update some compilers/interpreters that could no longer be updated for Windows 7.

Bad news: modern operating systems have deep protections against speculative vulnerabilities (like Spectre, Meltdown, and others), and therefore, on the whole, any solution will work a bit longer (by my estimates, plus 5-15%).

P.S. We will likely switch to LLVM's clang++ instead of g++. It performs better in some tests.

Are there general methods of constructing testcases that can exploit this GCC/Windows bug and therefore slow down solutions (like there are anti-hash tests for solutions using unordered_set)? Are there methods that slow down both contest mode and custom invocation mode (since 249807302 is only slow in the contest mode whereas 253289473 is only slow in the custom invocation)?

I'm not entirely sure about the current situation. I hope that after migrating to Windows Server 2022, there won't be such methods.

If so, should we expect such tests in future contests, will such methods be used during the test preparation in contests?

I hope that after my work, it will not be possible to exploit the operating system/compiler bugs so explicitly. Therefore, when preparing problems, one should not expect that such tests will be specifically constructed.

However, I urge you to clearly distinguish between two different scenarios in your mind: a bug and a feature of implementation. For example, exploiting collisions in hash tables, quadratic tests for sorting in Java — these are examples of using features of the standard library. The presence of such tests is a feature of problem preparation, and they may either be present or absent in the test set.

If this is deemed an unwanted bug, will participants be protected from unscrupulous hacks and tests? (For example, if the solution was hacked or FSTed because of a compiler bug rather than because of anything else, will the author be able to appeal against it and get their solution accepted?)

In any fairly large and complex software, there are bugs. This is especially true for compilers with their aim to optimize the resulting code. Sometimes the line between a bug and a feature is thin and undefined. I am not ready to take responsibility on our part for individual attention to cases of probable bugs in compilers. In other words, if a participant writes code that encounters unexpected behavior from the compiler/OS, we will not be able to individually understand and help in this situation. On the other hand, I am closely monitoring the situation and paying attention to widespread instances of any such problems.

Is there no bug in 32-bit version? If it's 64-bit only, maybe CF team could add a C++20 32-bit compiler? UPD: 32-bit version can also be slowed down, see 253400416.

I think that 32-bit versions are stable. But currently, I am working on resolving the situation more systematically, rather than fighting it using half measures.

Is there this bug in Microsoft Visual C++ compilers (32 or 64 bits)? If so, is it the reason why we cannot see it among the options?

Read the previous response.

Are there generic methods of protecting against the bug (except switching to 32 bits)? I read (and enjoyed) a blogpost by ToxicPie9 regarding this issue, he suggested to add several lines of code to the solution. However, it does not always help: 253290209 still gets TL (and the author said about that).

I really hope that switching to a modern OS (and likely to clang++ as well) will resolve the issue. However, I'm fairly certain that this won't be the last bug in the compiler/OS, and over time, we will discover new ones.

Has anyone a clue what is going on, why do the aforementioned programs run slowly? Is this bug reported to the GNU or Microsoft developers who are responsible for this strange behavior?

I haven't delved deeply into this issue. It seems that in the current setup, even if somehow resolved, the problem will emerge in some other unexpected place. I prefer to try moving forward.

Why does behavior in custom invocation and in contest/gym mode differ?

Many bugs in the world are quite unstable to reproduce. This is an example of one of them. The method of launching for solutions during testing and during custom invocation is the same.

Is this bug reproducible on other platforms? For instance, on godbolt there are some windows compilers. Is there a way to slow down this code: https://godbolt.org/z/5G73dfezK (by cleverly choosing the value of evil or anything else)? UPD: probably no way since Godbolt uses UCRT rather than MSVCRT. UPDUPD: this code is slow on Godbolt: https://godbolt.org/z/WE4Mxv9cK, so UCRT is also flawed.

It seems to me that this problem is clearly reproduced only on specific versions of Windows 7. This does not mean that there are no other problems in Linux or newer versions of Windows. Tests by maxplus (thank you!) show that there are some irregularities in other setups as well, but we expect that they have little impact on the testing of problem solutions. In any case, I plan to make the main setup one in which such problems do not reproduce at all.

Is there surely no such bug on Linux? (Anyway, with regard to Codeforces it is probably a pointless question. I am quite certain that transition of all invokers to Linux is the last thing MikeMirzayanov would prefer to do since it is for sure really costly in time, effort, architecture changes, and/or money.)

Apparently, this particular bug does not reproduce on Linux. However, even on Codeforces, we lived for a long time without noticing it. I don't know and cannot confidently assert what happens on other systems, especially considering the wide variety of versions, hardware, security patches, etc.

Regarding the switch to Linux: yes, it is a labor-intensive and complex task. We have a huge number of problems (checkers, author solutions, generators, etc.) that may not work or work unpredictably on other systems. In some cases, they are written in languages/dialects whose support in Linux significantly differs.

I should note that the number of participants using Windows is twice as many as those using Linux and iOS combined.

We are working to make our problem ratings better. Stay tuned for updates!

Actually, on clist there are many weird ratings too. I can't say it is perfect:

There are more examples like these if you examine clist carefully.

Curious, why does your example give TLE when submitted, but not during custom invocation?

I think so. If you're not sure, just stick to the time-tested 32-bit versions.

Thank you for this investigation. I have introduced a new language: C++20 (GCC 13-64). Please test it. My tests show that this issue has been fixed in this version.

The current plan is to hide all other 64-bit C++ compilers except for the new C++20 (GCC 13-64).

On farmersrice → account deletion, 3 years ago
0

Just erase all personal data from your user account and don't use it anymore. We don't store personal data long-term.

Just stop cheating:

Please, try again now

On orim → help Mike, 3 years ago
+118

6 times. I will repeat, 6 times in different contests you cheated, using someone else's solutions or sharing your own. After each time, you were warned that your account would be banned. What do you want from me? You deserved the punishment.

I fixed this problem.

Thank you. It appears that the issue was not the result of https://codeforces.me/blog/entry/105779 However, it can be seen here https://web.archive.org/web/20210416021306/http://codeforces.com:80/contest/382/problem/D

Thus, the incorrect rendering has been occurring for at least three years.

The most challenging aspect of addressing such bugs isn't the backend repair but ensuring backward compatibility. Fixing this might impact other problems that depend on the current behavior.

A probably better solution would be to correct such issues in Polygon.

Do you know of any other incidents involving similar problems?

You cheated in no less than 4 rounds. And those were clear coincidences. I think you're a hair's breadth away from a ban.

+21

I have switched Kotlin 1.9 to use the JRE 8 32-bit. Now, all your example submissions on 1.9 are as fast as on 1.6, or even faster. Please conduct your benchmarks and share the results.

Additionally, I will reinstate Java 8 32-bit but will remove Java 11 and 17.

You see, I prefer not to maintain more than two versions for a single language (ideally, just one). Supporting additional versions introduces complexities. C++ is an exception due to its popularity. However, even with C++, when adding new versions, I find it necessary to phase out the oldest ones.

+43

Everything comes to an end. We say goodbye to the magic of the New Year. Santa Claus is already actively casting spells to roll back the masquerade. Here are some statistics for you.

A total of 70712 magic masquerades were applied.

rank count
legendary grandmaster 17590
newbie 11895
candidate master 7461
expert 6669
specialist 5600
master 5461
pupil 5193
grandmaster 4159
international master 3433
international grandmaster 3251

Or by color:

color count
legendary 17590
gray 11895
orange 8894
violet 7461
red 7410
blue 6669
cyan 5600
green 5193