In short, Smokey detected your "command sequence" spam Irrelevant question posted as answer with a link to a product review at the end. and interpreted each of them as a reply command
[ SmokeDetector | MS ] Url-only title, bad keyword in body, blacklisted website in body, blacklisted website in title, body starts with title and ends in url, +7 more (891): nutritionless.com/stim-rx/ by toqlrfaz on askubuntu.com
@iBug - There's also a one off error - My commented feedbacks seem to start here: chat.stackexchange.com/transcript/message/48580529#48580529 -- If other people made comments after commands they probably got hit by a one off. Apologies but I'm both busy atm and interested in helping fix this, but no time available in the next 2 hours. :(
@iBug - Yes, I realize that and that's why I wrote it that way. It is correct above in the backscroll but wrong on MS. See the link in my prior comment:
SmokeDetector 10:05 PM [ SmokeDetector | MS ] Potentially bad keyword in answer (1): Will resetting Windows 10 also reset drivers for touchpad and keyboard? by Sumon Islam Sumon on superuser.com
Rob Rob 10:08 PM sd spam Irrelevant question posted as answer with a link to a product review at the end.
-
Last line says: "SD" "command is spam", comment is...
[ SmokeDetector | MS ] Url in title, bad keyword in body, blacklisted website in body, blacklisted website in title, body starts with title and ends in url, +7 more (888): nutritionless.com/stim-rx/ by toqlrfaz on drupal.SE
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, pattern-matching product name in body, pattern-matching product name in title, +3 more (496): Keto ultra Weight loss diet by Jon ceena on graphicdesign.SE
@iBug I get an error when I try to use that regex (multiple words groups). For what I think it's trying to do (?:\b|_)(?P<words>(?P<word>[a-z]+)(?:[][\s.,;!\/\()+_-]+(?P=word)){4,}) would work. Looks like you already applied a \b, but I'd suggest (?:\b|_) instead, as the pattern appears to be specifically looking for words separated by _, which won't trigger a \b. I'd recommend against the \b at the end. That shouldn't be a big deal, but it might limit some detections.
Interesting fact: my testing indicates using (?:\b|_) at the start of the pattern reduces the number of steps for evaluation by about 70%. Using \b reduces it by 80%, but doesn't catch _foo_foo_foo_foo_foo_, which appears to be something that's intended to be caught.
@Makyen hmm, you're right in terms of robustness. BTW, the multiple groups with same name is not supported by Python re. It's in the regex's docs, though. I have read through it multiple times and employed many of its exclusive features in Smokey code.
[ SmokeDetector | MS ] Bad keyword in body, pattern-matching product name in body, pattern-matching website in body, potentially bad ns for domain in body, potentially bad keyword in body (295): dinner manages, delightful conditions by Tra ted on ham.SE
OK. I can edit my comments on MS and make certain the correct comment is posted where it is supposed to be and is full length --- or is there an error in the code that will be fixed, and my corrections would simply be moved to a different report - so I should leave things as is ?
I was thinking you might (if you have root) run "sed -l 50" on the comment file going back just before my comments that were fouled to check for other 'short comments' from the various rooms and see if others had the same problem with their comments as was discovered with mine. Surely I was not the only one affected.
@Rob Yes, you are not the only one that's been affected over time. Some of the prior instances have been mentioned to individuals when they have happened and the comments removed, others have not. While it's something that would, eventually, be nice to have cleaned up, it's not a high priority for MS admin time.
Understood, not a high priority - and you two were working on fixing something else when I came back --- carry on, I'll fix my 2 or 3 and see if I can spot were they might have missed and landed.
@Rob No, the mapping from words to reports are processed by message, not by report. So sd k - k will give a tpu- feedback to the report associated with the latest message, skip the 2nd message, and another tpu- to the report associated with the 3rd message. If a word isn't recognized as a feedback and the mapped message isn't a report, it's silently ignored.
That's why people occasionally say sd k - k - k - k - k - k - instead of plainly sd 6k.
[ SmokeDetector | MS ] Url-only title, bad keyword in body, blacklisted website in body, blacklisted website in title, body starts with title and ends in url, +5 more (891): pilpedia.com/nitro-ultra-maxx/ by dintew201 on askubuntu.com
The first 4 words in your message were "spam", “Irrelevant", "question", “posted". The first word is a feedback, so it's sent as tpu- to the report associated with the first message. The second word "Irrelevant" is not a feedback, so it's sent as a comment to the 2nd report. The 3rd message isn't a report and neither is "question" a feedback, so it's simply ignored. Similarly, "reported" got sent as a comment to the report in the 4th message.
And that one was spillage - it looks like it took the third word from: "Irrelevant question posted as answer with a link to a product review at the end." and spewed it onto that report. I removed the comment, but you're admin I guess you can see it.
Got it, whole line split instead of landing in one place.
I've gone back a ways past the problem area and find that the remainder of the string went to nul. -- I went past the spot where MS offers this type of info when clicking the "MS" link: "Post not found for drupal/275391. It may have been reported during a period of metasmoke downtime." - in pink!
I understand why we are getting repeated letter and repeated word posts now, ReDoS Explanation / Tests Solutions, and Wikipedia's webpage - this is probably why earlier you were stuck at 100% CPU usage and undo made a similar complaint a few times these past few days.
I'll leave you to it and not interrupt. Thanks again for pointing out the errant comments.
@NobodyNada It was SD. The 100% CPU was reported by @iBug, and fixed, at least for the issue that was present. There are probably other potential issues, but that one should be fixed.
Always use safe input so you aren't accepting repeating sequences of invisible characters on the EOL. Using safe toascii and comparing raw with processed length will tell you if non-ascii chars are being stripped prior to feeding it to regex and risking a recursive death. Set a time limit on regex and handle the exception when it's made.
[ SmokeDetector | MS ] Bad keyword in body, blacklisted website in body, body starts with title and ends in url, pattern-matching product name in body, potentially bad asn for hostname in body, +2 more (463): Complement isn’t for those ✏️ by user159414 on wordpress.SE
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, pattern-matching product name in body, pattern-matching website in body, potentially bad ns for domain in body, +2 more (394): keto blast This to narration to loathe by Andook on wordpress.SE
[ SmokeDetector | MS ] Url in title, blacklisted website in body, blacklisted website in title, body starts with title and ends in url, pattern-matching website in body, +5 more (692): nutritionless.com/stim-rx/ by ygaikotri on wordpress.SE
tpu- feedback received
[ SmokeDetector | MS ] Url in title, bad keyword in body, blacklisted website in body, blacklisted website in title, body starts with title and ends in url, +3 more (492): www.jbngroups.com by kebir jons on graphicdesign.SE
@SmokeDetector While the question does ask for links to where to find lists, this is undisclosed affiliation (see other reported answer where the user says it's their blog), and the user posted 3 similar answers promoting their blog in a short time, which I consider to be excessive promotion, even if they were appropriately disclosed.
[ SmokeDetector | MS ] Bad keyword in body, blacklisted website in body, pattern-matching website in body, potentially bad ns for domain in body, potentially bad keyword in body (295): Low androgenic hormonal or androgenic by alemmi on meta.SE
@SmokeDetector What decides which messages are posted in which rooms? I have noticed that this is a post from Mathematics, but SD did not mention it in CRUDE.
@iBug - The 'user name' attributed to a comment in MS is shown as "SmokeDetector" when there's no command as in @MartinSleziak post 2 dozen blocks above. Your last edits have altered how MS comments are supposed to work.
@EriktheOutgolfer I'm guessing that SD has anonymized Martin's post much as the 'Community user' does on SE sites; of course the correct thing to do IMO is to not accept the comment. Several hours ago there was a bigger fish to fry and while a fix was made for SD/MS comments the focus of the attention was spent elsewhere.
@MartinSleziak - You are not registered on MS (and that link takes almost 30 seconds to load, at times, if you are curious enough to click it; and search for the absence of your name). The good news is that registration is open, but you'll have to wait for the admin to approve some switch somewhere (unless that has changed in the past couple of months).
@MartinSleziak look: you've got your privileges fairly, and, unless you do something bad, such as too many invalidated feedbacks, you don't have to worry
