[X] My Assistant

I see that you also haven't found a way to disable the LEDs from within OpenWRT.
Mine doesn't need active cooling but i use my server for torrents and stuff. I only needed a high powered router from SQM.

Fortunately, the LED's dim themselves out to near invisibility over a period of a couple years.

Not sure why; are they OLED's or something? I've seen a lot of modern things with white LED readouts go dim over a pretty short period of time.

Also nice, you have the 3200. I guess that makes three of us at least with this router series.
I was able to get the power LED to turn off; just not the others.
I wouldn't have minded if they were little green dots like on the WRT54G (my beloved).

Another thing I'd like to do with these 1900AC V1's is to replace their WLAN cards with Atheros ath9k ones, since the V1 has two mini-PCIE slots (later routers just have the WLAN chips soldered to the main board). Then I don't have to deal with Marvell anymore.

BTW - the fan is on stilts because it's quieter this way than if it's pressed right up against the grate. It's standing on RCA panel mount jacks as spacers with cotton swabs stuck through them to act as pins. I discovered that the solder tab on the RCA jacks is the perfect shape to slide into the holes on the router and make a snug friction fit without causing much damage. And no glue was required, so it's easy to remove.

The swabs are each cut to a specific length so that they extend to touch the board/heat sinks inside and appear level from above.

The fan's just an arctic F8 (the kind with ball bearing - the F8 CO or something like that) with the sticker removed. Unfortunately I can't control speed with the specific fan I have and the way I installed it: straight to the power switch, instead of by [github.com] repopulating the unpopulated components for the fan header that's still on the board as a remnant from the AC V1. I believe the V1 used PWM on the actual power rail to regulate speed (it's a 2 pin fan).

If lm-sensors is to be believed, my 1900ACS is currently sitting nearly idle at about 32°C.

This post has been edited by dragontamer8740: Jun 20 2023, 20:38

The generation when the WRT3200ACM came out was one of the last decent generations that Linksys did.

Nowadays they're trying to pressure users into installing their shitty app on their phones and creating an account for remote management while removing or kneecapping ability to perform true offline setups. The major example of this is their Velop and related systems that use WiFi mesh. Also, their node add functionality sucks balls.

Interestingly, was able to find all sorts of hidden pages and functionality within their normal UI like more in depth WiFi tweaking, disabling remote management access, etc. Annoyingly, they also removed the ability to perform firmware updates manually without a link to the internet, but one can get around it because they didn't really disable it. They just made it hidden.

Seriously though, one doesn't need a fucking cellphone to configure a mesh system. I'm familiar enough with the protocol having had to read the IEEE 802.11 spec to know that all you really need is to add a couple of MAC addresses and the software can take it from there. Linksys isn't the only one doing it too for mesh systems. Wonder why stuff became shitty all of a sudden...

It's finally happening. Infinity, a 3rd party app for Reddit downloaded from F-Droid, will become subscription-ony by next month.

For me this is just a hint that reddit doesn't want me to be on it.

I'd suggest others take that hint, too.

This post has been edited by dragontamer8740: Jun 27 2023, 04:40

I bought a used Dell U2410 (an old, no longer manufactured IPS monitor with a CCFL backlight).

I got into the service menu and it reports that it's been turned on for 421 hours (about 18 days, rounding up). It's super, super bright for a CCFL (~440 nits according to my colorimeter's readings).

By comparison, the one I already had (U2412M) - one that I got new a decade ago - has 9,500 hours on it (about 390 days).

I think I picked a good panel on ebay. For a CCFL, I really lucked out. :)

I have been trying to make a Lemmy account weeks ago. Maybe they fixed the registration by now.

I bought a used Dell U2410 (an old, no longer manufactured IPS monitor with a CCFL backlight).

I got into the service menu and it reports that it's been turned on for 421 hours (about 18 days, rounding up). It's super, super bright for a CCFL (~440 nits according to my colorimeter's readings).

By comparison, the one I already had (U2412M) - one that I got new a decade ago - has 9,500 hours on it (about 390 days).

I think I picked a good panel on ebay. For a CCFL, I really lucked out. (IMG:[invalid] style_emoticons/default/smile.gif)

did they specify the very good state of it at all? it seems every time a listing claims its great its terrible, and when something is in anunknown state or not mentioned is the only time I get the really nice stuff.

in thoughts/struggling:
it seems a lot of server boards use 'standard' atx mainboard connectors but then just change up the pins. didnt know this was allowed, they fit perfectly with a regular 20pin but then its nearly only 12v


server hardware is a struggle

This post has been edited by cate_chan: Jun 27 2023, 21:58

did they specify the very good state of it at all? it seems every time a listing claims its great its terrible, and when something is in anunknown state or not mentioned is the only time I get the really nice stuff.

Sent you a forum PM with the ebay link in case you want to see in more detail.
I do usually find your observation to be what happens in practice.
But in this specific case, the photos of the hardware looked good (could see the matted reflections on the screen, didn't see obvious dirt/scratches - also was described as something that had belonged to the seller's daughter, or something like that).
They did also say it was "in great condition," although that means less to me than photos.
They didn't show pictures of it on, which wasn't great, but the angles they shot it at were good for seeing overall quality.

After shipping, it was ~$120 - but given that they didn't detach the stand to ship the box was quite large so I'm willing to believe a lot of that was actually shipping cost.

I guess because it's a CCFL it's just not as desirable as the newer UltraSharp™'s are.

They also described in the listing how they intended to ship it.
It shipped fine, but unfortunately the packaging job also made me want to become an ecoterrorist.
[postimg.cc] (IMG:[i.postimg.cc] https://i.postimg.cc/2yKgjV4c/urge-to-ecoterrorism-rising2-sm.jpg)

I like that it has a shit-ton of video input choices:

• 1x VGA (15KHz supported - so I can use my Amiga or play my SNES on it!)
• 2x DVI-D
• 1x DisplayPort
• 1x HDMI
• 1x YPbPr component
• 1x composite (I am assuming NTSC and PAL are both supported)

in thoughts/struggling:
it seems a lot of server boards use 'standard' atx mainboard connectors but then just change up the pins. didnt know this was allowed, they fit perfectly with a regular 20pin but then its nearly only 12v


server hardware is a struggle

Interesting. Fuck them, but interesting. I wonder if all the 12V pins are tied together or going to different on-board regulators.

This post has been edited by dragontamer8740: Jun 28 2023, 00:50

Sent you a forum PM with the ebay link in case you want to see in more detail.

nice piece of equipment as I mentioned in the message.

Interesting. Fuck them, but interesting. I wonder if all the 12V pins are tied together or going to different on-board regulators.

its all tied together they seem to simply be doing it to get more power through, since most of the server power supplies just have big terminals for 12v they have to split it out in the thing that splits the cables out, or on the board. it seems hp went for splitting out the voltage on the board that also has the cables, but supermicro just does it all on the board and just shoves all the power supplies beefy 12v lines over a bunch of standard cables for the atx connector.

I've also found a worse offense in non standard wiring, consider the standard atx cpu plug:

now wait for it, the supermicro boards, which might I remind you FIT PERFECTLY with the actual standard:


I didnt want to believe it and looked at some reference pics, but they're really using a reverse wired 8 pin atx:


quick extra thought:
been using my old fluke 77 with the auto hold feature (beep and hold last stable non 0 measurement) and it's been great for this thing where I'm holding cables in weird places and cant see the display until after I get back and get the cables off. wonder why its not more common on all meters.

This post has been edited by cate_chan: Jun 29 2023, 21:12

Its a robust single rail design right?

Got lucky with a non scalpers gpu

Its a robust single rail design right?

yes instead of the psu supplying all the voltages its just 12v+ and the other voltages are split on a dingus board that splits the cables (hp) or on some random parts of your motherboard (supermicro)

Does the digital yuan really have an expiry date?

oh yeah, sign me up for that... I totally want government to just expire my money.

I didnt want to believe it and looked at some reference pics, but they're really using a reverse wired 8 pin atx

had a moment here i think, some pinouts show it reversed, the cables on some of my plugs are reversed. but they were actually the right polarity all along.
only found out after I alraedy repinned them so now I can do them again. at least finished the main 20 pin atx connector and then I can finally see if this new board posts at all or if it needs more random connectors, all so tiring.

in other tiring thoughts, not all that great with twitter atm platform just pretty much got nuked. some are fleeing to misskey.io and others instagram. but neither have everyone I followed on twitter nor the same format. bit of a struggle all around

I remember seeing "goodbye twitter" trending when I went to scrape some artists yesterday.
Guess that's related?

Also one artist I liked no longer has an account there. Wonder if it got deleted by them, by twitter, or by something else.

Latest:
"Elon Musk imposes daily limits on reading posts on Twitter"
"Elon Musk Places 'Temporary Limits' on Unverified Twitter Accounts"

HAHAHAHAHAHA

If it's based on accounts, rather than IP's, I suspect a lot of new ones are about to be made. Although they'll have to sidestep the phone number thing somehow, I guess.

Anyway I have two "unverified" (free) accounts, and both are quite old, so I should be able to continue business as normal given my low rate of scraping.

The limits appear to be 400 posts per day for new accounts and 800 per day for older ones.

Side thought:

I want to kill every artist who is moving from twitter to instagram. That's even worse.
Whenever I'm asked "what's your insta" or "what's your discord" i die inside.

(IMG:[i.postimg.cc] https://i.postimg.cc/vZNnZn15/F0-C-qk-Xg-AId-WRd.jpg)

Also I can't get gallery-dl to work with searches by date anymore (404 errors no matter what I try to do). So I've been feeding in tweet ID's manually by scraping from the search results in a browser for gallery-dl to grab.

This is for tweets older than the weird cutoff twitter has where it stops "scrolling."

This post has been edited by dragontamer8740: Jul 2 2023, 22:35

so now that things seem to be going slightly more okay at twitter it seems it was just a coverup of them not paying their bills and migrating, or them breaking some internal services.
also seems they changed some server-side code which caused certain clients to be stuck on refreshing effectively ddosing it.
no real sources yet just whats going around.

so just like reddit it might not be dying (yet). but another good showcase of how quickly it could fall apart entirely.
and I got to explorer some fediverse messing about with a bunch of japanese artists, all worked out.

No one asked, but here's a C program to turn the tweet ID's from twitter (post-november 2010) into timestamps. Tested on 32-bit and 64-bit x86 Linux but it should work basically anywhere.

CODE

#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <string.h>
long long int timecalc(long long int tid);

/* only for "snowflake" tweet ID's */

int main(int argc, char **argv) {
  if(argc > 1) {
    int i=1;
    long long int timestamp;
    time_t seconds;
    struct tm epoch_time;
    char buf[80];
    while(i < argc) {
      timestamp=atoll(argv[i]);
      seconds=(int)timecalc(timestamp);
      memcpy(&epoch_time, gmtime(&seconds), sizeof(struct tm));
      /* strftime(buf, sizeof(buf), "%a %Y-%m-%d %H:%M:%S %Z", &epoch_time);*/
      strftime(buf, sizeof(buf), "%Y%m%d_%H%M%S", &epoch_time);
      
      /* printf("%s (%d)\n",buf, timecalc(timestamp));*/
      printf("%s\n",buf);
      i++;
    }
  }
}

/* convert snowflake ID to Unix timestamp */
long long int timecalc(long long int tid) {
  return (tid >> 22ll)/1000ll + 1288834974ll;
}

No one asked, but here's a C program to turn the tweet ID's from twitter (post-november 2010) into timestamps. Tested on 32-bit and 64-bit x86 Linux but it should work basically anywhere.

pretty neat, similary to no one asking. I did some recreational golfing/refacturing of your code. but I gained 2 hours in the process

CODE

[cat:~]$ /tmp/cate 1675894539964358656  
20230703_174928
[cat:~]$ /tmp/dragon 1675894539964358656  
20230703_154928

the offending source:

CODE

#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <string.h>

/* only for "snowflake" tweet ID's */

int main(int argc, char **argv) {
  if (argc < 2) {
      return 1;
  }

  char buf[80];
  for(int i = 1; i < argc; ++i) {
    strftime(buf, sizeof(buf), "%Y%m%d_%H%M%S",
      localtime(&(time_t){(time_t)(atol(argv[i])>>22l)/1000l+1288834974l}));
    puts(buf);
  }
}

NOTE: I know I could correct the offset value to 1288827774l to fix this, just not entirely sure what caused it. I think it might be the casting to int and or gmtime's doing

This post has been edited by cate_chan: Jul 3 2023, 22:00

pretty neat, similary to no one asking. I did some recreational golfing/refacturing of your code. but I gained 2 hours in the process

CODE

[cat:~]$ /tmp/cate 1675894539964358656  
20230703_174928
[cat:~]$ /tmp/dragon 1675894539964358656  
20230703_154928

the offending source:

CODE

#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <string.h>

/* only for "snowflake" tweet ID's */

int main(int argc, char **argv) {
  if (argc < 2) {
      return 1;
  }

  char buf[80];
  for(int i = 1; i < argc; ++i) {
    strftime(buf, sizeof(buf), "%Y%m%d_%H%M%S",
      localtime(&(time_t){(time_t)(atol(argv[i])>>22l)/1000l+1288834974l}));
    puts(buf);
  }
}

NOTE: I know I could correct the offset value to 1288827774l to fix this, just not entirely sure what caused it. I think it might be the casting to int and or gmtime's doing

This is a difference of locales. gmtime() uses UTC; localtime takes your timezone settings into account.

I also suspect yours may fail to work correctly on 32-bit systems because of a smaller 'long.' I believe I chose 'long long' because it would be large enough on both 32-bit and 64-bit linux (at least under x86 and PowerPC).

Leaving for Europe tomorrow, so I might not have time to check that for a bit. Try messing with the 'TZ' environment variable in the meantime. I think localtime() checks it.

Nevermind. Something came up and I'll be staying here. In Europe now.

This post has been edited by dragontamer8740: Jul 7 2023, 01:45

Thought:
"Ha, the image in the signature of dragontamer8740 is a "variable" one. I remember when I was young and I had one like that, that worked through a .php renamed as an images file and loading other files. That's how it was done in the times of dinosaurs. I wonder how it is done nowadays?"

Yeah, it was long.