Technically a feature request, but I haven't seen an official thread for that: the 'set as avatar' link needs a confirmation window. Currently you're one misclick away from accidentally changing it.

Enable Roaming seems to be missing from the login page, this is useful for people who use vpns so when one area goes down they can switch to another without being logged out

stars12 said:
Enable Roaming seems to be missing from the login page, this is useful for people who use vpns so when one area goes down they can switch to another without being logged out

i think roaming is always enabled now

Is the blacklist supposed to persist in one state across pages until switched? Because my blacklist has ended up disabled over multiple pages after viewing a blacklisted image once and I have to re-enable it constantly.

bipface said:
even with an PRNG which outputs non-repeating values, you still need some way of mapping these to actual posts among the search results.
if you used it as a seed for the current get_random_posts algorithm, it would still lead to duplicates, so it offers little advantage compared to using a simple PRNG.

get_random_posts

is optimised with the assumption that all it needs to do is grab one page worth of posts.
in general, to have a randomly-ordered sequence in which you can go directly to any page, the DB has to read all the results which precede that page; that's simply the nature of random ordering, there's no way around it.

whether the random seed is persisted anywhere, and whether the RNG produces duplicates or not, are only a very small part of the solution for random-order pagination.

Let's see if my thinking is right. Say, you generated a list of post numbers, as integers, and then used the PRNG to select the member of the list. I'm thinking either you use an array and deal with (small but repeated for every instance) RAM space (or regenerating the list if needed), or you have an object-oriented implementation that has the overhead of all those pointers. I guess it still has that problem with having to do the entire list. However, keeping a cache of each tag's post IDs with ability to append more as site is updated would let you merge them using the equivalent of intersection in set theory. You'd still be stuck with the sort or insertion costs. :/ Yeah, guess not worth the effort. Quick and lazy fix: Make "next" button just refresh the page? :D Also, you'd get different order on newer posts appearing in the mix, even if you used a repeatable set of random numbers.

“Reply” seems to work here in the forums, but on normal posts when pressed it brings me to the top of the page and doesn’t add the other commenter’s quote to the text box.

iOS 13.4 iPhone Xs
Google App ver. 97.0.298258560

deya said:
“Reply” seems to work here in the forums, but on normal posts when pressed it brings me to the top of the page and doesn’t add the other commenter’s quote to the text box.

iOS 13.4 iPhone Xs
Google App ver. 97.0.298258560

That "back up to the top" happens too much for a bunch of things. Happens when I (unnecessarily) have to press a button just to display tags in the comment section as well

I have a lot of things blacklisted, and it makes almost every page just a few posts. I use to have very full pages before the overhaul, but now I have to look at four to ten pieces of art before going to the next page, which is really annoying.

dragonwithblankets said:
I have a lot of things blacklisted, and it makes almost every page just a few posts. I use to have very full pages before the overhaul, but now I have to look at four to ten pieces of art before going to the next page, which is really annoying.

this definitely didn't change, it's probably just luck of the draw that you're now getting pages with many posts that are tagged with stuff you have blacklisted. the only real possibility is that somehow your "Posts per page" value got changed somehow, but blacklists on the posts page works more or less the same now as it ever has.

With the removal of the XML API, there's no way to get a total post count for a search via the API (it was the count attribute of the <posts> element).

tonycoon said:
With the removal of the XML API, there's no way to get a total post count for a search via the API (it was the count attribute of the <posts> element).

adding to this, danbooru has the endpoint /counts/posts.json for this purpose, but it doesn't seem to be present on e621

I can't edit this post at all https://e621.net/posts/757231

"Description is too long (maximum is 50000 characters)

Log ID: (none)"

bipface said:
adding to this, danbooru has the endpoint /counts/posts.json for this purpose, but it doesn't seem to be present on e621

https://e621.net/tags.json?search[name_matches]=canine
You can use this

tsai said:
https://e621.net/tags.json?search[name_matches]=canine
You can use this

That only works to view post counts for single tags, not complex multi-tag searches.

The colors of Favorite Mode and Edit Mode are swapped. Their order in the dropdown list may also have been swapped. Favorites should be green.
Edit: looks fixed now ( '.')b

So do a range of mod actions generate a false 'last post by', or only locking? (for example, NMNY locking a thread seems to set the thread to show NMNY as last poster, even though there is no visible post by NMNY)
I ask because in this thread, Millcore shows as the last poster, but the last visible post on page 6 is by Lunacy. Previously, this would mean definitely that Millcore posted but then hid their post. Currently, I don't know whether I should assume that that is what happened, or whether instead some mod action Millcore took caused that.

This could be viewed as either a documentation bug or a feature request, I guess.

lunacy said:
The colors of Favorite Mode and Edit Mode are swapped. Their order in the dropdown list may also have been swapped. Favorites should be green.

edits have always been green for me

Two small things that make logging in annoying: Since the redesign, logging in boots you back to the main page instead of returning you to the page you were at. Another thing that maybe isn't a bug but is still annoying is that for some reason password managers could properly detect the password field but not the username field. Since the redesign the top field is now detected but is glitchy and doesn't work properly.

I've noticed a few things. Here is a list. Some of these may be duplicates of other stuff in the thread, which I haven't read through (apologies), please disregard any such duplicates.

- deleted posts are now viewable through the API, and expose certain metadata that was not available before, including the file size, dimensions, extension, and MD5.
- I can't view my dmail inbox via the API. I receive a 500 error if I try. It gave me a code, presumably that can be used to reference site logs, ask me on discord and I will share/attempt to help reproduce
- tag alias objects, in the api, have a confusingly named pair of fields, forum_post_id and forum_topic_id, and i can't make sense of them by looking at them
- the "are you 18" first visit banner has a horizontal scrollbar at the bottom under the text, and it's ugly and doesn't need to be there.
- API views of posts used to include the username of the poster, now they include only the ID, and reference several other users (such as the approver) by ID only, can you change this to include the usernames again somehow
- api inconsistencies: post endpoints return an object containing a post or posts field which has the array of posts, but some other endpoints don't do this (e.g. tags)

one major thing I'd like to request for future updates to the API is version isolation. it's easy to do by just sticking a version number into the endpoint name (e.g. https://e621.net/api/v1/posts.json instead of just https://e621.net/posts.json), that way if you decide to break compatibility in a fundamental way like you did between the current version of the api and the previous one, you can either give people time to update from the older to the newer version during a grace period where the old version continues to work, or you can invalidate all of the endpoints they were using and cut them off cold turkey, both of which are better than mishmashing old-client and new-server together and hoping for no unexpected cascades of bad behavior.

you can see that i mainly care about the api, i pretty much view furry porn exclusively via the matrix at this point

Blacklist isn't working. I'm using Opera on my phone.

snergal said:
- deleted posts are now viewable through the API, and expose certain metadata that was not available before, including the file size, dimensions, extension, and MD5.

That's only if you search for status:any or status:deleted , not a bug.
It's pretty nice, I can finally get the sources to my deleted favorites via the api.

bitchassfurboy said:
Blacklist isn't working. I'm using Opera on my phone.

Might it be an issue in how you entered it?
cat dog will only blacklist images which have both tags, for example, so if you want to blacklist images with either they need to both be on seperate lines:

cat
dog

bitchassfurboy said:
Blacklist isn't working. I'm using Opera on my phone.

Does it have commas?

Pool history is broken. Clicking the link just says "An unexpected error occurred."

wat8548 said: Pool history is broken. Clicking the link just says "An unexpected error occurred."

Some work, some don't. Weird.

In the bug resolution tracking thread it says range search syntax is fixed. However anything to do with dates i.e. "5_days_ago", "yestermonth", "date:(date 1)..(date2)" etc still doesn't work. You could click the example links in the search help cheatsheet to verify.

HTTPS doesn't work on certain older websites. The source link will break if they get changed to HTTPS.

Example:
http://www.jurassicbeauties.com/
https://www.jurassicbeauties.com/

The site is offline for the Booru Slideshow application.

strikerman said:
I'm using Chrome here, so that might be why?

I just tried chrome, it's giving the same error

I'm not fluent in code, but I'd like to say the site has images that don't load, I'm able to access my keyboard strokes from before the site change and before this I encountered a 500 error, but it was more than a 500 error.

Edit: This seems to be the old system this bug I'm encountering.

Edit 2: I've deleted all cookies from E621 and still my problem persists. I don't know what else I can do to resolve this and I don't know how to contact the admins.

Edit 3: I checked the site on mobile, signed in and it works smoothly, so the issue I'm having is on desktop.

Flagging a post as inferior does not properly mark the superior post as having a child post. The inferior post still displays that it has a parent post.
Additionally, the button on "parent" in "this post has a parent" will return the children of that parent post, instead of the parent post itself.

1. Searching for date ranges now works, cool, but differently than it has before the update.
e.g. date:X..Y
previously it searched from the beginning (00:00) of X to the end (just before 24:00) of Y,
now it searches from the beginning of X, to the beginning of Y.

date:X..X worked like date:X
while now it returns nothing.

on the other hand, id:20..30 now works as expected, including both 20 and 30. thanks.

2. Posts like https://e621.net/posts/8172 still think they have a sample and thus they show nothing unless you switch to viewing the full image.

3.

earlopain said:

earlopain said:
Another thing: this api call https://e621.net/posts.json?tags=id:<27500 status:any errors out. If you remove status:any or set it something else all works fine.

Turns out post #27224 is the problem. You can view it just fine on the site but if it should be returned through the api the request fails. https://e621.net/posts/27224.json
This also explains why it worked without status:any. Deleted posts where not returned.

This now works as well, thanks.
However, there are still some posts for which API calls don't work in the same way, like https://e621.net/posts/525433.json, https://e621.net/posts/694727.json, https://e621.net/posts/945332.json, and at least 250 more.

and to clear something out:

darkalex said:
https://e621.net/posts.json randomly returns null for file/preview/etc urls.

Example dump: https://pastebin.com/UDizhgVi, just search for "url":null

it's not random, at least in my experience. these are 'hard-blacklisted' posts with tags like 'cub', that you need to log in to see their URLs. You can 'log in' in API by generating and using an API key.

One thing I've noticed, and forgive me if this has been posted already, but on mobile (tablet) hitting back on a post sometimes takes me to the bottom of the page causing me to have to scroll up again and find where I left off. It seems to happen randomly so I can't quite pinpoint why it happens. Or maybe it's just because my Galaxy Tab E is old. I haven't really tested it on my Pixel 3.

Also, the simplified view option tend to pop up quite a bit on mobile as well. Not sure if there's a way to disable that through the website since I use it for some websites. A cursory google search doesn't seem to have yielded any results that specify if I can turn it off for specific websites.

When trying to search by description on the "Mod actions" page, your input is totally ignored, instead just pulling up all actions of the chosen types.

taranuka said:
I'm not fluent in code, but I'd like to say the site has images that don't load, I'm able to access my keyboard strokes from before the site change and before this I encountered a 500 error, but it was more than a 500 error.

Edit: This seems to be the old system this bug I'm encountering.

Edit 2: I've deleted all cookies from E621 and still my problem persists. I don't know what else I can do to resolve this and I don't know how to contact the admins.

Edit 3: I checked the site on mobile, signed in and it works smoothly, so the issue I'm having is on desktop.

My issue has somehow been resolved.

I can't seem to flag a post as inferior to another post that's already flagged (mistakenly, in this case). post #2179130 was erroneously flagged as a duplicate of post #2177890, though the former is better quality (larger and lossless). When I try to flag the latter as inferior to former, it lets me submit it and I don't get an error, but it doesn't get flagged.

Relatedly, the page to flag a post doesn't show the image or its information like the old site did.

I've been trying to upload a .webm conversion of an old flash animation. It's 13 seconds in total, but the site gives this error every time an upload is attempted:

error: ActiveRecord::RecordInvalid - Validation failed: video must not be longer than 3600 seconds

cloakedthief said:
I've been trying to upload a .webm conversion of an old flash animation. It's 13 seconds in total, but the site gives this error every time an upload is attempted:

error: ActiveRecord::RecordInvalid - Validation failed: video must not be longer than 3600 seconds

Double-check that it was encoded correctly. Some badly encoded files get recognized as having absurdly long play times, so it may be getting tripped up on that.

watsit said:
Double-check that it was encoded correctly. Some badly encoded files get recognized as having absurdly long play times, so it may be getting tripped up on that.

It plays just fine in any other media player. I had to use Swivel because originally it was considered an interactive .swf and no other encoder could convert that.

cloakedthief said:
…

how about providing the file to us, so we can do more than just guess at what the problem might be

bipface said:
how about providing the file to us, so we can do more than just guess at what the problem might be

Alright, how would you like me to send it?

Shingen said:
However, there are still some posts for which API calls don't work in the same way, like https://e621.net/posts/525433.json, https://e621.net/posts/694727.json, https://e621.net/posts/945332.json, and at least 250 more.

that is all (as far as i know) fixed now, thanks.

cloakedthief said:
Alright, how would you like me to send it?

would be a good choice

every time I click the edit button, the tags textarea grows slightly taller:
https://imgur.com/zyN01ub
firefox 56, win 7

I am unable to view my profile at all; logging out does not help.

Am I an idiot or is anyone else having issues with the related tags when uploading?

Usually when I clicked a specific tag before searching related tags it would bring up related tags for only the tag I selected.

Also is related tags supposed to be limited? It won't let me see any more tags past a certain point when trying to search for related tags amymore.

I don't actually have any bugs to report, I'd just like to guive thumbs-up to the extra RIS links on images' sidebars, the bulk edit request feature, tag auto-completion and the next/previous links based on search terms. The various new options in the profile settings are a great touch too.

post #1152845

More Format Bugs:

https://e621.net/wiki_page_versions/78147

Broken table formatting made the entire cloaca wiki disappear until I encased the broken format code with code blocks. Also headers within quote blocks don't work properly anymore.

I'm sure there are workarounds to this, but it doesn't seem like using invalid syntax or certain combinations of syntax should be breaking the entire page or changing format structure in unexpected ways.

Pool Issue:

Removing images from pools doesn't remove the association to the pool on the image itself.
https://e621.net/posts/331231?q=welcome_freshmen

The image on post #2164085 wouldn't load. I tried erasing the cache and also tried using a different device and browser. You also can't download the image.

bipface said:
webmshare.com would be a good choice

http://webmshare.com/play/EGJBg

Here ya go.

Dunno' if anyone's reported this yet or not (because six pages is more than I'm willing to sift through...) but this seems like a pretty HUGE and GLARING issue.

When in tag editing mode on a post, if you press the down arrow key on your keyboard to select an option from auto-complete, the edit field is completely cleared of all tags except the one you selected from auto-complete, as if you manually nuked all the other tags. This seems like it could lead to a lot of headaches, if people start using auto-complete in this way without paying enough attention.

This pic has the header of a pool to which it doesn't belong. I noticed the next/previous arrows are blanked out, which makes sense I suppose.

post #2078207

I don't know if it's a bug or intentional change, but e926 shows only a small number of posts and an annotation:

33 post(s) on this page were hidden by safe mode (e926). Go to e621 or disable safe mode to view (learn more).

In account settings I have set to display 50 images per page, but even when subtracting images hidden by blacklisting (as before site update - blacklisted images are hidden), page displays only 17 images (50-33=17), presumably because those 33 have rating other than safe.

If I apply filter by adding rating:safe to search it displays correctly (all 50 images that should be there minus those hidden by blacklist), but isn't e926 supposed to be displaying only safe pictures anyway?

Is it intentional change (for images that are not displayed on e926 to still take up place in the posts view when there is no filter rating:safe applied)?
----

And the other thing is bug in css - in e926 (and e926 only - e621 does not have this problem) using Hexagon theme tags completion hints use too dark font color (very dark grey RGB:51,51,76), instead of typical blue-ish (RGB:180,199,217), making them hard to read.

Edit: both of the above are on Firefox 74.0 (64bit)

jacob said:
When in tag editing mode on a post, if you press the down arrow key on your keyboard to select an option from auto-complete, the edit field is completely cleared of all tags except the one you selected from auto-complete

i haven't seen any mention of this one; what browser+version are you using ?
can't reproduce in firefox 56

thevileone said:

Pool Issue:

Removing images from pools doesn't remove the association to the pool on the image itself.
https://e621.net/posts/331231?q=welcome_freshmen

that seems like a silly data duplication issue.
remember boys and girls, do not hold the same data in multiple places, because eventually you, or someone else, will forget to add/update/delete one of them :)

irajedcoyote said:
The image on post #2164085 wouldn't load. I tried erasing the cache and also tried using a different device and browser. You also can't download the image.

when you look at https://e621.net/posts/2164085/events it seems the image was flagged for removal and probably removed (so those images became inaccessible), but i guess later the post was restored, but the images weren't made accessible again, so that's where we ended up.

Currently old unread (by the recipient) outgoing messages are treated the same as unread incoming ones and cause a counter to be displayed next to the Account button in the upper-left.
I resolved that by deleting the offending ones, but I still don't know if the Mark All As Read button will also correct those. I tried sending a new outgoing message to a blocked account but it didn't trigger the same message bug.
Also, the feedback counter isn't reporting properly and blocked accounts aren't listed as such, see here: https://e621.net/users/6572 Feedback:0 yet there is an entry there stating that he's blocked. I hope it's just not displaying the blocked-user text instead of all the blocked accounts being unblocked by the update, that'd be a mess!

In case it matters, I'm on old Chrome 49.

e926 probably doesn't need the 18+ confirmation box.

I'm trying to change my avatar, but it keeps giving me the message: "Comment threshold must be greater than -50000."

All of my favourited posts are mixed up in the new design

corinthian said:
I'm trying to change my avatar, but it keeps giving me the message: "Comment threshold must be greater than -50000."

Nevermind, I fixed it

[Bug] Set order Links broken

Do you still recommend creating a separate bug thread for each bug? I would, because new bug reports would drown in the discussion of existing ones, if we just throw them in here.
I'm having trouble with the /posts.json endpoint ignoring the "id" parameter and /posts.xml just plain 403-ing on me.

Here's a thread I made before I saw this general bug report thread:
https://e621.net/forum_topics/25740

Bugs are actually being tracked on Github now (ie. in a proper issue tracker). NMNY stated that he transcribes from here to there.

So if you want to do things the formally correct way, it's probably to go to Github and post the bug there.

bipface said:
i haven't seen any mention of this one; what browser+version are you using ?
can't reproduce in firefox 56

Not sure which version (how does one check that, again?), but Firefox.

jacob said:
Not sure which version (how does one check that, again?), but Firefox.

Open Help -> About, it should be right there.

Editing posts which have locked tags appears to give you the message "Forcefully added locked tags" even if you made no attempt to remove them.

The sorting method "date:" gives all posts posted exactly that long ago rather than all posts since then.

addicted1234 said:
Open Help -> About, it should be right there.

Gracias.

bipface said:
what browser+version are you using?

Now that I know how to check: Firefox 74.0 64-bit, if that matters at all.

jacob said:
Now that I know how to check: Firefox 74.0 64-bit, if that matters at all.

can you try safe-mode and see if it still happens:
'Help' → 'Restart with Add-ons Disabled…'

Some searches with "date" are not working, such as "date:yesteryear", "date:yestermonth", "date:month", "-date:year".

Others such "date:5_years_ago", "date:5_weeks_ago", "date:decade", "date:2012-04-27" seem to be working fine.

Sets are ordered in when they were posted instead of when they were added to the set, but when using the "previous" and "next" arrows on an opened post the set then follows the order they were added to the set.

On the same note, when searching with the original favorites order, using those arrows follows the chronological posting on the site.

This can get confusing and in my opinion beats the purpose of the "next" and "previous" arrows. I'd prefer if they fixed this according to when they were added to the sets/favorites, but anything goes, I guess. I'll just have to hope.

I don't know if it's been reported already, but searching "user:me" or "fav:me" returns nothing, while using your own username returns the correct set of pictures.