It looks like the rule34.xxx downloader isn't working. Direct links to images work, but the gallery and post downloaders don't. My guess is it's something to do with the API, but I have no experience with that kind of stuff.

>>24057 Running from source fixed the problem! It also seems to have fixed a minor issue I had with styles (default_hydrus_dark.qss had some white windows but now it's fixed and displays properly). Everything seems to be working like a charm now, thanks! >I don't know if that sio_flush call is an OS thing or an mpv thing I tried looking it up, what I found is that sio_flush comes from a libmpv2 dependency called libsndio7.0. This library is described as a "Small audio and MIDI framework from OpenBSD, runtime libraries" You can find more info about sio_flush here: https://man.openbsd.org/amd64/sio_open.3#sio_flush However, this still doesn't explain why Python didn't recognize the sio_flush function, maybe it's a version mismatch between libmpv2 and libsndio7.0 (i.e. libmpv2 is too new for libsndio7.0)? Well anyway, thank you again and keep up the good work!

>>24063 >However, this still doesn't explain why Python didn't recognize the sio_flush function, maybe it's a version mismatch between libmpv2 and libsndio7.0 (i.e. libmpv2 is too new for libsndio7.0)? Actually upon exploring that manual more I realized what the problem is: libsndio7.0 comes from OpenBSD 7.0 but that version does not include a sio_flush function! https://man.openbsd.org/OpenBSD-7.0/amd64/sio_open.3 The sio_flush function was actually introduced in version 7.2: https://man.openbsd.org/OpenBSD-7.2/amd64/sio_open.3 I think the older versions of libmpv2 utilized sio_stop instead of sio_flush but the new version of libmpv2 I have thinks libsndio7.0 includes a sio_flush function which it doesn't. It's possible that reverting back to an older version of libmpv2 would've solved this problem. Just putting this possible workaround out there in case anyone experiences this problem in the future.

I updated from 623 to 635 and now my "drag and drop" functionality is weird. Usually I could drag and drop images from hydrus to, say, an empty firefox page and it would load the file, or drag it to a 4chan page and it would attach the image to my quick reply window. Now when I do that, no matter the website, it will redirect me to a firefox "error" page, saying that I can't access the file. Drag and drop also doesn't work with windows explorer i.e. I can't "quick export" a file by dragging and dropping it between hydrus and explorer. Is anyone facing the same issues? I'm running W10 if that helps

>>24065 Seems to be an issue with file permissions. Check the permissions on your Hydrus and database folders then try running web browser or Hydrus or both as administrator. Also how exactly did you update your Hydrus?

>>24066 >Check the permissions on your Hydrus and database folders The folder was tagged as "read-only" (I think it has always been like this) so I unchecked that box. I am now also running the hydrus client as admin, and the issue persists. >how exactly did you update your Hydrus? I a run a backup, close the program and then run the .exe windows installer.

>>24067 (me) In the meantime I rolled back to v623 just to check and it works fine, I think I'll use this version for a bit more kek

>>24034 >Can you provide me an example URL so I can test it on my end? sure, here's a (nsfw) example I recently encountered. https://kemono.cr/fanbox/user/1504886/post/7792219 The titles and post texts seem to work fine, but the filenames are all nonsense letters and numbers. I also found this issue on the gallery-dl repo that seems to be about the same problem. hopefully this is useful to you. https://github.com/mikf/gallery-dl/issues/8031

>>24058 They did change the API to require an account, but I don't think that's the specific problem here. The downloaders don't use the API. I haven't run into any problems downloading posts. But I have noticed when I open a lot of pages at once in my browser it may ask me to solve a captcha. So I would recommend that you increase the delay between page fetches for rul34.xxx in network > data > review bandwidth usage and edit rules, and also add a normal non-hydrus browser header for rule34.xxx under network > data > manage http headers.

I had a good week with a couple frustrating bumps. I managed to get multi-domain URL Classes working, and some other fixes and little improvements, but unfortunately I ran into a hard problem moving the macOS App to a new build process, and I'm pretty sure tomorrow's App will be the last I put out. I will have a guide tomorrow for how existing macOS users can migrate to a source release. Otherwise, the release should be as normal tomorrow.

>>23993 >I fired up Inkscape and made a new SVG 'cog' icon, and a new one for 'retry'. I'm not the best artist in the world, but this is the another experimental step in migrating away from static 16x16 PNGs. Aren't there some icons under a Creative Commons license you could use?

https://www.youtube.com/watch?v=H3v9unphfi0 windows zip: https://github.com/hydrusnetwork/hydrus/releases/download/v636/Hydrus.Network.636.-.Windows.-.Extract.only.zip exe: https://github.com/hydrusnetwork/hydrus/releases/download/v636/Hydrus.Network.636.-.Windows.-.Installer.exe macOS app: https://github.com/hydrusnetwork/hydrus/releases/download/v636/Hydrus.Network.636.-.macOS.-.App.zip linux tar.zst: https://github.com/hydrusnetwork/hydrus/releases/download/v636/Hydrus.Network.636.-.Linux.-.Executable.tar.zst I had a good week with some frustrating bumps. URL Classes are now multi-domain, but this is the last macOS App release I will put out. Full changelog: https://hydrusnetwork.github.io/hydrus/changelog.html macOS App Github are retiring the old 'macos-13' runner that I use to make the App. For a few weeks now I have been trying to get the 'macos-14' runner, which is now Apple Silicon, to build both Intel and Silicon Apps. Unfortunately, I have run into hard problems with both (we've been skating by with a hacky App structure for ages, and Silicon is much stricter about internal layout and signing requirements) and do not have the time and macOS expertise to fix them. I'm not sure if the anon-signing I was planning would even work long term. Thus I will not be putting out new Apps with the 14 runner. I am sorry! This means that if you are on the macOS App and want to continue using hydrus, you can either stay at this v636, or you can migrate to running from source. That may sound intimidating, but it is much easier these days, and with luck it should shake out to just opening a terminal, checking a couple of things, and running a couple of scripts. A macOS source installation should have much better compatibility than the old Intel App, which was getting Rosetta'd on any new Silicon machine. Please check the guide here: https://hydrusnetwork.github.io/hydrus/running_from_source.html And there's a mini guide specifically for macOS App migrations here: https://hydrusnetwork.github.io/hydrus/running_from_source.html#migrating_from_an_existing_install Please let me know how it goes and if you run into any trouble. I regret this has come so quickly. The earlier 'macos-14' migration work went well, thought I would be launching a 'future build' test today, but unfortunately there were too many barriers at the end. There's a brownout period for a few more weeks, so if I need to make a hotfix, I can get another App out, but since we know the writing is on the wall, I have drawn a line here. It is possible that a user who knows more about Apps and macOS will put together their own App building solution, much like the flatpak or AUR releases. It will not be me, I'm afraid. nice stuff URL Classes now support multiple domains! I've written a new panel for the 'edit URL Class' dialog that starts off simple but will expand to let you put in multiple fixed domains and/or multiple regex rules. This panel looks after 'match subdomains' and 'keep matched subdomains' too, and it has a couple boxes for testing if domains match your current rules. If you are a downloader maker, please feel free to play with this. I recommend starting slow, since this is new ground and I'm not sure if there's some weird URL-matching situation I haven't thought of yet. Let me know how it goes. Thanks to a user, EPUB files now have thumbnails! Should work for any EPUB 2 or EPUB 3 files that actually have one. Please send in any files that you think have a cover image but don't show in hydrus. For duplicates auto-resolution rules, I am retiring 'set B as better' as an available choice. It was buggy and unwise (non-KISS) to try and include it. If you have any rules that were set this way, they will be paused and set to 'A is better' on update, and a popup will tell you what to do next. Greyscale jpegs (actual greyscale palette, as in 8 bits per pixel) are now handled properly in the extra file metadata window and the duplicate filter. Previously, it would say their subsampling was 'unknown'. next week With multi-domain URL Classes clear, I want to get back to finishing the duplicates auto-resolution project.

>>24086 >re: B is better I was actually going to suggest to you that you should get rid of this option entirely, since I can't think of any useful purpose of it, and it would likely just lead to a lot of confusion and incorrect choices being made in the auto-resolver. I was surprised when I saw that you could set B as better. If they're a better/worse pair, A should always be the better one. so yeah I think this is the right choice you made here.

Sometimes when searching files I see a bunch of thumbs that look like this at the bottom. These don't have tags or anything but a file ID or a sha256, and they cant be deleted. What are these?

>>24089 Those should be deleted files, that's what they looked like until recently, now they use the blurhash for a very blurry representation of what the picture used to be. Went to take a pic of some of mine but they seem to use a different icon, not sure what that's about.

>>24089 Right click on the tag you searched for, and go to "manage". It comes up with some command that seems to take care of this. I can't remember what it is called, but just click on it and run it. I've had the same issue and it seemed to clear it.

>>24058 >>24075 this caused my downloader to stop working too but i think i managed to fix it. I copied the api access credentials at rule34.xxx/index.php?page=account&s=options and appended it to the api/redirect url in the rule34 file page url class and it seems to be working fine. im a brainlet when it comes to this kind of stuff so i originally thought i had to put the api and user id into the https headers in hydrus lol

>>24086 >With multi-domain URL Classes clear, I want to get back to finishing the duplicates auto-resolution project. In case this hasn't already been suggested, would you be willing to add "OR" for comparators at some point? Many of my auto-resolver rules so far are almost identical, except that 1 or 2 comparators are different. If there was a way to have an OR block like with searching, I wouldn't have to keep manually recreating basically the same rule 7 times or more. I'm worried that I might enter something wrong each time I manually recreate the rule so it makes me a bit apprehensive. Just being able to make the rule once and have the cases were just one comparator needs to match be together would make this simpler. alternatively, just being able to duplicate the rule with a button like you can for subscriptions would also somewhat work. It wouldn't reduce the number of rules I have to deal with, but it would at least remove the possibility of me recreating a rule incorrectly.

>>24055 Yeah. The subs system is bad at handling certain errors--the whole network engine is--and the popup toaster is brittle. I've been thinking of better error handling at the domain level, but I was actually thinking the other way around--noticing 403s from cloudflare et al and delaying any request on that domain gracefully for the next n hours. I guess your subs here have frequent connection errors, so a similar sort of 'domain panic' that I have already implemented jumps in and says 'too many errors on this domain/sub, waiting ~24 hours'? I think moving towards more error-handling awareness and user options is the way forward here. You are correct that you should be able to say that a particular domain or subscription has a certain propertly, like, 'connection errors are typical, don't bail out' or 'only wait two hours'. I've been thinking about sketching out a skeleton of a 'domain manager' so I can start hanging domain-specific options off it. Maybe this is a good place to start. I don't know what the hell to do with the popup toaster. Maybe I should just bite the bullet and have a spawnable window that can scroll all the current messages or something. >>24063 >>24064 Hell yeah, glad you are working. That .so file stuff is a little over my head, but it is a typical pattern of 'the local OS has a dll that does one thing, but the github Ubuntu runner wanted to call another thing'. This often happens with Arch hydrus users, where, much like with your situation, they tend to get updates a little early and so we see latest-version regressions. I did try the very latest windows mpv dll here https://sourceforge.net/projects/mpv-player-windows/files/libmpv/ on my dev machine and it was fine. I was wondering if your version mismatch would be a mpv-python thing or an OS one. Not sure strictly what version my test dll was, but I guess we'll see how it all goes. Since the recent macOS stuff proved a bust, I should move onto migrating the Linux build up to the next version of Ubuntu (currently 22.04) to keep with the times better. >>24065 >>24068 Thank you for this report. I think you are probably getting caught by v624, which updated Qt: https://hydrusnetwork.github.io/hydrus/old_changelog.html#version_624 I guess the DnD code changed and now Win 10 isn't happy. Unfortunately, once I fire off a DnD, I actually lose control of most of the program, so it is usually Qt's fault. Windows has some funny rules about drag and drop. The basics are that normal programs cannot DnD to admin mode programs (and I think vice versa), with the exception of OS stuff like Explorer. I've also encountered problems where Discord will accept a 'move' flagged DnD but not a 'copy'. It was some sort of internal discord permissions policy. There is a BUGFIX checkbox under options->exporting that you can try if you like, but I can't promise your Firefox has the same issue. If you haven't got it on, try the 'copy files to temp folder...' one too. A dumb solution is to DnD your stuff to an explorer window and then to your problem program. If this is a critical issue for you, I recommend moving to running from source and then selecting which PySide6 (Qt) version you want under the advanced mode (6.7.3): https://hydrusnetwork.github.io/hydrus/running_from_source.html

>>24071 Thank you. I'm not too familiar with the kemono downloader, but it hits up this API call, right? https://kemono.cr/api/v1/fanbox/user/1504886/post/7792219 When I hit that in hydrus now in debug mode, it seems to deliver an explicit Content-Type with utf-8 charset and my decoder has no trouble with using that. I've attached the JSON that my hydrus saw when it downloaded it. All the unicode kana looks good. I don't know if the filenames are different from what they would normally be, but I see this part, for instance: "file": { "name": "AhLUPmDlNiQN3FfWcG3VwbDQ.jpeg", "path": "/c1/61/c1613404e4e97051f01bcf949478d524e6c3b8942bb3f4c0f60aa252c4544f5a.jpg" }, Which seems to refer to this file that I can click from from the main post: https://n1.kemono.cr/data/c1/61/c1613404e4e97051f01bcf949478d524e6c3b8942bb3f4c0f60aa252c4544f5a.jpg?f=AhLUPmDlNiQN3FfWcG3VwbDQ.jpeg Any chance that kemono have fixed this encoding issue on their end in the past week? If it is still broken for you, what do you see/parse in the test box of the parser in network->downloader components->manage parsers->kemono-whatever and what do you expect to see?

>>24075 >>24100 Is this with the user-made downloader here? https://github.com/CuddleBear92/Hydrus-Presets-and-Scripts/blob/master/Downloaders/rule34.xxx%20api%20-%202021-06-14.png I have had reports that the default hydrus rule34.xxx parser (which just hits up the normal post/gallery html) has been getting worse and worse recently as they have switched on stricter CloudFlare gatekeeping. It often fails with the typical CF 403 response. If getting into the site with your browser and then copying cookies over to hydrus with Hydrus Companion or similar is a no-go, then it sounds like the API may be the way, although I suppose you'll have to make an API key with your account because of their recent API policy change. >>24083 I'm mixed. There are a bunch of nice sets out there, but I always run into the issue where a particular free set has a very distinct style but doesn't have a good icon for x, so I end up with a pretty gross pastiche of different styles. This is what we have right now in several places. Also, most of the sets I have seen that have high numbers, like 3,000, are small pngs from the old days. Most of the hydrus png icons come from this excellent free set: https://github.com/legacy-icons/famfamfam-silk but as I move to SVG, I thought I'd try my hand at getting exactly what _I_ wanted. I was an amateur anime vector tracer in /wg/ about a million years ago, so I know the basics. I just have no artistic skill or sense of aesthetics or polish, hahaha. If you know any great free svg icon sets that'll look good in hydrus light and dark mode, let me know. I know one guy who is using the new /db/static override thing to create a suite of custom icons for his specific QSS stylesheet. I'm still thinking how we might package a combined stylesheet and icon set for hydrus; at the moment this is all a huge mess, but I'm ok to still be in the 'playing around' stage. I still need to start scaling buttons with these new SVGs to see how Qt handles it all. >>24087 Yep, just me trying to be overly smart and wrapping around to dumb again. 2024 has been the big KISS year, and I'm really happy with the push. >>24089 What's the 'file domain' in that search page? Are you searching in 'my files' or something like 'all known files with tags'. Hydrus is quite happy to search across filelists that include files it once had but has since deleted, or it knows about because of PTR syncing, and when there's no known thumbnail or blurhash, you'll get something like that. If you are seeing null thumbs in 'my files' or 'all my files' etc..., then there's an error somewhere. This can happen after database damage, if you've had that in the past, and from some legacy logical code errors I've done in tag processing and so on. I notice your two thumbs there don't have trash icons, so it doesn't seem like they are known previously deleted files. Them only having a hash and an id is also suggestive of this being a logical mess-up on my part somewhere, maybe from years ago. If you do not sync with the PTR, try hitting up the database->regenerate menu. Go for 'tag storage (all, deferred siblings and parents)', and the local hashes and tags caches. Let me know if that fixes things. If you sync with the PTR, then the tag storage regen job will take too long, so try to identify with tags, if you can, produce these ghost files, and then as >>24098 says, right-click the tag and hit maintenance->regen tag display. See if that helps.

>>24101 Thanks. An OR logic block doesn't sound impossible, so I'll write it down and look into it and hesitantly say I may back off if it looks like a nightmare behind the scenes. A duplicate button is a great idea. Duplicate resolution rules are trickier than normal objects, but there's no reason I can't wire that up with some extra thought.

>>24102 >24055 i've looked a bit more into the issue myself (the flaky part) and i see that some separated queries are 500'ing much more often than others. i might leave those paused while leaving the rest running...once i'm done dealing with the fact that those separated subs have downloaded a lot more than i anticipated and my nas is now full to bursting...probably will use tagrank for a few days to let the cream float to the top (that sounded better in my head) and prune a good chunk (read: several hundred gigabytes) off the bottom.

>>24102 >>24065 (me) >>24068 (me) Thank you for the reply. I don't mean to add more noise to this issue but I was going to follow your advice of running from source, but before that I decided to do a fresh install of 636 in another directory. Since it's not in C: anymore seems like I can run it without admin privileges and now drag and drop is working just fine. The only thing I noticed that's different is that my rule for DnD is to export it with the 'filename:' tag, and for files that lack that tag, it used to attach the hash of the file. Now, instead, it just gives the file a filename "empty.jpg". This is a total non-issue and I'm happy that things are now fixed, yay

>>24086 epub thumbnails for existing files have a wrong aspect ratio. This can be fixed by running a 'regenerate file metadata' job, then 'regenerate thumbnail if incorrect size'.

>>24086 >Please send in any files that you think have a cover image but don't show in hydrus. I had a good number of epubs with SVG covers from this publisher.

>>24086 >Please send in any files that you think have a cover image but don't show in hydrus. This one has a terrible cover image, but it's an example of the Apple IBooks flavor of epub, which has a different tag to specify cover. https://wiki.mobileread.com/wiki/IBooks >>24111

This is an IBook with an HTML cover page. Fuck this format in particular. It's almost the same as >>24113, but names it's cover "Cover Page" instead of "Cover Image", and the cover's mime is text/html. I don't think Hydrus has an embedded HTML renderer. Maybe it's okay to cheese it by picking an image from the HTML if there is one - à la 'simple downloader'.

>>24113 This patch breaks books that have both an IBook HTML cover and an epub2 cover image. IBook cover should be used only as last resort I guess.

>>17183 >why yes I would like to add more metadata to files shared across the internet >no I'm not some glownigger trying to track and deanonymize anons It's a good thing I'm not terminally online and can't relate to the usecase, otherwise I might fall for such bait.

>>24120 I don't think you have any idea what you're talking about. Where exactly does Hydrus add metadata? I think it can let you manually edit a couple things maybe, but a file posted directly from Hydrus should have no more information on it than any other file, besides defaulting to the hash as the filename and giving away that maybe you use Hydrus. Personally, I retain my filenames as tags when importing and reapply them automatically when exporting, so hash filenames are only used for files from my subscriptions.

I had a good week. I've added system:time tech to the duplicates auto-resolution system, so you'll be able to filter and arrange pairs based on their import time and so on. There's also some new Base64 tools for the parsing system, improved sync for certain large subscriptions, and some stability fixes. The release should be as normal tomorrow.

>>24098 >>24104 Right clicking the tag and regenerating tag display cleared them, thanks!

>>24120 You jumped the gun newfag. Read manual first, then come back and beg for forgiveness.

Hello, I have a some feedback based on my tagging workflow / experience over the last couple years. I am writing these declaratively ("need", "should", etc.), but of course these are my own personal opinions. I do not mean to impose; it was just easier to write that way. I am also curious if others have hit similar pain points. With that said, here is my feedback: My tagging workflow: - focus media viewer window - use user set shortcuts to add common tags - focus "manage tags" window - add manual (non-shortcut) tags Tagging problems: - can't see tags hover-window when content is wide - option to render tags hover-window on top - option to pop out tags hover-window (or equivalent window) - need to swap between "manage tags" window and media viewer window to use custom user set shortcuts - option to use custom shortcuts in "manage tags" window - need to use mouse to enable / disable custom user sets (shortcuts) - option to create custom shortcut for enabling / disabling custom user sets - example: F1 -> toggle "pose shortcuts", F2 -> toggle "clothing shortcuts", etc. - need to use mouse to copy + paste all/selected tags - add shortcut to copy + paste all/selected tags in the media viewer and the "manage tags" window My sibling + parenting workflow: - shift+tab 4 times to focus tags in "manage tags" or media viewer window (from input box) - hit menu key on keyboard and arrow down to siblings / parents - ensure the correct "tag service" tab is selected (with mouse) - type in the desired parent / sibling - shift+tab 8 times to select "add" - shift+tab 14 times to select "apply" (more likely, use the mouse) Sibling + Parenting problems: - no easy way to access the tag context menu from the "manage tags" window - add hotkeys to manage parents / siblings for the currently highlighted autocomplete tag - example: type "red sh" -> "red shirt" becomes highlighted in the autocomplete window, press hotkey to add parent - no hotkey to swap between "tag service" tabs in the parent / sibling manager - add hotkeys to switch tabs - no easy way to select "add" or "apply" in the parent / sibling manager - Pressing "Enter" when the tag completion is empty should "add" the parent(s) / sibling - Pressing "Enter" when the tag completion is empty and no pairs are uncommitted should "apply" - Tab order should prioritize (sibling / parent) and (replaced / ideal) tag input boxes

>>24120 >doesn't know how to use Hydrus >trusts internet tags LMAO!