Code Tour: 2018-09-01 to 2019-04-29

[momijizukamori]

So it's been a while!


Issue 1691: Banning an OpenID user exposes their ext_?? username. (pull request)
Category: Reducing Confusion
Patch by: Darael
Description: When you went to ban an OpenID user, the confirmation pop-up would use the 'ext_#####' designation they have in the backend system, instead of a human-friendly display name. This now uses the display name instead!

Issue 1913: Replace S2 layer editor with something slightly more modern (pull request)
Category: Styles, Getting the Coding Up To Plum
Patch by: rshatch
Description: We had a text editor for S2 layers that was... basically unchanged from what we inherited from LJ (and at that point, was unchanged since its creation, probably). As such, it mostly worked, but was pretty basic, and none of the extra bells and whistled functioned any more. This patch replaces that text editor with a new one with shiny features such as: syntax highlighting! Hotkeys! Auto-complete! Functioning links to documentation!

Issue 2230: TT/Foundation the logout page (pull request)
Category: TT Conversion
Patch by: zorkian
Description: One more step in our slow migration away from (old, badly documented) BML to (new, cleaner) Template Toolkit pages.

Issue 2249: supportdelete priv no longer exists, if it ever existed in GPL'd livejournal code. (pull request)
Category: Code Machete
Patch by: Darael
Description: There was user privilege called 'supportdelete'. It did not actually do anything, and possibly did not ever do anything. This removes it from the list of privs admins can assign, to reduce confusion.

Issue 2265: Basic API + sample routes (pull request)
Category: API, Getting the Coding Up To Plum
Patch by: rshatch, allen
Description: Among the things we inherited from LJ was the XMLRPC API interface. This was ancient, did not do a whole lot, wasn't maintained, and was largely undocumented. In short: bad. Also XMLRPC is a pain to deal with, and is increasingly unused for that reason. We now have all the backend code in place for a new, better-documented, easier-to-use JSON interface, as well as API keys for it and a few basic endpoints as examples. Once we're sure we haven't missed any major issues with this, we can start adding more complicated endpoints, but it's best to make sure the foundation is rock-solid first. This is actually a TON of backend work to do all the setup, and means that the user-facing bits (endpoints) will be a lot faster to create.

Issue 2278: Global backend changes to allow converting /customize to TT (pull request)
Category: TT Conversion
Patch by: rshatch
Description: While we had fancy TT elements for most form components, we were missing one for the color picker used in /customize. Also, it turned out those pages needed a few tweaks on the backend, like the ability to set form elements to nothing, and to be able to next template blocks inside themselves.

Issue 2283: Customize/advanced to tt (pull request)
Category: TT Conversion
Patch by: rshatch
Description: Beginning the long messy path to total TT dominance of /customize by starting with the /advanced pages (which, ironically, were less complicated code to change)

Issue 2356: remove Yahoo Messenger from user profiles (pull request)
Issue 2376: more "Other Sites" upkeep: remove Delicious (pull request)
Category: Profile, Dead Site Graveyard
Patch by: rshatch
Description: Some services have vanished since we last updated our profile pages, and it doesn't make sense to keep listing the ones that don't work any more. These fixes remove Delicious and Yahoo Messanger.

Issue 2358: add discord to profile #2358 (pull request)
Category: Profile, Getting the Coding Up To Plum
Patch by: srukle
Description: ...And some new sites and services have become popular in their place! Now you can put your Discord username in your profile and have it show up nicely under 'Other Sites'.

Issue 2377: whitelist embeds from streamable.com
Issue 2366: add pastebin.com to embed whitelist
Issue 2397: add gimletmedia.com to embed whitelist
Issue 2400: add random.org to embeds whitelist
Issue 2403: whitelist embed new spotify (pull request)
Issue 2402: whitelist embed from strava
Category: Embedding
Patch by: rshatch
Description: People like to show content from other sites in their journals! Now the list of sites you can put stuff from into Dreamwidth pages is longer. And Spotify works again!

Issue 2368: Attempt to make Travis run compiler (pull request)
Category: Getting the Coding Up To Plum
Patch by: zorkian
Description: Adds an integration to TravisCI, so that when people submit pull requests we can run very basic tests to make sure it won't make absolutely everything fall over if we use that code.

Issue 2372: Convert /latest to TT and spruce it up a little. (pull request)
Category: TT Conversion
Patch by: rshatch
Description: So in 2012, I told Mark that if he added Markdown support, I would clean up /latest for him. He added Markdown support that day, and I.... have finally fulfilled my half of the bargain. /latest is now using TT, and has been polished up a little.

Issue 2373: Move stc include to the controller so it's loaded for cached version.… (pull request)
Category: TT Conversion
Patch by: rshatch
Description: It turned out that including CSS in the template of /latest didn't work when it was cached (and caching is good, it makes things faster!), and it had to be in the controller for that page instead. This makes that move.

Issue 2378: properly anchor header navlinks in Tectonic (pull request)
Category: Styles
Patch by: rshatch
Description: If you used header navlinks, and made your header really tall in Tectonic, the navlinks would end up kind of.... floating in the middle. Whoops. Now they stay where they're supposed to be.

Issue 2381: remove fastclick.js (pull request)
Category: Code Machete, Getting the Coding Up To Plum
Patch by: rshatch
Description: Fastclick.js was originally a hacky workaround for an issue with very early versions of Android - on pages not flagged a special way for mobile use, it would delay sending the message that an item had been clicked on for half a second on touchscreens (which is long enough to be noticable). Fastclick.js got around this by making a pretend mouse click in the same spot as the touch click. Which.... worked, mostly. Except Android got its act together, and this library was abandoned, and it turned out it was causing all sorts of bizarre behavior on mobile devices. We had it because it came bundled with Foundation, but nothing else relies on it, so away it went!

Issue 2386: Update year in site footer
Category: Getting the Coding Up To Plum
Patch by: rahaeli
Description: It's 2019, yeaaaaah! Except our site still said it was 2018 in the copyright notice in the footer. This is now fixed (and was made live on production already)

Issue 2390: Librarian's Dream style doesn't use header image properly (pull request)
Category: Styles
Patch by: rshatch
Description: Librarian's Dream has the option to use a header image, but... turned out it was getting covered up by blocks of header color because of the way it was written. Now you can add a header image without having to do a bunch of faffing around with CSS to make it actually show up.

Issue 2391: Don't strip meta tag when 'use layout stylesheet' is unchecked (pull request)
Category: Styles, Getting the Coding Up To Plum
Patch by: rshatch
Description: It turns out we were removing the meta tags that make pages display nicely on small screens when people used custom CSS instead of the layout stylesheets. This was under the assumption that custom CSS layouts were probably not optimized to behave correctly on small screens, which may have been true at one point, but it's now 2019, and people are writing custom layouts with small screens in mind. And there was no way to add the meta tag back in without either including all the layout stylesheets (extra stuff to load) or manually editing S2 layers (finicky and basically a huge pain). Now all layouts include the meta viewport tags, unless you have explicity disabled it (Account Settings > Display > Mobile View > check 'Turn off mobile view for this device')

Issue 2396: Tagnav highlighting not always working (pull request)
Category: Reducing Confusion
Patch by: alierak
Description: We have a feature where you can navigate entries within a tag. It's supposed to keep the tag you're browsing by highlighted, for easier navigation, but this wasn't actually working when the tag had more than one space, or special characters. Now it does!

Issue 2399: Add classes to allow easy targetting of nav module list elements (pull request)
Category: Styles
Patch by: rshatch
Description: One of the most common questions asked on styles_system is 'how do I hide [x] nav menu item?'. Before, there wasn't a particularly clean way to do it that wouldn't either leave weird extra spaces, or be prone to breaking. Now each of the menu items has their own CSS class so they can be targetted and hidden easily.

Issue 2401: typo in faq #175
Category: Docs
Patch by: kaberett
Description: Typos happen to the best of us, but now FAQ #175 will have one fewer!

Issue 2412: checkconfig.pl: don't re-invent version comparison (pull request)
Category: Reducing Confusion
Patch by: nfagerlund
Description: Our script that checks that required dependencies are installed and up to date had some homegrown logic in it that was causing it to report that version 1.5 of something was not up to date with the requirement of version 1.45 (because 5 is less than 45), despite it being a higher version (because in the Perl world, 1.5 is the same as 1.50). We probably needed the homegrown version when the code was originally written fourteen years ago, but Perl now has a nice way to get version numbers for comparison that actually correctly reports 1.5 as being newer than 1.45.


28 total issues resolved
Contributors: Darael, allen, alierak, kaberett, nfagerlund, rahaeli, rshatch, srukle, wahiaronkwas, zorkian - welcome back to a few old faces, and a special shout-out to nfagerlund for their first contribution!

Comments

[azurelunatic]

The text on my profile page is now in exciting different sizes. I draw attention particularly to AO3 & Discord in the other services bit.

[hitchhiker]

ooh, awesome! really looking forward to seeing where the api cleanup goes.

[saga]

Hi Team! Thanks for the hard work and the additional icon slots IT'S NICE. I also like how viewport meta tag is also gonna be usable in custom.

I use the Tectonic layout for one of my accounts and after the code push, it's pushed everything up and the headers now float all over the navigation strip. I also use Firefox. Is there a setting I am missing? Thanks!

[beatrice_otter]

Whereas mine has my name and number all correct just like I cut and pasted it ... but doesn't link like the other site profiles do. Is it supposed to link?

[alexseanchai]

I'm not sure why it would be? Discord operates differently in a crucial respect from sites like AO3 and Instagram. There's no profile to link TO.

[momijizukamori]

This sounds like a bug! I will see about having a look later today.

[momijizukamori]

Did the hotfixes for discord last night fix it? It all looks the same size to me.

[saga]

Thanks for the heads up, and thanks for all the hard work too!! 😘

[delight]

I am going to assume no, the way other messenger service listed usernames didn't link; they're provided for ease of access.

[momijizukamori]

Had a quick look, and it is in fact a bug - hopefully I can sort it out and get a fix in today.

[saga]

Thanks for the really fast update!! Hope it all goes well!

[momijizukamori]

ME TOO.

[syntheid]

Oh man this is just full of good news. New api backend! TravisCI! S2 editor update!

[momijizukamori]

That's one of those so close items - we're definitely keeping it on the radar, for sure!

[momijizukamori]

Exciting and useful places, I hope!

[momijizukamori]

It doesn't do a lot yet, but now that all the backend pieces are in place.... soon.

[momijizukamori]

Yeah, LJ did that a bunch of times. And then like, never properly documented the work for other people to build on, so it got left in the dust (see also: BML)

[momijizukamori]

I'm like, unreasonably proud of that one, maybe.

[momijizukamori]

Many shinies!

[momijizukamori]

They are, and hopefully the things people make with them end up making stuff better for everyone :)

Re: API and CI!!!!

[green_knight]

A short while ago, the client I was using - asLJ - started crashing the moment it connected to DW, which is probably related to API changes as nothing else on my computer changed. Since it's abandonware, I haz a sad.

Luckily, I have an app that would be almost perfect to convert into a DW client (and I'll be happy to do this), but I took a look at the API some years ago and took another look a couple of months ago and STILL ran screaming for the hills.

I'm hoping that we'll get not just the API but good documentation on how to use it.

Also, are we going to get an API for uploading images?

[roadrunnertwice]

Right, those LJ-descended technologies were all so wild! Every time I learn the story behind one of those it's like running into a platypus for the first time, and I'm just like, "that's bizarre, but given what you had to work with I'm impressed you managed to get any kind of mammal off the ground at all, like DAMN."

[sylvaine]

Hooray, thank you everyone! \o/

[momijizukamori]

Hmm. We haven't actually touched the old API at all, so nothing there should have broken. I wonder if it happened when we switched everything to https-only? I know that caused havoc with a few other clients/addons until people updated the paths.

One of my big goals with the new API (and the reason there's SO much backend stuff) is that I wanted it to be explicitly self-documenting - you actually write the spec for a path in OpenAPI format, and then the backend takes care of routing and API keys and validating inputs (as well as outputs, to guarantee that we're returning what we claim to return - which is incidentally why one of the two routes doesn't actually work right now lol). Right now the spec only gets outputted as JSON and only the spec and comments/screening paths work, but I have a bunch more paths ready to be tested once we're pretty sure the backend work is solid, and there's also a number of parsers that can take that spec file and generate easy-to-read doc pages. Because yeah, the XMLRPC API is a hot mess and the docs for it are even more of a hot mess. And yes, an API route for image uploading is on the list of plans :)

(this has been my baby for like, the last year and a half, can you tell?)

[momijizukamori]

Yes!