GitLab Pair Programming - Frontend, 31 Mar 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: FE Pairing - 20230331 - Detecting tests to run based on fixture changes

Description

In this paring session we work on configuring Jest to properly detect what tests to run when an HTML fixture changes in a pipeline.

00:00 Detecting tests to run based on fixture changes
33:00 Working on tests with cookie expirations

A

Having is so I'm looking at that Master broken it's from six months ago, so things have changed a lot in the in the I'm just investigating, essentially um playing around with jest trying to understand what caused that broken master. So VMR was green for this particular broken master and those three specs started failing for everybody else.

A

um The Mr is um this one here so introducing pipeline schedules, which is mostly front, end related I, would say, and the pipelines were green because we are running, we call them predictive tasks so essentially only a subset of this, not the entire Suite right, at least at the time. We did that um and my problem, or what I don't understand, is the spec that is failing. One of the specs that is failing is this one.

A

Ci variable lists and um I see this file on at least the source file is being a reference is being used by pipeline schedules, but we don't touch it in the Mr and the only relationship that I see between pipeline schedule and CI variables list is the fact that pipeline schedules uses CI variable lists, but I don't understand how changing pipeline schedules or introducing them would break um CI variables, lists and I've been stuck there for a.

B

While so don't use to go into that Mr, so this is all six months ago, yeah.

C

B

Okay, um so this introduced I, say pipeline scheduled or I, don't know if those are new components or yeah. Those are new components. Anna.

C

B

A feature flag, yes,.

A

Yeah, that's correct, oh and I must I must mention something else. I didn't understand this fix, but it was mentioned that it was easy blah blah. Where was it? I will find it here. Yeah we needed to stop stop feature Flags so that apparently fixed the problem, um but to be fair, don't understand what that does. Oh yeah yeah.

B

um Oh I think I may I think I might know. uh Can we go back to the alarm yeah? Of course.

B

So this is also.

B

um Sorry I'm gonna I'm, gonna, I'm, Gonna Keep about we're gonna bounce around I'm. Sorry, man.

A

Yeah yeah, that's that's fine! That's fine! That's right! Should we go.

B

To the test that failed, can we look at that and maybe even what it do? We need to look at it. What it looked like six months ago, are we yeah.

A

Oh yeah sure um so we could check the job and then yeah no wait well. I could check the MMR.

B

Now, like, okay, okay, I.

A

Mean it would be interesting, maybe, but what it looks like now is this one.

B

Hoppers fixtures load HTML fixture, ah so interesting line. 14.

B

yeah this one pipeline schedules edit HTML- oh oh yeah! So let's go back to that. Mr.

A

uh Which is yeah here.

B

Pipeline schedules: controller: if we go to the controllers project.

B

um Okay, yep and if we go to edit HTML Hamel done on up there, yeah okay yeah, so previously we rendered a form.

B

But now, if the feature is enabled, we don't render the farm anymore, we render those pipeline schedules form edit, but the CI variable spec was using this HTML file. The whole form it was it was trying to read from the form so like I would I think we would see that in the let's con Let's test that hypothesis, but can we look at what the failed messages were for that test?.

A

Yeah, of course, uh it's right here um so anyway, that's native form.

B

Yeah yeah: it's not even able to find certain things. It's trying to find yes, because the fixture changed under the hood, because we wrapped the previous Behavior we introduce new Behavior with a feature flag, but in test environment feature flags are always on and fixtures count as a test environment so generating these fixtures. From for the just test to read, uh we run that those fixture, R, spec files and feature flags are on by default, so we broke. One of the fixtures we've been absolutely insane, though, is if that test passed on the Mr.

B

That's the weirdest part. So did the test actually run on the Mr.

A

uh No, it did not that's the part I'm trying. That's a follow-up story at the time. What we did we, we just heavily relied on. How is it called that just flag called find related tests? We heavily relied on this one and it did not apparently detect anything based on the changes so based on those files that changed here. It did not detect um right right.

A

Yeah, though it didn't basically didn't, run those tasks and after that I think it was right after this that we decided to run the full pipelines like front-end pipelines, front-end tasks for every Mr, at least once um yes, yeah okay. So so that means that's very interesting. So pipeline schedules were actually it's. The other way around I didn't see that so pipeline schedules are actually included inside the CI variables inside the CI variables list. So CR variables list depend on those on Pipeline schedules.

B

The tests too, it's possible that none of the production code does it's. Just the production code needs is kind of taking some sort of like integration, test approach and or the test is kind of taking that approach. So it's like hey, let's just run our bit in the context of this pipeline schedules. Page um I have no idea. I haven't looked at the production code, but yeah, okay, but yeah. The find related test is interesting because that's only going.

B

So it's only going to look at what just in the front and know about, but these fixtures are a whole other variable that can be adjusted.

B

um I think that uh uh I have a feeling that.

B

When a flag is being introduced, we've opted to run all things yeah, and so that sounds like a pretty good.

B

um It sounds like a pretty good solution for this problem, but I could see it still being possible for maybe instead of a flag, but I've just touched that HTML file um and wrapped something that a unit test was expecting so that that thing's not rendering anymore, whether it was at the flag or whether it was some sort of application setting or whatever it was uh so it would be really cool to find related tests based on touched HTML files.

B

That's a really that's a very hard problem, I think and I'm nowhere eager to solve that one. Maybe we are.

A

Yeah, actually we're looking at um so what we did for our spec. We did a lot of things like related to that, a bit more advanced mappings like for instance. If you change a view, an HTML file will try to run the feature specs that we think are associated. It's mostly probabilistic. It's based on the name like on the path of the file, but, like we've done a few of those and we're looking now for jest, at the very least, to have a static list of files.

A

So if we know um you know like for the basic ones like you say, okay, if we change those files around here, we should run those texts around here, trying to be a bit more advanced, but um so far it's only relying on that um on that flag, which I had troubles understanding so I asked a lot of questions already got some pretty cool insights about it.

A

um I thought the dependency was reversed for some reason, I always thought that yeah, it's I, looked at it as a dependency tree, but the other way around I thought. If you include something it will run the tests for that. For the file you're, including, and it's the other way around um anyways.

B

A

Not great, like the documentation for this flag is not great, so yeah, you kind of have to figure it out.

B

Right right, yeah, I, wonder if.

B

Yeah the load HTML fixture thing is interesting. I wonder if um yeah I wonder if we'd be able to see a list of all of those kind of things they get pulled.

A

Because this one will um load area I need to look into this, what it actually does, because I have a very basic understanding of fixtures.

A

For me, it's just like, instead of relying on the actual implementation of the back end, we just generate the signatures so that the front end can just with mock data right and then the front end can just pull. That. Is that what it does yep.

B

You can actually look at um a line: 14 yeah. If you want to copy this string and and switch the project for it, you can see where it gets defined. I think oh, not not going to the file. Actually, okay, do a search for this, a project-wide search for the screen for the screen, just the inner part, okay, yeah.

A

Gotcha, sorry, no.

B

B

So they're in the front end fixtures. So all of these fixtures have a the the name of the test is what actually gets written is the file it gets written to, um okay, and so so these are all targeting the online 19 the controller.

B

So sometimes these fixtures are just API responses and it looks like sometimes it's actually whole views.

B

So this is how you can get the names of fixtures, so this is so here's here's an interesting way of looking at it, though you can get the names of fixtures by looking at all of the tests that get declared in our fixtures.

B

Based on the names of fixtures, you could build a whole dependency of what front-end tests are, depending on what fixtures, uh yeah yeah.

A

B

The big question is: what back in code touches what fixtures so I think all of it all of it can be answered. um I mean it's a pro. It's a beefy! It's a it's! A lot of work.

A

Yeah yeah I see I start to see that and uh oh, but you were saying something before with the stick, just though the feature Flags whenever so they're always enabled they're always done in the tests which yeah I guess makes sense to me, but then so that you know, because we want to fail early, we want to test it. Okay, you want to test it all the time and here so we do that.

A

Okay, we have that one here we're going to generate we're going to store it somewhere, most probably um okay, then this one runs because it was mentioned in the Master broken here. um Actually another issue. That was that we have for ourselves we mentioned, but I don't have data for that, so I couldn't find like. However, the find rateist is enabled to find any cells when front test pictures are loaded with a helper method like load HTML feature.

A

That might be true, but I just couldn't verify it or I did some basic rudimental testing and it was like, even if you had load HTML fixture, it was somehow working like at least a fine related test worked out, but um that sentenced a bit cryptic for me. I, don't understand where the author.

B

Came from find is currently being used around a selective, Jazz test.

B

B

Yeah, it's really interesting um is that the same broken Master issue. Yes, yes, it is okay, so I would be really so. There's there's other ways we load fixtures and um doing the load. Html fixture is probably a helper that adds more problems than helps.

B

How familiar do you have you seen the other way we load fixtures in the just test? No.

C

B

Haven't if we just do, if we go back to the vs code, we could do a project-wide search for um we could do a project-wide search. If you turn on the regex search yeah, we could do import dot star fixture.

C

B

There's one in mock there's a whole bunch in mock date that you see.

A

This case here, yep.

B

No all right, so we have a jest Alias for test fixtures to go to that temp front. End ee like to go to the temp directory that fixtures get written to so this is just like. We ran import to that exact Json file that the fixtures get written to. Okay.

B

So all that to say um we run, find related tests and it'd, be news to me and I would be really surprised if something that changed one of these fixtures, it was able to find the Jazz test using it, um because what I find interesting is what do we pass to find related tests? Do we just pass the Mr changes? Yes,.

A

uh Yeah and no way there's something else. There's the we haven't no, but that's not for chest, it might be we pass. Let me find it um because I've been working with this recently, we have something I think I need I, just need to Game On fixtures, because every time we talk about those I feel a bit off um in here yeah. It's actually this one predictive tasks. We are find changes, excellence, and here we are using fixture mapping, oh wow, and where do we pass those um the front end?

A

We essentially add so at the end of that class we yeah. The first thing we pass is the changes from the MR, the file that change in the Mr and we're. We have that mapping front ends fixture mapping, uh which is coming from the file which is I just need to find the file, so that I can find the format. But essentially we have that mapping key value um where we would refine changes. You hear a feature, so this environment variable here, that's what I was looking at a crystal ball: front-end fixtures, mapping.json, yeah,.

C

A

Just so it there is something I'm not actually sure whether we.

B

Pass this checked in this: this is something that's generated, or is it get checked in.

A

um I think we cache it it's. So it's not check I, don't think it's checked into the repo unless we cache it yeah um and this one I will need to see what this one actually does, because I don't know in detail what it does, but I'm not sure whether we use that for jest. We, whether it's only for our spec, but it's probably for just it- would make sense so.

B

Look at let's see where these pictures I think I see. Look at the spec support here on front end. Fixtures!

B

um Sorry, if you go to the fourth or fifth search result: oh.

A

Yeah front and fixtures.

B

So after spec support, after all the front and fixtures run if we have front and fixtures mapping path and we write to it there, you go yeah. All of the of that and that'd be really really interesting thing to see. So that's something that gets passed to find related tests.

A

um I need to check that I need to check that, but easiest way right now would be to go on Master, because I've been working on this quite a lot and I changed the structure quite a lot. So yeah.

A

Where is it detect? No sign changes, but that's the original one. Okay, fixture mapping, foreign.

A

Then so, then we have that file with changes.

A

um Oh, but I have a better idea. Let's see what find related test, what we actually use here so predictive there you go just CI productive coverage find related test. We pack our spec change, yeah. We passed it to find related tests because that files- that's the file I talked about before it can. It contains both the changes from the MR, the file names plus those mappings um wow. So so we are having the fixtures to find related tests. Yeah.

B

But at the very end of this line, you're going to see it's at the very end of this line, just keep going. Oh that's the sequencer um yeah. That makes sense. That's really interesting!

B

So if assuming that, that means, if something would it have changed a fixture, we know how to tell what just tests would requires that fixture I could see it still having a problem when we call them load HTML fixture method, um because the find related test that Jess does I think is only able to work off of imports and the load HTML fixture I. Think if you, if you could jump into the jump into that module helpers, is actually an alias for uh yeah helpers, fixtures, yeah um yeah there. It's actually I, think I.

B

Think down one more, not this one! It's okay.

B

Oh this one, you meant down one more sorry, yeah yeah.

A

Okay, yeah, oh yeah, of course, and this one is.

B

Git fixture is what is it doing it's joining to path and it's reading the file from the file system, so none of that import tree dependency stuff. That just depends on to find what's related, and none of that could see us when we use this thing really you're right yeah. So we really shouldn't use this thing.

A

It's actually one of the suggestions from the from the issue. I saw here, it was to say, encouraging static, import statements rather than helping methods or that test mapping, yeah I think now the person that wrote this was really yeah she's on PTO for two weeks, so I couldn't ask her, but she I think she she's way deeper in the problem than I am uh but yeah I see the problem, at least with HTML fixture.

B

Let's see if we can um how many of these, how many of these do we have it's, not that many right question. Oh, it says a hundred and oh, it's or 49.

A

55, sorry, it's all right um and and some are loading something called fixture: ah yeah. Okay, it's just different. Okay,.

B

um I think we could do you want to see if we can fix one of these while we're here. Maybe we.

A

B

To the CI variable this one sure sure.

A

Ci variable oh yeah, this one here.

B

Raise so the one thing I think we'll need to um one thing I think will need to figure out, is.

B

uh Because we don't, we need to tell Jess how we wanted to treat these files we want to import importing. Json is easy for just for node, because no just trees like a Json object. We want these to just import it raw, don't do any transformation to it or anything like that. I, don't know if we're already doing this with any other kind of just thing.

B

So can we go to our Jess config file? First.

B

um It'll be the base one yeah, uh so, let's scroll down maybe do file search for raw.

B

Oh, oh, there we go the last one.

C

B

I think that's.

B

I think that sounds sounds right. Can we look for.

B

Can we look for HTML here at all? We don't do. We have anything that does HTML.

B

um Oh, hey miracle.

C

B

Yeah, let's, let's just for right now, let's add HTML to the end of this uh at the end of this list here, yeah.

B

I wouldn't want to commit it. I wouldn't want it to merge to merge it like this, because I think I'd want to scope it only to fixtures that we're loading with HTML right now. This is saying any HTML import that Jess finds we're going to do the raw transformation which I'm not entirely sure if this is in sync with how we load webpack modules. So we unfortunately Jess and webpack load things in different differently, but we really want them to kind of load them the same. So again it's this config.

B

We have to keep an eye on um yeah, but let's see if we can import it like this and it raw transforms. Let's go to the CI variable list and see if we can change our load HTML fixture to just an import of test fixtures pipeline schedule, edit.

A

um So import I'm going to copy one that you said before um yeah.

B

Yeah yeah I think it's underscore fixtures.

A

More sense: yeah go on yeah, okay,.

A

um And I'm gonna call it. Oh, it doesn't matter right or could I.

C

A

Doesn't because we're not referencing okay, um statistic, Capital team, uh David testing, so that it doesn't uh from this one and two six shoes with fixture that they sent as well. Nice.

B

Yeah well so we want to do HTML. um Html is not only telling it where to look, but also we're using the raw Transformer. Yes,.

B

The uh we're not going to use load HTML fixture is there any other load HTML fixture in this file. Oh.

C

There's another one.

B

Oh, is it just those with variables.

A

And yeah I did with variables both.

B

Of them, okay, let's, let's get that one, then maybe we want to call it. Maybe we want to promote it. Okay, we can call it David.

A

Yeah, that's good all right, uh edit with variables there you go yeah.

B

Can we go to um that place where we did load HTML fixtures and that, because I think this one just called set HTML or something? Yes? Yes, yes, so, instead of load, HTML fixture, we'll just set HTML fixture to the content that we imported.

A

Okay set this channel fixture. So what should I change here? Sorry.

B

um Sorry I was just double checking that we had that function on line 28, I didn't know if that was actually a thing or not. So we do that's good. We can go back towards. We can go back to our spec file and instead of load HTML fixture, we'll import uh on line two.

B

Html fixture, oh.

A

Okay, let's set this channel fixture.

B

Yeah on it and then on line 15 will replace r or we'll add into the new line uh instead of load HTML fixture we'll do set HTML fixer for data test. Okay,.

A

Oh yeah, okay,.

A

And just like that, okay, yeah yeah I'm just going to do the other ones. uh Do this thing too.

A

Good, sir, that should be it yeah.

B

Yeah so let's run, let's run this test and see see how it goes.

C

A

You'll have to tell me if I run, Jeff Bex correctly, it's what I'm doing is I'm doing yeah that guy that's without.

A

Like this right, yeah.

B

I used I would just do I just run yarn tests and pass the path to the test. You don't have to deal right.

B

A

Awesome: okay,.

B

So I think I think what we have stumbled upon is a is an epic that we can create till move everything away from using load HTML fixture, and we want to ultimately remove load HTML fixture so yeah. That makes a lot of sense. That's something we can. We can crowdsource to contributors in front end and stuff. We just highlight all the offenses, and maybe we can uh add something that we don't.

B

We don't use it anymore and here's what we do instead. So.

A

That'd be super cool yeah, because then then he would definitely unlock Ghost testing yeah. It's it's really really interesting. Thanks for the for me, that was a really deep dive. So thank you so much for that. Anyways.

B

The reason one more thing just to highlight is- uh and you may or unlikely are already aware of it, but the I'd be I I, don't believe jest can follow dependency trees unless the Imports are statically at the top, because you like Dynamic importing and stuff I, don't think it's going to be able to follow those and that's why we wouldn't just be able to replace the implementation of load HTML fixture, because we we need these Imports to be static at the top and I know that Jess handles that specially and node as well.

B

Like that's a general Bank there. So yeah.

A

Cool yeah that makes the area that that makes sense. It opens the yeah I'll check this one and I think with having those mappings. If we were to engineer those mappings, we probably do something even without the place I mean replacing, it would be great, but imagining assuming like there's another kind of dynamic loading, I think we could in in Ruby or whatever we could just parse those you know, I could check for all the files having load HTML fixture I take. What's afterwards and I just run the specs Associated.

A

To this one I mean we could probably it's hacky yeah more hacky than than importing, but it's yeah. Something can be done actually.

B

Right right, I think I I agree with what you're saying I. Think, though, like since we don't well since we're not abusing it that much, and we really should avoid like file system reads, and you know, tests and like let's just avoid it. All together is like.

C

A

B

You so do you want to create a do. You want to create, you can even just stub the content of it and ping me on it. We can collaborate and fill out the list of how we the Epic for it. If you want to take a whack at creating an epic to clean these things out and then of.

A

B

Can collaborate on the description of, however, we want to do it, so it.

A

Sounds fantastic yeah, I'll, just um yeah, exactly I'm gonna, stop it I'm, just gonna put the things and then you can come and oh.

B

Yeah I think that sounds great uh and then, if you want to create a Mr of this one, that would be fun. You can just create a VMR. The changes we made, except for one thing, I, think we should update, which is the Jess config yeah um this one. You mentioned yeah right. Instead of um transforming all HTML I'm inclined to want to only transform test fixture HTML, and so that would just mean below this.

B

We would add a regex that starts with test fixture. It ends with DOT HTML, and that is the one that goes to Rodger's one. So yeah.

C

We could do that here.

A

Yeah yeah not just to do yeah.

A

That's cool, yeah, yeah, I'm, afraid I'll, try something here that one so Mr and epic sounds great. Yeah awesome.

B

Yeah yeah feel free to ping me on it. I'm I'm, happy to I've we've put together a number of these like Mrs. A number of these epics with issues targeted at fix, fix this file. Here's the instructions for how to do it and that's always super helpful, so nice yeah thank.

A

You so much for really really interesting cool.

B

Now, I'm glad this is an interesting problem. Thanks for bringing it up, uh yeah uh hey did you have Miracle? Did you have something that you were hoping to work on.

D

I am working on something it's just, not anything friends so.

B

Is it cookies of.

D

B

Cool, let's do it yeah and that's probably great because David here with a strong back end experience much he'll be able to help out way more than than I would uh I.

D

Actually, finally figured out um the issue that I had brought up um and what it was that Jesse met when she was talking about. Is this the right? This is the right area. um Sorry I have so many vs code windows, open I, wasn't sure this was the right one.

D

um uh What Jesse meant when she was talking about that active session Ruben file? Oh.

C

D

Well, my cookie now works. The only thing that I'm working on is my final test um and I haven't really wrote that many R-Spec requests tests, and so this is not the right spice.

D

um Yeah this is not. This is not right. Sorry give me one second nope.

B

You're fine, but that's great so now this is so now's just testing it. Yes,.

D

I'm. Writing my last test and I. Don't know where everything went.

D

I thought it was here.

D

Okay, you know what let me stop sharing and reopening and reopen stuff um but yeah. uh My issue is that I, don't like my last test: I need to uh expire a session and make sure my cookie has been deleted with the session expired, but I'm not quite sure how to do that in a request. Spec.

D

Just because it's more of a controller spec thing.

C

D

um I've tried a few different ways.

B

That's a really good question.

B

Yeah, that's a really good question.

D

Oh, that explains why I couldn't find anything I'm on the wrong branch.

B

D

uh But yeah so.

B

Does it is it uh foreign? Are you trying to fix? Are you? Is this a new test, or is this we're trying to fix a test that got broken.

D

It's my very first student Hall, oh.

B

That's no, no you're totally fine. If you want to uh share your whole screen too, that's that works as well, but this is this is a nice font too. This is a good font size.

D

Cool, so these are the tests that I've done so far. um Basically, when the user signs in the cookies there when they signed out the.

C

D

um And then, when they hit, remember me the cookie. Is there the one that I'm struggling with? Is this one.

B

D

um I've tried deleting the session cookie I've tried using a few methods like reset session um session Inspire, um and then I also try um setting up a helper method and then calling it here and for whatever reason, I think it's. The session expired. My cookies, don't delete and I know that this works locally, because I've been sitting here watching it happen for a few hours. Oh.

B

Okay, so one one idea is: is it possible for us to.

B

Rather than waiting for this to expire, could we test that the response cookie about gitlab active user was created with some sort of expiration?

B

Does that make sense.

D

Yes, but also um I think that's done by default, so when um so, because I didn't actually set an expired uh key.

A

C

D

The session locally, it expires with session.

B

I see I see oh interesting.

A

uh Could you please go back to the session spec quickly, I? So how do you? How do you actually so first of all, is your experience in that domain? I just did a lot of our specs so happy to help um here. How do you see me so then the session expires so that your contacts here, it's implicitly done right. So it's as you said it's implicitly when you do that post on user session path.

A

We assume that the selection, the question, is expired, or is it when you actually delete the cookie in the in the test? I'm, not sure.

D

This is just a sign in the user, so I actually have a session running and then I wanted to expire the session and then make sure the cookie was deleted, and so this line is the one where I was just throwing caution to the win. This is actually like. It's set when you sign in and I thought, I just delete this cookie, then it'll expire session and my cookie will no longer exist. Okay,.

A

Yeah I see I, see so: okay yeah, okay, so you drop the session and then you expect the other cookie to be done. Okay,.

B

So does the cookies all right.

B

Can you explain what's going on in the active session thing because yeah that seems really interesting and.

D

Oh yeah, because.

B

We were like there was a lot, a lot of cool things to learn there. You you glanced over I was like oh man. This seems so much fancier than what we were doing. Oh.

D

Yeah, so all I did was create um okay, there just two methods that set an unset at cookie cool.

C

D

Then um what I finally figured out is that device used important to start sessions.

D

So there's two methods in here called after authentication, uh so rather three methods after identification um after set user and then before log out so after authentication, is where the remembering variable lives, and so, if I have my cookie set here, whenever someone clicks that remember me check, then my cookie will be set. It also expires with. The remember me variable.

D

um The only thing I couldn't quite figure out is how to uh directly pull the expiration time from the remember me Cookie into my cooking, because if I did it here, it blew up the pipeline. So it expected me to test something here and I was expecting me to test the method here and I didn't have a method and then, if I hold the the um the sorry, if I pull the inspiration from that cookie here, they just didn't line up at the same time, because this was running before this.

D

So this was always later than remember me cookie, but because I have this all set up in this method. Where remember me is set, even though my cookies, expiration technically says session, it's follows the remember me, cookies, expiration.

C

D

um And then this yes right.

C

D

Then this is for a regular garden, and so after you sign in it sets your user. um So if my user just hits the signing button, then my cookie is set, and then this one delete everything when the session ends or when you log out.

B

Got it got it? Okay, okay,.

B

I see and I think that may be the the the the difference um or I think I think my my brains hooked on something.

B

D

B

Deletes that what we're gonna say.

D

Oh I was thinking like I could call this method inside of the active session, destroy method and maybe that'll solve all my problems.

B

By calling it from the spec.

D

um No so I have this active session or active session, destroy session method and I could call my unset active user cookie inside of that destroy session method. Maybe that solves all of my problems.

B

That'd be great yeah.

C

um Yeah, sorry, no.

B

No, no, no, that and.

B

I think I, don't I'm, not sure what the problem that that might solve, but I can see how that'd be really helpful. The problem that I'm I'm I'm thinking of right now is yeah. How do we test the case of this new cookie?

B

You know last as long as the session when it expires with request specs, because that's really interesting because and looking at Warden um I think we would have to trigger another request, but the time would have to be after the expiration, uh and so you could do like time traveling in our specs uh and so then we can trigger another request and the request then like that's when it would delete stuff.

B

And that might work, um but then the other, my other perspective of it, is like, if things get too hard to test.

B

Have to ask yourself I, guess: you're always forced to ask yourself like how how much benefit is this test really doing.

C

B

I say this: I say this, because if you go back to active session and you've talked about setting these cookies and you didn't have to give it an expiration- that's kind of already already handled for you. That means that there's not really a line of code where a bug could be introduced and that's really like the tests we can get. We can get from the test perspective. You can get like full use case coverage, but it really means like you're gonna, probably be covering multiple lines multiple times.

B

The other way of put thinking about it is like is, is all of the code sufficiently value covered and because, if we don't have to test the framework itself, um so if something is just too tough to test, it's I wouldn't be afraid to just tap out and ask a maintainer or someone and be like I can't figure this out, but but all the other lines are covered. So we should be good okay, um but that's that's totally. Your call, but I do think it'd be fun to try to figure it out uh yeah.

B

But what was the problem that you're thinking of with uh adding a call to unset with the destroy session.

D

Oh as I was explaining to you what was going on in the sport file. I realized that before log out only really handles the log out, it doesn't actually destroy the session. This is what destroys the session and I did not call my unset active music at all. Inside of this one.

B

D

See I did that when I intentionally destroyed this um it'll be called explicitly incident implicitly.

B

um Yeah, so when you hit the request, specs does it hit these like? Is it does these initialize I have no idea, do these initializer modules get loaded for the request, spec.

D

Yes, if you do it.

D

In a weird lines, I, um let me see here so like if I use this method, login user, it doesn't actually always doesn't actually meet my initializers but I use session pass or users actually pass. Then it does. Oh.

B

C

B

Oh and there's a whole okay, all right! Okay, can you go back? Let's go down to the your tests again that you added I'm one of curious about something. So we already have where the user signs out. Is that one working where we post the user session path? Oh then, we post the destroy user. Yes, Scotty gotta got it got it cool cool all right, but now we're here just saying: okay, have we just completely expired, yeah? Okay?

B

Could we um and I think the way to test this would be to time travel? uh Do you know how to I? Don't remember how to time travel in rspec.

D

I know how to do it for controller specs and not request aspects.

B

I think it's the same.

D

Okay, um then I did it in.

A

um Travel too much probably um would be the one I'm, not sure. If we don't use time Cup right, we use a rails. It was probably for that, but I think it's traveled to travel from yeah. One of those.

B

C

B

Like next year, yeah.

C

D

uh Yeah, it only lasts two weeks all.

B

Right, let's, let's I'm a silver compensate and travel the next year. Okay, just to be sure is it. Travel too is that is that what we.

A

Yeah travel two and block.

B

Oh, it is travel too cool.

B

um Okay, you could do travel too and then parentheses, oh man. Why is why is Ruby, so you could do one I know from now Dot from now. What kind of language is this.

D

And these 2B nails like, why is it that b, nil has an underscore and 2B does not I.

A

Mean yeah, it's really excellent. Oh sorry, I cut you off go.

D

Ahead, oh, no, it's okay, I mean like technically I could add it, but.

A

So the the answer is actually amazing. I find it amazing, because B underscore will match a method. So Neil is is going to be the nil question. Mark method you could be, you could do be false, be true, B empty, be any method your object has. It will be really the underscore yeah. For instance, if you have an empty method, Define you can say BMT or it's just to stand more in English yeah, oh okay,.

C

I find it super nice.

B

That is really interesting. I had no idea.

B

Okay, travel to I'm, looking at some examples, and it looks like oh yeah, it looks like we can just do it like that. It looks like the problem now is actually has an underscore. So it's from underscore now.

A

Yeah and you need block afterwards so do an end just like you're testing, okay,.

B

Do we need we need to do a block.

A

um No, what was wealth I think ah wait, wait just calling it like this and then but then you would have to reset it right. Then you would have to say apparently travel back and I. Think if.

B

A

Do the block I say That's Just Gonna Roll it back afterwards. Yeah I see.

B

It's probably easier so then, then I think we just want to post to anywhere like I, don't think like, maybe rather than or just get anywhere I, don't think we need to post, but just some sort of.

C

B

I mean maybe get user session path. Yeah that'd be interesting and let's see.

B

Have you already tried something like this.

D

um Not exactly exactly I try to just um just setting the session um expert or the session date to be like in three weeks and then the expiration should have updated. Good I haven't actually used like the travel 2 method.

D

um So when you were saying you have something else,.

A

Do you happen to have buy bug installed the gem by the way? Okay, because then I think I mean I found that when I cut them out three times every minute, I would probably do a buy bug right after they get user session, so that you can inspect the response.

A

Okay, but it's just so we're gonna, say yeah, it should be nil. It's not yeah! It's up to you and I think also a note on this on debugging on buyback I'm wondering if it would even wouldn't even work in the initializer in your word in that RB. If you wanted I think you could also put because I suspect that the initializer is run when you boot the app yes.

D

A

I think the block is lazily loaded right, so it should be. The method should be called when your spec is called and I suspect. If you put a buyback there, it's gonna block and you can inspect also or what you were, what you were curious about before in order like. Is it actually going there or is it? You know resetting or not yeah yeah, just to sign out.

B

hmm I'm gonna I'm gonna search for any other kind of test where we do um we're testing for expiration stuff, I'm gonna see what we do.

D

D

Duration for, like the reddest store and everything, but not exactly like our session store, yeah.

D

Okay, what returning true in my response.

C

B

Is that the only cookie is the about gitlab active user? It.

D

Looks like it because I yeah I printed the whole response.

B

B

Can we go ahead and and exit and continue this one and enter just to double check that we're actually changing something by traveling? Can we travel to zero years from now and and it's traveling even my question- is: is traveling even changing anything at all like we can comment it out if we want to or travel to one second from now or something like that.

A

Or call uh time dot now and then you should see 2024.

B

A

I was actually sorry.

B

I was actually saying: can we try running it again without the traveling and and seeing what the cookies returned are? uh Yeah, okay,.

B

B

Time is time is weird. You know, uh I think it's it's fascinating. How people can can do things that cause them to perceive time slower and faster than they normally do, which kind of hints at this whole thing of time, kind of being an illusion.

B

But it's not it's! So not yet that's strange uh yeah, that's it um so yeah! That's the only cookie. Getting back. Is this about gitlab active user? Yes,.

C

D

Maybe maybe it explicitly call this um where's destroy session.

C

B

Well, right now, we're not um in our tests we're not just we're not uh destroying any sessions like nothing, no right! Nothing no session should be destroyed in our tests. Yet.

D

Oh yeah, but if that's the case, then I should at least have like the known sign it. Where am I the known signing cookie and like the git lab session cookie.

B

Right, that's. My big question is yes? Yes, where are other cookies.

B

Oh there's the other cookies.

B

But request stock cookies is not oh wait. We did response.cookies response.cookies is not only has that one.

B

um So in the buy bug, if you do response dot cookies, that's that only gives us that one cookie we don't get all of them.

A

Which raises the question: what is the difference between cookies and requested cookies yeah cool? There is a stack Overflow about this. um They should be the same. They say, however, in post railsbury finds the X as RF token.

A

B

Maybe it's because um some of these start with the underscore maybe like that causes them to be hidden when we try to do just regular cookies.

B

um I find it interesting too, that that cookie is being sent back.

D

B

Muted, myself, no you're fine.

B

Were you gonna say something.

D

D

I find it weird that it's only in response that hook using that in any other type of cookie jar.

B

C

D

If I, um let's see ER well again,.

D

I guess this would Now fail.

B

Do you um do you want to try? uh Oh, what was I thinking.

B

Yeah I don't know, have you have you? Do we wanna look into feature specs at all? Do you think that environment might be a little more helpful because it's.

B

It's more fully integrated with a browser and all the stuff there.

C

B

Don't know if you've already thought about it or if you've you've tried that out already, but that might be easier than request specs for something like this.

C

B

I also check just one one random thing: can we do a search across the project for all references to about gitlab, active user.

B

Okay, yeah, it's just in that active session. One okay, yeah yeah,.

C

B

Yeah, that's interesting.

D

um Yeah I guess last dish. Effort was.

C

D

D

um What did they call this.

D

I also should be.

D

C

B

Is it a um if it's in the same model doesn't need the I, don't I? Don't this I have no idea about Ruby? Does it need the active session name, or can you just call it without the accession name, or do you have to do self or something I have no idea.

D

C

D

Shoot what was I doing was it session expired.

B

um With the time traveling or with the redoing.

D

What you've done before I understand what I was doing before, but I.

B

Could try to find a problem.

A

B

A

Just dropping the quick hearing.

C

D

B

Yeah I I think before we were deleting the response. Look. We we called like delete on the response. Cookies.

C

um Oh, that's right.

D

Yeah, that was my last Stitch effort yesterday,.

C

D

Why how did I feel the first test? That's on the test, I wrote.

B

Off cookies, oh, it looks like.

D

B

D

I forgot to pass it so I already met, um oh I mean off here and I. Think we have. We have something weird going on with this method, where I can't change Beyond arguments given without blowing something up, and it's only called it's only creative here and cold and Warden.

D

Well, if I could spell all.

B

Right I think the I think the error uh I think they are. If you go back to the act of Session One.

B

I think uh I think it wasn't about this method. I think it was about on line 112.

D

A

um Probably like 112 is calling this one, because you see right under the in the stack Trace right under it says, line 147.

A

and which was the line you were at before yeah.

D

B

How that needs I see? Oh that's, interesting and I see how it needs. uh Okay,.

A

But it should be reversed, like the snack Trace is misleading.

B

um That is misleading yeah.

A

Yeah I think it's because yeah, when you have those before blocks and stuff like that, it would just.

C

A

um I got used to it, but it's it's! It's completely. Unintuitive yeah I see what you're saying as you're saying: okay,.

B

I had no idea what the stack Chase was saying that was so weird. It was oh, okay, all right.

B

Here we go, this: is it uh yeah this? Is it.

A

B

I yeah I, don't know.

B

I think it might be really really hard to figure this out uh and I'm inclined to you know what you just added of like. Oh yeah, we're hooking into the destroy session method, which is the single source of Truth, for whenever we need to destroy a session. We're gonna get rid of this thing like oh, that sounds like yeah. You can just cover that with the unit tests. We don't necessarily need a more integrated test for that.

B

But if we wanted to do a more integrated tests, I could see the benefit to it, but it sounds I'm not sure how to trigger that would be, and I couldn't find any other kind of examples that were triggering some sort of expired session thing either.

D

Yeah and then they're ending loop with this specific problem.

B

Yeah I I think the fact that we may not have a whole lot of other specs dealing with testing at this level might indicate that we're kind of okay, just leaning into the framework and assuming that the framework does its thing.

B

But I don't know yeah.

B

C

B

Seem seem great, though,.

D

Cool then yeah I'll, just ping, my maintainer yeah.

B

Man, these cookie problems are hard.

D

B

I'll be glad when it's over yeah well, thanks for bringing it up, that's really interesting appreciate it. Miracle yeah thanks for thanks for working at it and I as you keep bringing it up, I'm, I'm learning! So much about and I really appreciate, seeing this latest iteration of like.

A

B

That what the active session thank us for and like all of the things I had no idea meant anything. So that was really really helpful and educational.

D

Actually BT couldn't touch, which is where I solved the problem and I was like okay, you know what maybe the original search is always better.

B

um Did you get to meet with um uh Julie Jamie Jessie.

D

B

So bad about that, um did you get some? Did you guys get to record the that pairing session, or did that not did not work out to record it.

D

I think that was.

D

That was an engineer and a lot of the things they talked about, went right over my head and they were talking about the specific problems.

B

Wow wow: that's really interesting! All right cool all right! Well, thanks! Everybody! Yeah I, didn't realize sorry so late. So yeah, thanks for hopping on thanks for pairing, appreciate bringing up the interesting issues, all right, I'll catch you all later, bye.