VBA Rubberducking

Duga

12:00 AM

RELOAD!

[rubberduck-vba/Rubberduck] 1 opened issue. 2 closed issues. 12 issue comments.

Duga

12:14 AM

[rubberduck-vba/Rubberduck] build for commit 333c89ac on unknown branch: AppVeyor build succeeded

[rubberduck-vba/Rubberduck] comintern synchronized pull request #4727: Code Explorer hotfix 3

Comintern

@this I just pushed my current local. I think I only have 3 more test sets to add - structure tests for added and removed components, and tests for the declaration sync.

...which will push us over 5K unit tests. Yay arbitrary thresholds!

IvenBach

12:35 AM

:Cheer:

Duga

12:48 AM

[rubberduck-vba/Rubberduck] build for commit 2fb73ef0 on unknown branch: AppVeyor build succeeded

this

1:41 AM

hmm. weird. I thought we already passed that.

But yes, yay for arbitrary threshold!

Comintern

I might just be VS messing with the count for different test groupings.

this

yeah could be

Comintern

#4358 should probably be addressed at some point. Rendering two identical tree views on every parse is a killer for huge projects.

If we are going to have 2 trees on the same project, they should share a common set of nodes - something along the lines of a ProjectTreeNodeProvider that hands them both a view on the same underlying collection.

this

Just to be clear - why did we add ref to a bunch of members?

I was going to ask about that. If they all come from the same list, they definitely should be one big list (!) projected accordingly.

Comintern

It reduces the computational complexity.

this

1:51 AM

by not materializing the lists every time it gets passed around?

not sure I follow.

Comintern

Dramatically. It allows the list to be rebuilt with .Except() instead of removing members one by one.

this

ah ok

Comintern

There's no way to do a RemoveRange<T>(IEnumerable<T>)

Although in hindsight, I probably could have cooked up my own extension method.

this

perhaps but best to use native methods anyway

Duga

2:24 AM

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.VBEEditor/Utility/ISelectionService.cs of pull request #4733: Introduce the ISelection service

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.VBEEditor/Utility/SelectionService.cs of pull request #4733: Introduce the ISelection service

Comintern

2:35 AM

@Duga So what do you set the out parameter to if it's a TryGet*? default?

this

hmm. blargh. It's a value type.

Duga

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.VBEEditor/Utility/ISelectionService.cs of pull request #4733: Introduce the ISelection service

Comintern

I'm kind of up in the air on it. The new out var syntax makes it easier to bypass the return check.

this

With the design as is, you still have to check for the null.

using default on a value type would be confusing.

so if you're going to the trouble of using a nullable....

Comintern

...at least it throws an NRE early instead of who knows what down the line with a bad value?

this

2:45 AM

IDK. I'd rather get a NRE than a weird behavior

Comintern

^

And the reference folder test finds its first facepalm.

public override string Name => ReferenceKind == ReferenceKind.TypeLibrary
    ? Resources.CodeExplorer.CodeExplorerUI.CodeExplorer_ProjectReferences
    : Resources.CodeExplorer.CodeExplorerUI.CodeExplorer_LibraryReferences;

"Hi, I'm Comintern, and I've used a ternary before."

this

I figured you were going for an early April's Fool joke.

Comintern

I've probably seen that on a daily basis for the last couple months. I think it's been that way since the references PR.

Tweeting Duck

4:26 AM

And sometimes, just sometimes, amazing humans walk into your little project and share the most amazing bits of wisdom you’ll not believe.

Tweeted by martinwoodward on January 21, 2019 at 6:56 PM in reply to martinwoodward

Duga

5:41 AM

[rubberduck-vba/Rubberduck] comintern added label support to issue #4725: Indenter does not indent preprocessor commands within procedures

[rubberduck-vba/Rubberduck] comintern added label status-norepro to issue #4725: Indenter does not indent preprocessor commands within procedures

[rubberduck-vba/Rubberduck] comintern created comment on issue #4725: Indenter does not indent preprocessor commands within procedures

> Closing as no-repro. This can be re-opened if additional information is provided that allows it to be reproduced.

[rubberduck-vba/Rubberduck] comintern closed issue #4725: Indenter does not indent preprocessor commands within procedures

Duga

5:59 AM

[rubberduck-vba/Rubberduck] comintern synchronized pull request #4727: Code Explorer hotfix 3

Duga

6:19 AM

[rubberduck-vba/Rubberduck] build for commit 320ee737 on unknown branch: AppVeyor build succeeded

Duga

7:39 AM

[rubberduck-vba/Rubberduck] MDoerner commented on Rubberduck.VBEEditor/Utility/SelectionService.cs of pull request #4733: Introduce the ISelection service

Duga

11:10 AM

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.VBEEditor/Utility/SelectionService.cs of pull request #4733: Introduce the ISelection service

Duga

12:48 PM

[rubberduck-vba/Rubberduck] MDoerner commented on Rubberduck.VBEEditor/Utility/SelectionService.cs of pull request #4733: Introduce the ISelection service

Duga

1:56 PM

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.VBEEditor/Utility/SelectionService.cs of pull request #4733: Introduce the ISelection service

Comintern

2:08 PM

Well there's the problem...

github.com/rubberduck-vba/Rubberduck/blob/next/…

this

ew. Shouldn't be touching COM there

It looks like it's outdated, too - it should be using the same changes that CE is using to retrieve the document name.

Maybe extract that into an extension method? a service?

Comintern

That gets run for every identifier reference for a worksheet.

It should run at most once per.

I was considering making a DocumentDeclaration subclass and making it a lazily evaluated property.

this

Hmm, IDK. Wouldn't it make more sense to evaluate it all once during the parsing run?

That way, we touch COM only once and in that one instant.

Comintern

My thought was it would probably be safer to defer it.

I don't think we want to hit that with multiple threads.

The highlighted line above accounted for 21 seconds in one inspection run BTW.

this

2:23 PM

Yeah, definitely not. But doesn't inspections run on multiple threads anyway?

Comintern

Yeah, that's true.

this

and I thought I understood that the parser is set up to do COM stuff on one thread?

Comintern

I'm pretty sure we don't.

this

OK

Comintern

I need to check to make sure the parser or resolver isn't already enumerating that collection.

this

2:29 PM

I guess for normal typelib consumption it won't matter. For those that touches VBA's COM, well, hazmat suit required.

Comintern

It if is, we should just grab all the results we need at a single point.

this

Yeah that's why I was thinking that might be better since I thought we have to do that anyway.

Comintern

Maybe copy it out to a Dictionary<string, object> instead of just grabbing one property.

this

Either way, I do agree we should have a DocumentModuleDeclaration

I don't know - it's not always safe to read values from the properties at all times.

Comintern

It has to inherit from ClassModuleDeclaration. That's about the only special consideration.

this

2:31 PM

It's usually safe to enumerate the properties' names, but not necessarily values.

Comintern

Right. That Access thing.

this

I guess Microsoft like to put little firecrackers in every nooks and corners.

Must have been a bunch of practical jokers back then.

Hosch250

More like, must've had deadlines.

this

You'll subscribe to your boring old reality. I'll subscribe to my much more interesting and fantastic reality.

Comintern

Do we want a naive easy fix in there for 2.4?

It would be simple to just cache the result in the inspection as a short term fix.

this

2:49 PM

at this point, only bugfixes should go into 2.4

as long an issue is opened and a TODO marker left, I'd be fine with that.

Comintern

That's why I mentioned it. I consider that a "bug".

this

yeah, the whole thing about COM-touching is a bigger but ultimately a separate issue.

Feeds

Hosch250 has made a change to the feeds posted into this room

this

At least you could just copy-pasta the changes that CE has for getting that information (primarily for safety)

Feeds

Hosch250 has made a change to the feeds posted into this room

Comintern

2:53 PM

Nah, CE only hits it once per declaration.

The inspection hits it N times where N is the number of identifier references for worksheets.

this

Right but the code the inspection has now is unsafe

e.g. it should not be hitting the VBComponent.Properties

Comintern

I don't know if I'd go as far as "unsafe". Unwise maybe.

this

I would because it will leak pointer, which causes zombifying on the exit

FreeMan

@this The cake is a lie! Seriously? Nobody else went there?

This is an interesting bit of chaos:

0

Q: Office 2019 VBA IDE

After installing office 2019 I encounter in some very strange VBA IDE problems (32 bit) (I check it in EXCEL, WORD and ACCESS VBA) Ctrl F - Find dialog, when hitting the "Find Next" the dialog closed without finding anything (even a single character). sometime shows "Out of memory" message -...

Hosch250

@Vogel612 I've heard of Cake. It's modeled after CMake.

I've never needed it, but it sounds pretty good.

Comintern

3:06 PM

Mmmm... cake.

this

Let them eat cake.

FreeMan

@Comintern I prefer π

Hosch250

If we are covering sweets, kisses are some of my favorites.

Comintern

Given the choice, I'd take the π too. Just not the squared ones.

Hosch250

Talking about that, my sister is a piano teacher, and she gave her students some chocolate kisses, depending on how much they practiced.

So, one of her kids comes out of the lesson and tells his dad "Xxx gave me 5 kisses!" The dad thinks she's being a pedophile.

Comintern

3:09 PM

lol

Reminds me of this:

Little Girl Confuses Weed and Weeds

►

Hosch250

Today is NOT a good day so far.

Tired, and already ready to snap about the bugs coming in.

Oh yeah, who thought it was a good idea to give us 1 week testing to shift our site to UTC (other than the devs testing as they make the changes).

this

Well, Tuesday is the day you have to pay your dues.

Hosch250

Hmm?

Comintern

I thought that was Duesday.

Hosch250

What dues?

this

3:18 PM

#Pun Failure

Hosch250

Oh, thought it was something about year-end stuff.

Like taxes, and stuff.

this

well, it is almost the time for The Duesday.

Hosch250

Yeah, not until April.

I'm waiting for a 1099, or something, from Robert Half, from a recommendation bonus.

this

Probably a bit later than that, due to the current terrorist attacks.

Hosch250

Otherwise, I'm ready.

@this You mean the shutdown? No, taxes are due anyway.

And the IRS is working and going to be giving refunds as normal.

this

3:20 PM

Hmm, I read that they may be backlogged because of that.

Comintern

We get our tax rush in January. Our customers are generating the 1099s as opposed to consuming them.

Hosch250

The Ds wanted them to, but Trump issued an executive order that the courts upheld based on a previous president who did the same.

And the court upheld it then, so this time they just said BTDT.

this

Don't you love it when the terrorists argue over how to spend other people's money and using human shields at it?

Mathieu Guindon

I just hope things will be resolved by mid-March :-)

Hosch250

That's the MVP meeting?

this

3:25 PM

yeah

Mathieu Guindon

^

Hosch250

Did you get your MVP back, @this?

this

otherwise the flight is going to be fun.

yeah, back in July

Hosch250

Nice.

this

the renewal cycles used to be on quarterly basis but recently they now do it on July for all MVPs.

(but even back then my cycle was on July)

Hosch250

3:27 PM

Had you lost it for a year, or something?

this

nope. been at the mast for 10 years straight.

Comintern

Hosch250

Nice. You just didn't go the meeting last year?

Comintern

Well that's new.

this

you didn't notice it until now? It has been up there for awhile

Correct, Hosch - that was my 2nd meeting I missed, though. :(

Sucked but life don't always consent to oblige.

Hosch250

3:28 PM

Well, go this year!

Say hi to Mat for me :P

this

yep

just have to book the flight.

KySoto

well... didnt realize i closed this tab

oops

Comintern

You should get in the TSA line now.

this

or just... not fly.

I bet you you'd arrive sooner by bus than by plane.

Hosch250

@KySoto Hi. Get your marriage license?

this

3:31 PM

or just go to Vancouver, then hop on the bus.

Hosch250

Congrats, BTW. I wasn't in chat yesterday, so I missed it.

@this Haha. I was looking at bus routes just to see how "easy" it would be to use public transportation to get to work.

this

Is it?

Hosch250

Turns out it would take me like 6 hours to cross the cities and get here.

this

oooo that's brutual.

Hosch250

I have to cross 33 miles through 2 cities, crossing through several sub-cities to get here.

this

3:33 PM

shouldn't they have express routes?

Hosch250

In the city, yes.

this

so not between the cities?

Hosch250

I don't go through the city, I skirt the city.

I live in the eastern St Paul suburbs and work in the western Minneapolis suburbs.

this

it's just that one'd think that for two city so near each other, there'd be express routes going directly from one hub to the other hub.

Hosch250

The freeway system here splits. So we have 35 E/W on the east/west side of the cities.

Then 94 goes through the city, and 694/494/394 kind of wrap around the cities.

I follow 694 to 494 and work in Plymouth.

Duga

3:36 PM

[rubberduck-vba/Rubberduck] comintern opened issue #4734: SheetAccessedUsingStringInspection takes way too long.

> My stress project for Excel has somewhere around 100 worksheets, and I noticed that the `SheetAccessedUsingStringInspection` takes around 25 seconds to complete. So... I attached a profiler to see where the bottleneck was and found this:

![pqlau](https://user-images.githubusercontent.com/8944005/51545570-deadca00-1e27-11e9-8b24-08fe8d89c4c9.png)

That points to [this line of code](https://github.com/rubberduck-vba/Rubberduck/blob/next/Rubberduck.CodeAnalysis/Inspections/Concrete/SheetAcce

Hosch250

@this There are, but I don't go near the hubs.

this

I see

Hosch250

They have express lanes on 94, but not 694 or the part of 494 I use.

Not sure about the rest of 494, haven't done the whole loop.

Mathieu Guindon

@Duga ouch

Hosch250

They have express lanes on 35 E/W too.

this

3:39 PM

down in SA, it's basically one big perpetual construction project

Hosch250

Ouch :(

this

they're building the cloverleaf bridges for the 281 and 1604 which really hurts the commute for those in the northern edge. They even say it'll be 3 more years before it's done.

Hosch250

I think they might redo the worst part of my drive in a few years. They redid large stretches of the freeways around it.

M.Doerner

Regarding the release, I think until 2.4 has been released only bug fixes for the CE and translations should be merged.

Hosch250

This part is bad because 100 is the chokepoint, and 694 has 2 lanes split off and become 100, then merges with 94 coming out of the cities.

So, until I pass 100, it's really slow. If they redo the stretch of 694 coming up to 100, it's gonna suck.

Same if they redo the stretch going the other way, where 94 splits off 694, and we merge with 100.

I might leave the state before then, though. The wide open west is calling me.

Montana or the Dakotas, perhaps. I'm looking at Sioux Falls and Rapid City and Fargo, especially.

Mathieu Guindon

3:44 PM

@M.Doerner agree in general - but if I can slip a fix for the SCP issues, I think they should go in too :)

this

The real question is when to release the 2.4?

Hosch250

@MathieuGuindon Stupid Code Pane?

this

I tried to test CE changes but I cannot because the application I'm working on OOMs.

LOL

Hosch250

Or Source Control Provider?

Mathieu Guindon

@Hosch250 that's a very, very good one :)

this

3:45 PM

I'm wondering how many else has tested the latest pre-release. The lack of new bug reports might just be that nobody has ran into it yet.

Mathieu Guindon

what about the IR TW work?

Hosch250

Ahh, Safe COM Provider. I always forget that one.

this

inspection? I think that should be 2.4.1?

Mathieu Guindon

@Hosch250 Self-Closing Pairs, actually

FreeMan

@this You're referring to TSA (or lack thereof) screenings? It's all just security theater anyway. Search for TSA red team to see some failures

Mathieu Guindon

3:49 PM

@this yeah, probably

M.Doerner

I am rather against merging anything that might even potentially break something not already broken at this point.

We are already missing the original release date of the references explorer by a month.

Hosch250

Like ISelection?

@M.Doerner RD always misses release dates.

We missed the release date of 2.0 by over a year.

M.Doerner

But we are mostly holding off from merging PRs because of the release since the original release date.

this

@FreeMan I didn't say TSA were effective. Only that it's slowing everything from excruciating crawl to agonizing wait of Death™

So to be clear --- we should at least merge the CE hotfix #3

then what? Release 2.4 after tomorrow or on the weekend?

Hosch250

Just release tomorrow, and do a quick hotfix or two.

M.Doerner

4:00 PM

We said that we want a week of field testing after the last feature related merge.

Mathieu Guindon

I really need to do the translations

M.Doerner

Moreover, we still need translations.

Mathieu Guindon

@Comintern is the CE PR good to go or you have more commits coming?

IvenBach

@Hosch250 don’t expose our shortcomings. It’s unflattering exposing what goes into the sausage.

Hosch250

I don't care.

I think the lack of deadlines helps keep code quality up.

IvenBach

4:03 PM

@this PBAP pineapple by agonizing pineapple seems more appropriate.

this

lol

IvenBach

So glad I can avoid travel.

this

@Hosch250 and that's what gets us Duke Nukem Forever 3D, right?

Hosch250

@this That was closed source, so...

Open source, anyone can work with what's there.

So, not really the same thing.

In open source, everything more or less is a release.

Mathieu Guindon

meh. the numbers disagree.

Comintern

4:08 PM

@MathieuGuindon That was the last of them last night.

Mathieu Guindon

people download the "green build" orders of magnitude more often than the latest pre-release build

Hosch250

That's where marketing comes into play.

IvenBach

Green release would mean we’re confident in stability and comfortable with a public offering of it. No?

Hosch250

@IvenBach No.

That's what the public thinks, but it's really not.

I've seen beta versions more stable than final releases, and most releases have an immediate hotfix.

Comintern

@Hosch250 Try installing a beta x-server or dbus on your Linux machine sometime and your attitude toward the non-stable branch will probably change.

Mathieu Guindon

4:11 PM

@Hosch250 except if the product is Visual Studio

Hosch250

@MathieuGuindon My daily driver is VS 2019 Preview 1.

It's nice and stable so far.

this

There's a difference between RC/preview vs beta/insider

blargh my grammar are break

Hosch250

Yeah, to the end user, there's a difference. But to a programmer, I'd say less of a difference.

I don't care if there's a bug in an open-source product I use if I know the language(s). I'll just fix it and use my version internally.

And maybe submit a PR.

IvenBach

You're definitely an outlyer.

this

^

Hosch250

4:15 PM

Mission success. That's been my goal for years.

this

i'd wager that a large majority of OSS consumers don't even know how to fix a bug, much less read source code

Hosch250

@this Mmm, I'm not so sure. VS is the open OSS tool I have on my work system.

IvenBach

You'd win that by a wide margin.

Comintern

If RD was Debian, master would be "stable", green release would be "testing", and the current next would be "unstable".

Hosch250

*nix users usually know some code, even if it's just Bash.

If Office or Windows goes OSS, yeah, easy.

this

4:16 PM

@M.Doerner / @MathieuGuindon OK to merge CE hotfix #3?

Comintern

Some of the config files in *nix almost qualify as code.

this

^

Duga

[rubberduck-vba/Rubberduck] retailcoder merged pull request #4727: Code Explorer hotfix 3 from comintern/Rubberduck/next into rubberduck-vba/Rubberduck/next

[rubberduck-vba/Rubberduck] retailcoder did something with some project card

this

and may require you to persue the man to know what it really does.

Duga

[rubberduck-vba/Rubberduck] retailcoder closed issue #4726: Folders multiply every time RD is refreshed

M.Doerner

4:17 PM

I have no idea about the CE. So I have no opinion.

Duga

[rubberduck-vba/Rubberduck] retailcoder is bored so why not move a project card

this

I guess my reluctance is more because I don't want to unnecessarily delay the release as I already did last week.

Duga

[rubberduck-vba/Rubberduck] MDoerner demilestoned issue #4609: ShadowedDeclarationInspection description text should mention fully qualified identifiers

[rubberduck-vba/Rubberduck] MDoerner milestoned issue #4609: ShadowedDeclarationInspection description text should mention fully qualified identifiers

[rubberduck-vba/Rubberduck] MDoerner demilestoned issue #4666: Data binding leak in the InspectionResultsControl

[rubberduck-vba/Rubberduck] MDoerner milestoned issue #4666: Data binding leak in the InspectionResultsControl

Comintern

IR should probably merge shortly after the green release to get more "in-field" time.

Duga

[rubberduck-vba/Rubberduck] MDoerner milestoned issue #4637: Code Inspection freezes when using the disable this Inspection feature

this

4:20 PM

yeah. About that....

Duga

[rubberduck-vba/Rubberduck] MDoerner demilestoned issue #4637: Code Inspection freezes when using the disable this Inspection feature

this

we have about 10 PR that needs to be merged.

I don't know about you all but I don't know if I really want to merge them all at once.

Duga

[rubberduck-vba/Rubberduck] MDoerner milestoned issue #4338: VB6: Code Explorer does not show Related Documents

[rubberduck-vba/Rubberduck] MDoerner demilestoned issue #4338: VB6: Code Explorer does not show Related Documents

[rubberduck-vba/Rubberduck] MDoerner demilestoned issue #4660: Clean up WPF leaks

this

At the same time, they all need to be merged sooner than later so that they can get the in-field testing for the same reasons.

Duga

[rubberduck-vba/Rubberduck] MDoerner milestoned issue #4660: Clean up WPF leaks

Comintern

4:21 PM

@this Why's that?

this

Decisions, decisions, decisions.

Duga

[rubberduck-vba/Rubberduck] MDoerner demilestoned issue #4085: AutoComplete - Exit Function bug

Comintern

Most of them are dealing with things that don't have a whole lot of interaction.

Duga

[rubberduck-vba/Rubberduck] MDoerner milestoned issue #4085: AutoComplete - Exit Function bug

this

a number of them have large scopes so if a new bug report comes in, it would take a while to track them to which the PR was the offending one?

Comintern

4:23 PM

And if they do, the interaction surface is the more important thing to check.

Duga

[rubberduck-vba/Rubberduck] MDoerner opened issue #4735: Release 2.4 needs French translations

> The title says it all.

[rubberduck-vba/Rubberduck] MDoerner assigned retailcoder to issue #4735: Release 2.4 needs French translations

Comintern

For the record, stuff like #4582 and #4731 shouldn't count for that determination.

Duga

[rubberduck-vba/Rubberduck] MDoerner added label resx to issue #4735: Release 2.4 needs French translations

[rubberduck-vba/Rubberduck] MDoerner added label user-interface to issue #4735: Release 2.4 needs French translations

[rubberduck-vba/Rubberduck] MDoerner milestoned issue #4735: Release 2.4 needs French translations

[rubberduck-vba/Rubberduck] bclothier merged pull request #4731: Add German translation for 2.4 for resources added late from MDoerner/Rubberduck/MoreGermanTranslations into rubberduck-vba/Rubberduck/next

[rubberduck-vba/Rubberduck] MDoerner pushed commit 18b1505f to next: Add German translation for 2.4 for resources added late

[rubberduck-vba/Rubberduck] MDoerner pushed commit b3052b74 to next: Improve some resources in English and German.

[rubberduck-vba/Rubberduck] web-flow pushed commit 4bfdbbb3 to next

Merge pull request #4731 from MDoerner/MoreGermanTranslations

Add German translation for 2.4 for resources added late

this

The 4582 probably should be merged soon as possible after the 2.4 release, yes.

Duga

[rubberduck-vba/Rubberduck] MDoerner opened issue #4736: Release 2.4 needs Czech translations

> The title says it all.

this

4:27 PM

That will hurt all of us so I'll get to know about it very quickly.

Duga

[rubberduck-vba/Rubberduck] MDoerner added label user-interface to issue #4736: Release 2.4 needs Czech translations

[rubberduck-vba/Rubberduck] MDoerner milestoned issue #4736: Release 2.4 needs Czech translations

[rubberduck-vba/Rubberduck] MDoerner added label resx to issue #4736: Release 2.4 needs Czech translations

this

So I'm thinking that after the 2.4 release -- merge the #4717 and #4072 -- both PR are UI-facing and hold off any merge for another week?

Duga

[rubberduck-vba/Rubberduck] MDoerner edited pull request #4733: Introduce the ISelectionService

Comintern

@this I'd consider #4717 almost completely bug fixes at this point. The only functionality I added was an additional grouping.

this

There's another issue, though - we have zero UI tests

Comintern

4:33 PM

Also, the sooner #4582 merges, the sooner I can do meaningful non-work at work.

this

so how do we prove our UI changes aren't regressing?

right now, it's clench our posteriors and hope for the best.

Mathieu Guindon

@this we have @FreeMan for that :)

Comintern

^

this

har har but you know I meant automated UI tests. ;)

Comintern

That's the problem with UI - you need a meatbag.

this

4:34 PM

Well not necessarily anymore.

Comintern

? Should I anxiously await your MockWpf implemenation?

I'm talking about things like "the cursor doesn't stop spinning". Or "the context menu pops up really slow".

this

lol, no. I'm talking about the UI unit test framework they released some time ago

Those stuff you cited, unfortunately might be impossible to test in an unit test

That's more of an integration test which probably needs a meatbag.

Duga

[rubberduck-vba/Rubberduck] rubberduck-vba-releasebot published prerelease Prerelease-v2.3.1.4456

[rubberduck-vba/Rubberduck] rubberduck-vba-releasebot created tag Prerelease-v2.3.1.4456

[rubberduck-vba/Rubberduck] build for commit 747189b4 on next: AppVeyor build succeeded

Comintern

I should start on my implementation of MockMeatBag.

Mathieu Guindon

lol

this

4:40 PM

shouldn't be too hard. Just toss in some Rnd() function and you're good.

FWIW - apparently that framework is recommended now

Comintern

Interesting. Is it compatible with AV's setup?

this

well it installs as a nuget package?

I don't know if it can be run headless, though. I just found out about it now

Comintern

And then communicates with the running application via a server.

My read of it is that you'd need Office installed to use it with RD.

this

github.com/Microsoft/WinAppDriver/issues/…

looks like someone has done it

and we definitely don't want to interact w/ Office. That should NOT be used in a headless environment. We probably have to provide a mock application

Comintern

Interesting.

I might need to read that more thoroughly.

Mathieu Guindon

4:46 PM

sounds pretty complicated for something that won't catch half the things F5+eyeballing would... no?

this

well, right now we avoid regression in the functionality by writing a red unit test with a complicated mock parser/project builder.

Comintern

Yeah, there's really only a small subset of stuff you could catch with it.

this

since the mock parser/project is already all set up it's easy to write a new unit test.

Comintern

Like circular bindings for example.

this

But we have basically nothing to verify that the UI is the way we expect it to be.

Comintern

4:48 PM

Eyes?

this

If it could catch data binding errors that's a win, no?

Eyes are unreliable and can't be relied on to do repeated/automated tests to prevent regressions.

Comintern

Catching binding errors would definitely be a win.

this

A different route would be to use something like this: github.com/bblanchon/WpfBindingErrors

Though I'm not as clear as how we'd use that in an automated unit testing system.

Duga

[rubberduck-vba/Rubberduck] Vogel612 project card. Enough said.

Comintern

@this That would be worth it just to get the damned things in the log files.

this

5:01 PM

doh - try and read a bit more: github.com/bblanchon/WpfBindingErrors/tree/master/…

now that'd be very simple way to just validate that none of our WPF stuff has data binding errors.

Duga

[rubberduck-vba/Rubberduck] rubberduck-vba-releasebot published prerelease Prerelease-v2.3.1.4457

Comintern

I'm sold. I may need to play with that.

Duga

[rubberduck-vba/Rubberduck] rubberduck-vba-releasebot created tag Prerelease-v2.3.1.4457

[rubberduck-vba/Rubberduck] build for commit 4bfdbbb3 on next: AppVeyor build succeeded

this

we might need both, actually. Having it in log would make it easy to write new red unit tests.

Just to be clear, is it really all about data binding errors, though?

Could there be other classes of UI errors that we need to be testing for?

KySoto

@Hosch250 going to in about... 5 hours or so?

Comintern

5:24 PM

@this The binding errors are the ones that are most difficult to find IMO - about the only way I've figured out to find them is by running it and watching the tail of the debug output.

Mathieu Guindon

and then scratching your head for 3 days trying to figure out what the heck the stupid errors mean. at least, that's been my experience with that damned CRUD app so far.

Comintern

IKR?

this

Mat, curious, was that going to replace the Excel-as-an-application?

Comintern

My favorites are the ones on properties that you aren't even binding to directly.

Mathieu Guindon

@this yes

this

5:27 PM

Comintern, would I be correct to presume that in the cases of the tribble trouble or missing menu we can write an unit test without involving the UI, right?

Comintern

Yes.

There are about 60 of them in the last push.

this

Yeah, I only skimmed. :(

Mathieu Guindon

"context menu is slow to open" is meatbag stuff though. or profiler stuff involving every command's CanExecute

this

Agreed.

I don't think that's what UI testing should cover anyway.

Comintern

The race condition I found on the EvaluateCanExecute for the refresh command is a perfect example of something you'd never find without meatbagging.

this

5:31 PM

The samples I saw so far basically covers the questions like "is the menu item enabled and available at that moment?" or "does it pick up the shift modifier modifying my keyboard input"

Comintern

@this The UTs are set up this way: github.com/rubberduck-vba/Rubberduck/blob/next/RubberduckTests/…

There's another set that pushes them through the main view model too.

this

Right. In that case, though we're testing the viewmodel in isolation.

The UI tests was supposed (at least I think so) to enable testing of the view (e.g. the XAML itself)

Comintern

IfIJamOnTheRefreshButtonLikeAJackHammer_DoesItCrash

this

lol

Transcript for