VBA Rubberducking

this

18:02

the inputbox should have its own property on the Vm to bind

You should then have the OnPropertyChanged(nameof(InspectionSettings)) called in the setter

then within the InspectionSettings, you should read the content of the inputbox.

Mat's Mug

I'd try to leverage something built into the grid control, if it exists

this

#NeedToLearnMore

IvenBach

^

That's why I'm attempting to solve this.

It's the only way I'll learn.

Mat's Mug

it's entirely possible the grid doesn't have anything prebuilt for that though

Vogel612

I'm pretty sure that it does..

this

18:07

remind me - wasn't that a customized control we created ourselves?

Mat's Mug

the GroupingGrid, yes. but it's based on an existing one

this

gotcha

IvenBach

ListCollectionView InspectionSettings is what everything is contained within

Mat's Mug

and even so, anything is 100% customizable in XAML

IvenBach

@AndrzejO Welcome to the pond!

this

18:08

except for that silly dropdown arrow. :\

well it is customizable but last time I looked it meant opening up the control template.... waaaayyyyyy too complicated

Mat's Mug

yup

hence why there's no damned dropdown arrow on these menu buttons :)

Hosch250

OK, I just pushed what I have.

this

i just wanna ShowDropdownButton=Always, for goodness' sake.

Hosch250

It recovers, but it doesn't do anything.

this

maybe that's what we can get Hosch to do? Gimme us sane dropdown arrows?

Mat's Mug

18:10

@Hosch250 what does that mean?

Hosch250

The parse always "succeeds".

this

will we get Error if there were?

Hosch250

Nope.

this

Hm. I just realized. Refactorings assumes that they are operable if parser is successful

Hosch250

I haven't been able to hack that to work. I need to think about design some before I finish that.

this

18:11

we need to update those so that they are disabled if we've selected a block of code that failed to parse

Hosch250

It's in my ANTLR-recovery branch on my repo.

@this Nah, they actually take that into account.

this

oh? Cool.

Mat's Mug

the parse tree should have error nodes when that's the case

it's these nodes we'll want to red-squiggle

this

cos for EM, I think there was a check like RubberduckParserState.State == OK in the command or soemthing

which would be now incorrect, I think?

Mat's Mug

not incorrect, just irrelevant

Hosch250

18:14

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

M.Doerner

I have just tested the logging of COM releases and I really thing that we should only log that in debug builds: we release quite a lot.

Mat's Mug

we'll need some HasErrorNodes extension method if such a method doesn't already exist on IParseTree

Hosch250

TFL.

this

@M.Doerner the only thing that need to be in release is if there's an error condition (e.g. when _release is non-zero

the rest can be a trace.

M.Doerner

I made them warnings, but we usually get trace level from users if there is a bug.

this

18:16

oh, i thought trace wasn't hte default

is there something lower than that?

M.Doerner

The default is NoLogging.

Threre are hundreds of COM releases.

this

just for my clarity -- Debug < Trace < Warning < Error ? or is it Trace < Debug < Warning < Error?

M.Doerner

And quite a lot fail.

Trace>Debug>Info>Warning>Error>Fatal, I think.

Mat's Mug

^ yup

this

well, poop

if we already ask users to use Trace for troubleshooting, that might be too much

DeepTrace>Trace>Debug>Info>Warning>Error>Fatal?

M.Doerner

18:19

To be fair, we already generate walls of text on trace level in the reference resolver.

this

I'm thinking primarily about what we currently do when we are helping users troubleshoot

M.Doerner

We just ask them for a log and usually we get the trace level.

this

I see

thinkinga bout it more, if we had DeepTrace, they might just select that anyway

Mat's Mug

^

this

and we'd just have bigger log

meh. I'm fine with it being at the Trace level

since it's not the default

and we can come up with a strategy to parse the wall of text if we really need to.

what I do want, is that we are able to collect COM info from users

M.Doerner

18:21

Btw, the logging in the reference resolver takes a considerable amount of its runtime.

this

cos from what I saw, it's going to be very hard to reproduce.

so it's better that we are able to collect all info we can from user's environment

so I vote to leave them at Trace

M.Doerner

Then we should really make sure that before we release the next time, we are doing things right. I am seeing soo many failed releases.

this

hmm. interesting.

any patterns to the failed releases?

and exactly how do we fail to release them?

M.Doerner

I have not really investigated so far.

this

Oh ok. and just before you run down that rabbit hole

did your change still allow you to exit cleanly?

M.Doerner

18:25

Yes

this

well, that's very very good

M.Doerner

Same amount of errors as before.

this

so we now have much more information but no new errors. Excellent!

M.Doerner

I have just added capturing the exception failing the release.

That will make the wall of text quite a bit larger.

this

just realized now - it's very possible that the reason last attempt in 2.0 bricked could very well because we did not unsubscribe the events which was fixed very recently in one of pRs while we were releasing everyting.

Well, that might necessitate us to write a log parser but if that's what we must do, then we shall do.

heck, just dump it in Excel

slap a table on it

boom, filtering!

Mat's Mug

18:29

hehe.. that's quite exactly what I do

(with my VBA LogManager outputs anyway)

this

curious - does your LogManager split fields?

I think our logs are just one big sentence ATM. Would be good to split them so they are more amenable to filtering

Mat's Mug

the LogManager isn't responsible for log message formatting :)

but the DefaultLogMessageFormatter does this:

Private Function ILogMessageFormatter_FormatMessage(ByVal level As LogLevel, ByVal loggerName As String, ByVal message As String) As String
    ILogMessageFormatter_FormatMessage = Format$(Now, "YYYY-MM-DD hh:mm:ss") & vbTab & loggerName & vbTab & FormatLogLevel(level) & vbTab & message
End Function

so, yeah

this

ok so yeah, it's tab-delimited.

M.Doerner

Ah, my bad.

Nothing failed

I just have been blind.

this

what was it that you missed?

M.Doerner

18:31

I have been modifying a finally block, not a catch block.

this

ah, that'd make a big difference. :)

M.Doerner

stupid me

this

it's monday.

Mat's Mug

@M.Doerner if you are stupid, Humanity is doomed

=)

this

^

Hosch250

18:32

@this LOL.

@Mat'sMug It is anyway.

Mat's Mug

yeah I know. just don't tell my kids just yet.

this

so in fact, we are now releasing everything correctly, @M.Doerner?

Hosch250

Want to bet on how long it is before they realize they didn't ask to be born?

@this AFAWK.

Mat's Mug

sucks even more to be their kids

Hosch250

Which is partly why I'm not interested in having kids.

Mat's Mug

18:34

@this wait so crashing on exit has nothing to do with releasing any of the COM objects???

this

it's only one of factors

the other access violations that @M.Doerner was observing is long after the vbe7.dll has unloaded

which may be.... 1) the CLR runtime is trying to access COM object when it shouldn't have been, 2) VBE is confused about who owns the COM object, 3) our Win32 handlers are all wrong, 3) @this doesn't know what he's talking about

But having more data than we originally had --- that's a big step forward because now we can point to something that's working right and rule it out from our list of possible causes.

as opposite to playing lotteries.

Mat's Mug

#ANewHope

WaynePhillipsEA

hello everyone.

Mat's Mug

hi!

WaynePhillipsEA

Looks like you're talking about the COM issue you wanted me to look at.

Mat's Mug

18:42

yeah

WaynePhillipsEA

So you're releasing all COM objects explicitly at disconnection now? Do you mind if I run it through my COM proxy logger to check that

?

Mat's Mug

@M.Doerner can you push what you've got to [next]?

M.Doerner

Really, we are not failing on the ones we do release.

This does not mean that we release everything.

WaynePhillipsEA

Right, it's probably worth me redoing the log anyway for you, as it will give you some clues as to what you're not yet releasing properly (if any)

Mat's Mug

last I tried releasing everything this is what I was getting:

that was quite a while ago though

this

18:45

that means you released it too early

WaynePhillipsEA

releasing too early

^

Hosch250

Oh hi, @WaynePhillipsEA.

M.Doerner

I am right in the process to do my git routine to push to the PR.

WaynePhillipsEA

howdi!

M.Doerner

Hi

Hosch250

18:47

So, you are the VB Watchdog guy?

this

While @M.Doerner is working on the push - do you think it's possible that the access violations we observer could be not anything to do with COM?

Yes, @Hosch250

WaynePhillipsEA

yup, that's me

Duga

[rubberduck-vba/Rubberduck] MDoerner commented on Rubberduck.VBEEditor/SafeComWrappers/Office.Core/CommandBar.cs of pull request #3615: COM Release for menu items and command bar

Mat's Mug

@WaynePhillipsEA did you glance at our source code? anything glaringly stupid jumping at you?

WaynePhillipsEA

@this anything is possible :)

this

18:48

heh. should have known better than to ask

WaynePhillipsEA

I had a quick look. lots of possibilities. Like the window hooks

this

not sure if I mentioned that but it may help to note that we have some win32 handler stuff to intercept WM messages yeah

that

apparently it was because we weren't getting what we needed from VBIDE api

WaynePhillipsEA

yeah, not exactly .net friendly

this

so the guy, @comintern, set it up

he was the RD's "COM expert" who got everyone that far along.

WaynePhillipsEA

yeah, I can appreciate you'd need some window hooks to do some of your magic

Mat's Mug

18:49

the VBIDE/Extensibility API isn't quite extensibility friendly for all I've seen

WaynePhillipsEA

VBIDE is crap for events

IvenBach

@this I did that. I know I have other things wrong, no results are showing when I type anything in. But it only filters when I click OK which isn't what I expected with the binding.

WaynePhillipsEA

bbl got to see to the kids

Hosch250

@IvenBach Try setting the binding to update on property changed.

Mat's Mug

come back anytime!

Hosch250

18:51

See you.

this

@IvenBach hence the need for the previous line that you linked, as @Hosch250 said.

IvenBach

@Hosch250 I don't understand what you're saying.

It's my lack of coding experience.

Hosch250

Show me what you have.

IvenBach

private string _inspectionSettingsFilter;
public string InspectionSettingsFilter
{
    get => _inspectionSettingsFilter;
    set
    {
        if (_inspectionSettingsFilter != value)
        {
            _inspectionSettingsFilter = value;
            OnPropertyChanged(nameof(InspectionSettings));
        }
    }
}
private ListCollectionView _inspectionSettings;
public ListCollectionView InspectionSettings
{
    get
    {
        if (string.IsNullOrEmpty(_inspectionSettingsFilter))
        {
            _inspectionSettings.Filter = null;

Hosch250

And the XAML?

IvenBach

18:54

                    <DockPanel Background="DarkGray" FlowDirection="LeftToRight">
                        <StackPanel Orientation="Horizontal" DockPanel.Dock="Left">
                            <Label Foreground="White"
                               FontWeight="SemiBold"
                               Content="{Resx ResxName=Rubberduck.UI.RubberduckUI, Key=CodeInspectionSettings_InspectionSeveritySettingsLabel}">
                            </Label>
                        </StackPanel>
                        <StackPanel Orientation="Horizontal" DockPanel.Dock="Right">

Hosch250

What do you see when you press , here? Text="{Binding InspectionSettingsFilter^}"

IvenBach

While RD is running?

Hosch250

No.

NVM, I got it in my IDE.

Stick this at the caret: , UpdateSourceTrigger=PropertyChanged

Otherwise, it won't update the binding until the textbox loses focus.

IvenBach

Even then it wasn't.

So I get what that addition does. Text="{Binding inspectionFilter, UpdateSourceTrigger=PropertyChanged}" tells it to update whenever it changes IE key is pressed.

Mat's Mug

no, whenever the property changes

Hosch250

18:59

Whenever the UI or VM property changes.

Mat's Mug

which may or may not be because of a keypress

IvenBach

Words fail me again...

Mat's Mug

we'll make you a perfect little pedant before you know it

IvenBach

It's going to take a long while for that to occur.

hitting a key doesn't cause the property to change. I have F9 breakpoints on both get and set of InspectionSettingsFilter.

Duga

[rubberduck-vba/Rubberduck] MDoerner synchronized pull request #3615: COM Release for menu items and command bar

IvenBach

19:03

Had the wrong name for the binding... #FailingFaster

Duga

[rubberduck-vba/Rubberduck] build for commit a398ebbc on unknown branch: AppVeyor build failed

BUILD FAILURE!

M.Doerner

Interesting

IvenBach

Should a CodeInspectionSetting.Description ever be null? I'm getting another one in language opportunities without a description. Wasn't that tied to a config file?

M.Doerner

Something went wrong in the build process on AppVeyor.

It cannot find some files.

Mat's Mug

@IvenBach they need to have a description...

Hosch250

19:14

Stepping out for a bit. Can't focus.

Mat's Mug

@M.Doerner I'll have a look in a few moments.. can't right now

M.Doerner

np

IvenBach

@Mat'sMug I'm getting one that doesn't. You helped me delete a config file before that caused any blank ones to be properly generated. Do you remember which it is?

On a positive note I was able to achieve the filter properly.

Andrzej O

@IvenBach Hi Iven! Hi All1

IvenBach

If you have any questions about RD don't hesitate to ask.

@Mat'sMug ^ That's the null Description I'm getting.

Mat's Mug

19:28

@IvenBach the config file - there's only one

%appdata%\Rubberduck\Rubberduck.config

IvenBach

#WordFail

Duga

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.VBEEditor/SafeComWrappers/SafeComWrapper.cs of pull request #3615: COM Release for menu items and command bar

IvenBach

Hrm... Still didn't fix it.

Mat's Mug

typo in the resource key name?

(assuming there's a resource key for it)

Duga

[rubberduck-vba/Rubberduck] IvenBach synchronized pull request #3622: Issue3485 inspection sorting order

[rubberduck-vba/Rubberduck] build for commit 43e75ddc on unknown branch: AppVeyor build succeeded

IvenBach

19:38

I don't see it in the resource.

Confused as to where it's getting the initial values from then.

Mat's Mug

look at the xaml bindings; they're binding to a property of InspectionSetting. then look at how that property is implemented

it's fetching a specific resource key

Duga

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.VBEEditor/SafeComWrappers/Office.Core/CommandBar.cs of pull request #3615: COM Release for menu items and command bar

Mat's Mug

@M.Doerner hmm seems the post-build copy is pointing to \Debug

> C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\MSBuild\15.0\Bin\Microsoft.Common.CurrentVersion.targets(5‌074,5): error MSB3073: The command "copy "C:\projects\rubberduck\Rubberduck.Inspections\bin\Release\Rubberduck.Inspection‌s.dll" "C:\projects\rubberduck\RetailCoder.VBE\bin\Debug" /Y" exited with code 1.

check your .yml, line 50

Duga

[rubberduck-vba/Rubberduck] Greedquest opened issue #3625: Keep Code Explorer permanently docked

> I've recently discovered the code explorer that Rubberduck provides; looks great. I'd like to replace the default explorer with it permanently, but every time I close and reopen Excel, the Code Explorer window loses its docking (so I have to go into the Rubberduck menu to get it back)

Is there a way to keep the window docked, or if not, could this feature be added? Even better might be to have the Rubberduck Explorer replace the normal explorer after the first parsing.

[rubberduck-vba/Rubberduck] Vogel612 added label duplicate to issue #3625: Keep Code Explorer permanently docked

[rubberduck-vba/Rubberduck] Vogel612 added label bug to issue #3625: Keep Code Explorer permanently docked

[rubberduck-vba/Rubberduck] Vogel612 added label user-interface to issue #3625: Keep Code Explorer permanently docked

[rubberduck-vba/Rubberduck] Vogel612 commented on issue #3625: Keep Code Explorer permanently docked

> Related: https://github.com/rubberduck-vba/Rubberduck/issues/3214
Duplicate: https://github.com/rubberduck-vba/Rubberduck/issues/3215

Long story short: we know. We're working on it :)

[rubberduck-vba/Rubberduck] Vogel612 closed issue #3625: Keep Code Explorer permanently docked

[rubberduck-vba/Rubberduck] Vogel612 commented on issue #3625: Keep Code Explorer permanently docked

> See also #3615 and #3614

IvenBach

AFAICT it looks like it's binding to InspectionSettings.LocalizedName which doesn't exist.

Mat's Mug

19:51

if it didn't exist, then none of the inspections would have a description

IvenBach

<controls:GroupingGrid ItemsSource="{Binding InspectionSettings}" ... confuses me. I don't understand how the binding can reference itself when the file is InspectionSettings.xaml

Mat's Mug

it's not referencing itself

remember all bindings are against a DataContext

and the DataContext is inherited, for every single node, from the parent

and we make that an instance of InspectionSettingsViewModel at the top-level node

IvenBach

d:DataContext="{d:DesignInstance {x:Type settings:InspectionSettingsViewModel}, IsDesignTimeCreatable=False}" is the main context for the user control.

Mat's Mug

which means the GroupingGrid's ItemsSource is binding to InspectionSettingsViewModel.InspectionSettings property

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

..which is assigned in the constructor:

        InspectionSettings = new ListCollectionView(
                config.UserSettings.CodeInspectionSettings.CodeInspections.ToList());

..which pulls a bunch of CodeInspectionSetting instances from the configuration

and, oh look!

=)

so this is the resource key:

 InspectionsUI.ResourceManager.GetString(Name + "Name", CultureInfo.CurrentUICulture);

IvenBach

^ I'm working through your breadcrumb trail here. I'll be a few minutes.

@Mat'sMug Mr smiley is taunting me. I'll pay him back with good PR one day.

Vogel612

19:58

relevant: Case 211: "No Magic"

Mat's Mug

IIRC the Name is literally the type name, e.g. FoobarInspection

so the resource key for that would be FoobarInspectionName

Transcript for