(BTW, trying to get dates prior to 100-01-01 gives weird results but not errors)

Also don't forget the precision - technically Date can be only precise up to a second. there has been articles out there where you can abuse a Date to report milliseconds precisions and cases where using SQL Server's old datetime can cause problems when comparing with VBA Date due to the precision mismatch.

I seem to remember an SO question about that.

I'm fairly sure there's lot more but all those add to this premise: "don't try to use Date like it's a Double or you'll get pwned."

and it does grates my nerves when I see some lunkhead going around "it's just like double! You can add to it and stuff!"

Oh, look, your duck needs its battery changed!

for funzies i found

?cdate(-657434)
1/1/100

add one more, and it overflows. Add a fractional and it goes in wrong direction.

yeha

it gets all nice and funky

Suggested inspection message: "Ducky gets very irate when you use date like a double. Do not taunt the ducky."

"Double Dating may lead to unexpected results."

mwhaha

now that's a gold badge I wasn't expecting today

not sure if I should be proud that this is my top post across the entire SE network

anyone else notice this:

nope

feel free to report on MSO

(or MSE if not just SO is affected)

the logo's been cut in half. I don't have any browser zoom on and I've not messed with Windows zoom, DPI or otherwise

looks normal here

and the favicon is grey scale - has it always been that way?

@FreeMan grayscale favicon sounds like meta

derp... that was a MSO question.

and judging from the colors of your screenshot, you seem to be using stylish

soo ... Try disabling that before opening a bug at SE

@Vogel612 ah, good point! All other SE sites seem to be OK, maybe it's just a meta-barf...

it's fine on regular SO:

if only I was that good at finding my own bugs...

All bugs are naturally repelled by their author.

ok that Excel ref bug might very well be the "VBA Question of the Year"

@MathieuGuindon What's your view on asking FizzBuzz to know if someone can code?

I've got an angle on it I think.

@FreeMan Any code one writes *never* has any bugs. That's the lie our brains force on us.

I just need to call the AddRef and Release functions for the underlying Range's IUnknown. In theory AddRef should return 1 while it's in that state. Otherwise Excel would probably leak memory like a sieve.

What's up with all those questions about web scraping & Excel?

Did I miss the memo that it's "Reinvent your wheel" day today?

My hobby is to point out the public REST APIs that they could be using instead.

That, too.

@IvenBach weeds out the clueless, nothing more

Yeah, there's two points, really.

I do wonder how many times the no-web-scraping clauses in all websites' TOS are violated daily.

1. Does the person say "wut??"

Wouldn't be surprised if it's somewhere in thousands but if millions... O_O

2. If they try to solve it, can they handle simple conditionals and loops?

Typically, if it gets a response, it's something like "I know about that but I don't want to pay for it".

@Hosch250 fizzbuzz is theoretically a simple design exercise for modularity

Huh. There was a reference count overflow vulnerability in VBScript.

because you can check how the interviewee handles and anticipates changing business requirements

OK, now I'm curious if Excel did the same thing with a dangling object reference prior to the fix for CVE-2018-1004.

> May 2018: Modified the behavior of VariantClear. Originally, when clearing a VT_DISPATCH, VariantClear would first release the dispatch interface and afterwards clear the VARIANT structure. After the patch is applied, it sets the VARIANT to VT_EMPTY immediately, prior to calling Release. (Note that VariantCopy makes use of VariantClear, so it inherits this behavior change as well.)

@Vogel612 It can really be almost anything you like it to be, TBH.

You can use it to weed out the clueless and incompetents.

And to discuss higher-level architecture.

And to make sure people aren't the type to get too involved in higher-level architecture.

I mean, if you say "Do FizzBuzz" and someone starts discussing high-level architectural patterns, would you hire them?

@Hosch250 so it's ok if the person is concerned with minimizing the register accesses? ;-)

@this It depends on which position you were looking for.

@Hosch250 If it's related to the problem at the hand, sure. If they just start randomly naming GoF patterns in abstract, probably not.

@Hosch250 if you say "do fizzbuzz" and they immediately started coding and it's not a junior position, they'd be out in my book

if they started abstractly discussing high-level arch that's also a red flag

FizzBuzz is not about the code

^ though in my case, we're usually looking for a business consultant more than a programmer.

Well, I don't know...

FizzBuzz is honestly not about weeding out the people that have no clue

At some point, it takes less time to risk redoing it than to discuss.

I mean: hello basic trivia quizzes

And FizzBuzz is definitely one of those problems.

It takes like 30 seconds to write.

ya, and probably either has two bugs or doesn't keep out the folks that have no clue

if you can memorize the solution to fizzbuzz, that doesn't mean you know things

Well, I had a remote interview once, and they had a shared-screen type thing. I got to the meeting early, so I just wrote fizz buzz and number guessing games to calm my nerves.

I ended up not getting the position, but it was because my math was weak.

TBH, though, fizzbuzz is kind of a s***ty interview question.

At the end of the day, it's the programmer who's left guessing what they want.

some business requirements are like that...

@Hosch250 In my book, that's a no-hire. That means they didn't listen or didn't ask questions to understand the requirements.

See, with fizzbuzz, the requirements are common knowledge.

it reminds me of the multi-inheritance koan from codelesscode

There's no need to ask the questions to understand them.

@Hosch250 nope. not really

If they want something other than the standard fizzbuzz, then they have no right to call it fizzbuzz.

At least around here if you're not a programmer, you probably haven't heard of fizzbuzz

@Hosch250 that's a humpty-dumpty argument.

It's like telling someone to go to the grocery store and get a banana.

You don't get to prescribe the language and terminology that your customer uses

I mean, you could discuss it and make them define the size, weight, shape, and type (red, yellow, Ecuadorean, etc)...

Or you could go to the grocery store and get a banana.

And if they complain, it's on them because they didn't say they wanted the standard yellow, but one of the sweet red kinds...

you're still missing the point....

because fizzbuzz is not about the solution to the problem

it's about how you got there

I don't think the analogy work.

The point of it is "Cook me a risotto"

Well, in the interview I can see they'd care about how I got there. But if anyone told me to drive to the grocery store a half block away for a banana, I'd laugh at them.

@Vogel612 I think that's a better analogy.

risotto rice and water are definitely givens, but d'you want onions?

what about peppers? zucchini? ham?

That's the whole point. Fizzbuzz is too standard and has well-defined specs. If you want something more vague, then use different terminology.

FWIW onions are a must-have, everything else is optional, but preferrably in tiny pieces

Maybe call it a fizzbuzzlizardspock.

Because then you don't know what the lizard is. It could be 7 (the next prime), or it could be "print 'lizard' if any number is matched by two subpoints".

For example, 15 would be lizard instead of fizzbuzz.

That one doesn't have the specs broadcast across half the tech blogs and tutorial sites out there.

Oh FFS, Microsoft. Why doesn't Excel show up in the ROT?

wut? that's now how automation works

support.microsoft.com/en-us/help/238610/…

> an Office application registers its running objects in the ROT once it loses focus.

wtaf

.... that's not right.

that's not how I thought it worked.

I don't think that's how it's supposed to work.

It "optimizes".

That part is true, yes

I wonder if Bing and Google have different definitions for "optimizes".

e.g. not loading into ROT immediately

but it has never AFAIK, been contigent on focus or lack of focus.

but spend a minute loading a bunch of bloatware COM add-ins

LOL - maybe that's why it needs to optimize the OLE stuff.

Hmmm... I wonder if I could use DispCallFunc to call AddRef...

Personally, I'd love for a plugin that will block the application from adding itself to ROT.

You can just remove it, can't you?

Alternately, you can hook RegisterActiveObject and discard calls originating from the target application's process.

@Comintern whats ROT?

Running Object Table. It's an old school OLE thing.

@Comintern because bit ROT?

LOL

@this I think "Do not taunt the ducky" should be the new slogan.

RE FizzBuzz: I've started asking for more information when given a request.

What is this accomplishing. Why is it structured this way. How can we expand on this.

@Comintern not if you are not the host, AIUI.

Hooking is probably necessary but you'd have to hook it before the host application calls that function

> Rubberduck.Setup.2.4.0.4488.exe (4.03 MiB) - downloaded 83 times. Last updated on 2019-01-28

I'm off to bed. Attempting to keep actually reasonable bedtimes.

Wow, that's early.

Well my GF sleeps for ~12 hours a day recently and shoots for 9AM

and her other BF needs to get up at 5:30 AM, sooo...

I don't care much either way, so long as it's consistent and it's not that I go to bed at 4AM (like the last three days)

Hm, after mergin next, I cannot build anymore.

unrelatedly: I'm hoping to get ReleaseCleaner into a state where I can automate removing old prereleases from the github release page.

we still have the 2.2.0 prereleases around

It says that license.rtf is missing.

and it's terribly tiresome to klick so often after scripting the removal of tags

Ok, where is that being used?

The case of the filename is wrong.

IIRC it's the copy that's displayed in the installer

I mean, where in the code?

that would be in RD.Deployment

@this?

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

@Vogel612 It might actually be easier to get GitHub to do auto-delete of pre-releases.

Send them a feature request to only keep the last N pre-releases.

whoops. Looks like it should be License.rtf. Sorry, @M.Doerner

Was that all you needed to get it to build?

@Hosch250 hmmm.. oh well. Gives me an excuse to work on something a bit differemt for a change any do some dotnet on linux

In other news, my boss's boss just asked me to stay on and co-architect the next rewrite of our project.

Not now, but sometime down the road.

What do I have to do to build the builder?

I think nothing.

since you didn't make any actual code change, right?

I do not have Rubberduck.Deployment.Build in my solution.

If you did make any code change, then you have to build the RubberduckMeta.sln

that's because that is now a separate solution.

Ah, OK.

you must unload the Rubberduck.sln.

The differing case is in the builder.

once htat's built, you would go back to Rubberduck.sln and it should then build successfully.

Where do I find RubberduckMeta.sln?

it is in the same directory as Rubberduck.sln

In retrospective, I should have made all filename checks be ToLowerInvariant() to avoid that being a problem.

The problem is that my reporitiry directory is case-sensitive.

I know.

However, my aim was to make it easier for contributors to contribute, that does matter to me.

It is the low-level IO erroring out, I think.

I think I am an outlier with the case-sensitive directory.

@M.Doerner so you could conceivably have a License.rtf, and a license.rtf file?

Yeah, because he made his folder containing the repository case-sensitive.

By default, Windows uses case-insensitive names.

But git and Linux generally are case-sensitive.

@Hosch250 yay! (right?)

Yeah.

I said yes, contingent on who the other architect was.

@KySoto Yes

funky

You can do that after installing the subsystem for Linux on Win10.

i guess me being a lazy windows 10 user

i just dont see it

I just did this, because we had wired case-incompatibility issues with folders before.

i dont REALLY have a good reason to have hte linux subsystem installed

I only installed it for the case-sensitivity choice.

Btw, it is per-folder.

I do really wish git or at least GitHub would enforce lettercasing in a consistent manner. It's very annoying when a Windows-based contributor makes a lettercase change. That can cause phantoms in the git repo. Not fun.

Darn, I changed the case of the wrong license file.

D'oh!

this is weird:

yeah tis

Pineapples need hobbies...

Meh, better that than popping off some bullets, I guess.

I still stand by my comment.

If they had hobbies they'd not be as high-strung.

Maybe stalking is their hobby.

Hm, now it does not find Rubberduck.idl.

Oh, no. could it be rubberduck.idl?

I have no idea.

it should be in the Deployment's output folder

Hm, it is there.

and the lettercase is correct?

Could it be a problem that it is not in the Debug subfolder?

Rubberduck.idl

I don't think so - I think all meta projects are suppoed to have their build unified.

what is the full error message?

one more thing - do you have C++ build tools installed?

I don't think so.

hmm then you shouldn't be trying to compile using MIDL.

you should be using tlbexp

obviously there's something wrong with the logic for switching there

@Comintern "Hey you. Get out of the bushes!"

rubberduckiegod?

To further confirm -

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

what message do you get in your output? You should see one of either message

I do not recall having the build tools, but apparently I have midl.exe.

16>N:\Repositories\Rubberduck\RubberduckBaseProject.csproj : warning MSB4011: "C:\Users\Dörner\.nuget\packages\sunburst.net.sdk.wpf\1.0.47\Sdk\Sdk.props" cannot be imported again. It was already imported at "N:\Repositories\Rubberduck\Rubberduck.Deployment\Rubberduck.Deployment.csproj". This is most likely a build authoring error. This subsequent import will be ignored.
16>N:\Repositories\Rubberduck\Rubberduck.Deployment\Rubberduck.Deployment.csproj : warning MSB4011: "C:\Users\Dörner\.nuget\packages\sunburst.net.sdk.wpf\1.0.47\Sdk\Sdk.targets" cannot be imported again. It was already impor…

Ok, so the logic is probably OK

since you did get a header for the midl.exe (rather than unrecognized program name errro)

It seems to fail to resolve the path expression.

The midl is kind of weird with how to form the path.

so this is the command we built:

midl.exe /amd64 /tlb "Rubberduck.x32.tlb" "Rubberduck.idl" /out "N:\Repositories\Rubberduck\Rubberduck.Deployment\bin"

@this I thought we determined when I was testing this that it didn't like the `.` for the local directory.

where am I setting the .?

It's in the output that Max pasted above.

Cannot open source file: '.\Rubberduck.idl': No such file or directory

Is that coming from midl.exe?

yes but I'm not passing that in

Gotcha.

in fact, I'm trying to see where the working directory is set - this seems to assume the path is relative for the inputs

@M.Doerner, if you were to paste the same command into your VS developer prompt, does it work?

(and if it does not work first time, does it work when you change directory to the bin folder)

Are the exit codes the same as the MIDL# codes?

I don't think/expect so - MIDL errors/warnings are emitted

Where do I find the prompt?

easiest to just type in start -> run => visual studio developer prompt

(or developer command line)

hm, on mine, it's actually Developer Command Prompt for VS 2017

@Comintern bTW, it starts at 2000

Nope

docs.microsoft.com/en-us/windows/desktop/Midl/compiler-errors

@this Those are the compiler ones.

Ah, leaving out the run did the trick.

do you not even see it under Visual Studio's Tools folder in the start menu?

sorry i meant run as the command.

that did not work

The preprocessor ones start at 1000: docs.microsoft.com/en-us/windows/desktop/midl/…

just opening the start menu and using the search did the trick.

1003 => The preprocessor returned an error. The error message is directed to the output stream.

so, no not applicable here, I'm sure.

it'd have been in his Output, I'd think.

What I'm finding indicates that there are missing header files.

wut

IKR?

I get the same error from the console.

even at the bin folder as the current directory?

and no more output?

@M.Doerner Do you have the Windows SDK installed?

@Comintern if he didn't, midl.exe wouldn't work at all

The .idl file we generate pulls in 2 required includes:

#include "rpc.h"
#include "rpcndr.h"

This is strange: using dir inside the folder does not show the finle, while the file explorer does show it.

If those aren't in the build chain - the preprocessor will throw.

finle?

Oh, just overlooked it.

@Comintern where do you see that?

Mine does not have that.

I cannot remember that I have ever installed the Windows SDK.

@this Line 27 and 28 of Rubberduck.h

we don't use that

damn, I need to make sure that midl stop spewing those stub files.

those are totally irrelevant.

(and besides, those are generated as result of midl.exe, not as an input)

Right, but it seems to indicate that it needs a valid build configuration that has a path to a headers directory that contains those 2 files.

That's why you use the developer prompt

it will set up the environment correctly so that midl.exe will function

without that, the midl.exe is all "WHERE IS MY KIDS!"

Huh? How does that ensure the headers are present? It can point to a directory, but it can't materialize the SDK files.

I'm pretty sure the header files aren't needed. midl is just generating the stub files based on the IDL file as the input.

@M.Doerner to confirm - this doesn't even work if you do it from the bin path, right?

nope

Correct, I tried before.

You have to use the dev command prompt.

@M.Doerner as a test, try this....

midl.exe /amd64 /tlb "Rubberduck.x64.tlb" "N:\Repositories\Rubberduck\Rubberduck.Deployment\bin\Rubberduck.idl" /out "N:\Repositories\Rubberduck\Rubberduck.Deployment\bin"

blargh, got that wrong - edited as above

that's interesting. /no_cpp causes this error.

64 bit Processing C:\Github\Rubberduck\Rubberduck\Rubberduck.Deployment\bin\Rubberduck.idl
C:\Github\Rubberduck\Rubberduck\Rubberduck.Deployment\bin\Rubberduck.idl(1) : error MIDL2025 : syntax error : expecting an interface name or DispatchInterfaceName or CoclassName or ModuleName or Libra
ryName or ContractName or a type specification near "midl_pragma"
C:\Github\Rubberduck\Rubberduck\Rubberduck.Deployment\bin\Rubberduck.idl(1) : error MIDL2026 : cannot recover from earlier syntax errors; aborting compilation

hmm, @Comintern you might be right after all: docs.microsoft.com/en-us/windows/desktop/Midl/…

Has not worked either.

Ok - could you find a rpc.h?

if you ever had windows sdk installed, it generally goes into the program files by default

I think 32-bit if the OS is 64-bit but not sure 100%

On my 64-bit Win7, it goes into `C:\Program Files (x86)\Windows Kits`, generally

I have to step away for a bit.

OK, apparently I installed a Windows SDK as part of something else in Octobre.

The other thing you need to check is to make sure that the path to the rpc header files is in your INCLUDE environment variable.

I has to be something with the case-sensitivity.

After disabling it again, the build succeeds.

TTQW, snowpocalypse expected tomorrow, again

I'll just keep it turned off.

@Comintern was thinking about that -- since midl.exe and several other c++ build tools needs SDK to run, I don't htink it's possible (typically at least) to have a midl.exe without the requisite headers.

@M.Doerner That is curious. I wonder what else I missed

@this I'm thinking that sounds right. My guess is that those headers are where IDispatch and IUnknown are defined.

Yeah

which would be why I'd get the weird compile error w/ no_cpp switch

anyway #TIL that midl.exe does need headers and preprocessing.

So for #3960 I'm adding collapse all and expand all buttons to the inspection results window. Do we do that anywhere else that I'm just missing, or do I need to track down a pair of icons to use?

@Comintern doesn't CE do that? ...with folders though