Fr0sT.Brutal

You could take current mouse coordinates but this will fail if button is clicked by key. Not a problem if the button is unfocusable

Yep and they're with us since D2009 or so. Even in D7 there were $message (without modificator). Though, it could be just a plain text or anything else: {$IF ...} BADABOOM! {$IFEND} but this makes source parsing harder (like generating docs) so $message is preferred. I also add such kind of checks to places where I rely on some assumptions. F.ex., I send short 4-char string code via PostMessage directly in WPARAM so I add compiler directive to check whether the code truly fits in SizeOf(WPARAM). If sometimes the code grows, these checks won't let me forget where I should modify the algo

Compiler directives come to help. This is to ensure TEnum is starting from 1 (needed to loop through a string that contains chars for every item in enum) const EnumLiterals: array[TEnum] of Char = ('Q', 'W', 'E', 'R', 'T'); function SetToFixedStr(aSet: TSomeSet): string; var elem: TEnum; begin // check {$IF Ord(Low(TEnum)) <> 1} {$MESSAGE FATAL 'Must start from 1 (using as string index)'} {$IFEND} ... end Or, in every case that handles ALL values in enum: case item of enumItem1: ...; enumItemN: ...; {$IF Ord(High(TEnum)) <> 5} {$MESSAGE FATAL 'Implement it'} {$IFEND} end; // case This will catch addition to every place of enum. In case of enums starting from custom number, use Ord(High(TStatsType))-Ord(Low(TStatsType)) to get number of items. Btw, FPC catches switches that don't handle every item in an enum. Sometimes it helps, sometimes annoys 🙂

Good point - I wasn't aware these two differ. And Neslib's implementation has stream parsing as well.

If the file is large enough to exceed available memory, probably you'll want to use stream-like parser, like SAX from XML. Grijjy has something alike https://blog.grijjy.com/2017/01/30/efficient-and-easy-to-use-json-and-bson-library/

But you can load them from config. In your sample, there are repeating 70, 40 etc. Unclear what these numbers mean You define structure in declaration but mix it with representation properties I'm not against attributes, I just don't see their real benefit. Any other examples are gladly welcome!

That's good but doesn't software implementation makes whole idea of TPM useless?

Honestly, this example rather shows how awkward attributes are. All options you provided seem to be better defined in form props/code than in class declaration. Drawbacks of attributes IMHO: - hard-coded values - constants are scattered and repeating - violation of abstraction of structure from style (like visual styles defined right in HTML) I could be wrong but so far I see only one really useful application - defining structures for (de)serialization but even here attributes are not required for simple cases. 2nd application, handy but not necessary (the only one I use myself) is marking DUnitX's test methods

My guess: M$ either looses TPM requirements or loses significant piece of virtualized systems market. Anyway W11 is not so much killer-featured to run and buy it. Some will reject upgrading, some will switch to Linux.

TPM sucks. Requirement of TPM sucks twice! AFAIU it could be used for DRM stuff so the soft, games and media you bought would be tied to the chip. Once it breaks, you lose all you've paid for. Moreover, it provides a unique identifier of a PC which destroys privacy.

According to https://m.eet.com/media/1089230/an087.pdf , function must be stdcall

BTW, you can debug DLLs by defining host process in Run options dialog. Then you can set breakpoints and see what's going on in fact.

Unit tests check just a part of the whole app. The main adventures frequently start when these tested units are combined into an app. That's where leaks could appear not detected by even the most severe test.

Sure

As for me, BUILTIN_TIMEOUT is MUCH more reasonable as default than USE_SSL xD

Change "true" to "not true" )))

And you either live without memory leak detection or get huge leak reports after every close.

Shouldn't CAPTCHA fields have "autofill =no" (don't know exact naming) property set?

Don't server's sockets inherit from TCustomTimeoutWSocket which has IdleTimeout that seem to do what OP needs? (never tried ICS-powered TLS server so just supposing theoretically)

Nice note. How did you produce custom RTL build? Just putting Classes and all its used units near the project or recompiled all units somehow?

Seems like it is. I misunderstood your point

Server-side: listening HTTP server processing subscriptions Client-side: HTTP client for subscribing to server and listening HTTP server for receiving events All could be implemented with ICS

Currently "out" is more just a usage hint than real compiler-controlled restriction. But even this state is better than "var". Hoping sometimes in Delphi v.25 Tegucigalpa they will add "use uninit'ed" compiler warning.

How would "if smth = Boolean(0)" help in this compared to "if not smth"? The only code I've seen where this could benefit is turning "if Condition then i := i + 1" to "i := i + Ord(Condition)"

I have x64 DLL that uses SysUtils and DateUtils weighting 590 Kb compiled with XE2. Unfortunately, RTL is not designed having little binary size in mind (bloated Classes unit that you have to include just when you need TStream).