Fun fact: That "MZ" required at the beginning of the header stands for Mark Zbikowski, who is credited with creating the .exe file format at Microsoft.
@@RvnKnight Well I've taken no such official classes. Just an obsessive need to learn the 'why' as much as the 'how' and whatever experience comes from having to find my own way through this never-ending cave.
@@RvnKnight I don't think it's that sad that people aren't aware of the MZ standing for Mike Wazowski. It's pretty much the most useless trivia imaginable.
But the file is nothing but its contents. Unless you're considering the directory entry of the file system part of the file, in which case you would also need to consider the filename and all the files metadata part of it. To my knowledge it isn't though. Directory structure is kept completely separate from file contents
@@HeyImSolaceFiles on your computer are stored with overhead. Some tools will tell you your file has a size of x bytes, and a size on disk of y bytes. This difference will exist even if the contents are 0bytes, since usually the system reserves a section of disk for it. This difference is op's 'being pedantic'
@@mars5train601 I know that this overhead exists, but this overhead is not part of the file. This is like saying a Google search result is part of a Website. It takes up space, but it's not part of the website itself, it merely shows some data of it (Title and a little preview) and points to where it is (its web address).
@HeyImSolace nah, thats the difference in memory vs storage. The file's metadata in the inode table has a minimum size. In windows ntfs (from googling), the equivalent looks to be the mft record, which clocks in at 1kb or 2 512B ssd sectors.
When I started watching, I somehow only expected the usual "overlapping the DOS and PE headers" and hello world message box, didn't expect you to enter the
@@TheSulross thats stupid, he found out how small it could be and then went on to, after answering the question, making something else related to the same topic, which again happened after answering the main question
17:46 bad video for security. Now the number of executables is finite and we can brute force through all of the 2^2^34 executables to find all of the penetration vectors.
@@samuelthecamel Well, it probably would. As technological advancements are constantly happening, once we have, per say, consumer-grade quantum computers, we just write the program, sit back, and enjoy.
The smallest piano having 555 Bytes, "like" THE one IC used for making simple sounds, the one and only NE555... That's just awesome and it cant be a coincidence
@@ping_th the NE555 is a small IC (integrated Circuit) / Chip used in electronic devices for timing/clock signal (and more) purposes. This chip (and its clones) are probably the most used IC for such stuff. It is so easy to use and cheap and widely used it became synonymous for this kind of usecase. Since any note you can play is nothing more than a sine wave with a specific frequency and the NE555 can easily produce such frequency you can use this chip to build a small piano (ignoring it not producing sine waves but square waves you then have to filter)
@@technikschaf1574 do modern computers still have this circuit, and if so, is it still possible to use it to make sounds/music that come straight from the computer?
Nearly 40 years ago, my first Computer was an Amstrad / Schneider CPC 464. I bought a monthly magazine which featured type in programs, back then called listings and there was always the 1kb program section, containing fantastic tools and full games with sound, colour graphics and joystick controls. I remember several Space Invaders clones, Shoot'em Ups, Racing and Puzzle games and an absolutely awesome Fractal generator, everything less than 1kb in Locomotive Basic and less than one page in the magazine.
Back in the day before magazines had download codes. Or DVDs. Or CDs. Or floppies. No, they printed the code directly and you had to type it in manually.
I was confused when I saw the video in my feed, thinking it would be basically the same as your video, but I love how this one explains everything visually! Kudos to you both
@@stalker32041i am not sure if it even has one, as i saw one being a symlink to systemd, which checks its filename on start and if it was started via the reboot link recommends the main systemd to reboot.
@@RWBHere that is a command which executes the program known as "shutdown" with the arguements "-r now" and has nothing to do with the size of the shutdown program, which includes reboot functionality
Fun video. ANSI (usually spoken as "ahn-see") control sequence originated with the Digital Equipment Corporation VT100. The SGR (Set Graphic Rendition) sequence was extended in "ANSI.SYS" on DOS to allow setting foreground and background colors. The glyphs the video shows beyond the ASCII characters are the VT100 graphics set, so the same program game would work using a serial port connected to a VT100+ terminal emulator that supported the color extension. The flickering could be reduced by only redrawing the bottom row except as the meteors shift down. Using save cursor and restore cursor sequences could improve keeping the little house's position from flickering. This will take more code code, of course.
Yeah and similar usage of ANSI is also great for anyone who enjoys torturing themselves making unnecessarily complicated batch files. Using it to clear lines and restore cursor positions instead of simply clearing it is much more pleasant and opens up some cool possibilities that wouldn't be possible (or practical) otherwise.
@@JJFX- as someone who has made unnecessarily complicated batch files, I appreciate all methods that make it less complicated, because that allows me to add more stuff.
@@HappyBeezerStudios It can simplify things but can become a headache to decipher later unless you mask repeated commands as vars. Aside from totally necessary colors and animations I mostly find it useful for advanced menu navigation, the ability to display information underneath an input line, temporarily replacing an invalid input with text, etc...
If I remember correctly, if you disable ASLR, you can simply jump to Win32 API function addresses. This will eliminate some strings in IMPORTS section of a PE file. I haven't tried it myself.
5:17 wait a sec, wait a sec, does every windows executable really contain ASCII art of a shrub by default? I'm tempted to assume that was a joke so you would have something to rhyme, but honestly I'm hoping that it's true because that would be a hilarious and genuinely beautiful absurdity.
@@avantesma1 one of the code bases my teams work with has a big old dragon ASCII. It isn't super uncommon. Though that is just made with comments so not technically the same
@@zpa89 If those etched Milhouses weren't functional parts of the chips, I'd say it's indeed, technically the same. Comments are the non-functional part of code, right? 🙂
One thing about compressors like crinkler is that you'll always have the potential hazard of the output executable being detected as possible malware. Most AV is leery of anything that looks like it's compressed, since it's a common way to hide telling features of the executable.
With the parameters now all in place, we can build Jorge Luis Borges' library of Babel, but for exe files. Since there's a limit to the number of bytes that can go in an exe and a finite number of values each byte can have, we can imagine a filesystem the contains every possible exe file. But it couldn't be a Windows filesystem.
10:56 those aren't actually ASCII characters (except for the @) ASCII is only a 7-bit encoding, and thus only contains 127 characters. anything beyond that into the 8-bit realm is a vendor-specific extension beyond ASCII
@@shallex5744 not only the value (not number) 0, but also the NULL byte. Sometimes used as a string terminator so not exactly a character, but if you think of the other control characters as a character, it can be argued that NULL is also one, which makes a total of 128 characters. Also, back in the day the remaining 128 values used to be known as high-ASCII which were more or less vendor or machine specific. Later they became code pages which were more or less standardised.
If you actually look at the image that you complain about, you first notice there are 16 rows with 8 columns, which excluding the NULL character makes for 127 characters. Then you notice that columns have 3 bits and rows 4 bits. Which is exactly 7 bits. And then you notice that those are precisely the ASCII characters.
Most useful usage of chatgpt for me when coding is simply to have it return documentation info...... and then hope it doesn't decide to just make something up which ends up wasting more time than simply digging into it myself :/
Another demoscene-friendly language is JS. In my experience, it's quite versatile. Stuff like fitting a game that was 30,000 bytes into 3,081 bytes. People have also made 140-character JS programs back in the olden days of X/Twitter when there was a hard 140 character limit.
it's the superior way to pronounce it. less syllables and easier to say. I've heard some people say ASCII one letter at a time, one of the worst experiences of my life
for anyone wondering, the character at 4:06 is "嗯", which (according to Wiktionary) is a Chinese character that can mean doubt, curiosity, surprise, agreement, "pledge", or just onomatopoeia for groaning.
You brought me back in time to my computer science days, making games on console applications. . Those were the simplest of days. . Thank God I still have them. . .
Back in the days of DOS, we (Crescent Software) sold a library for BASIC (called PDQ) which would allow for the quick compilation & linking of EXEs less then 512 Bytes. Why 'so big' (LOL)? Because that was the size of *ONE* sector on a *FLOPPY DISK*. Anything less was moot. It could also create TSR (Terminate & Stay Resident) programs, which were available with a keystroke or interrupt. And we could shove them up into Extended, or of you remember it, "Expanded" memory. Since this was in the days prior to math coprocessors, I wrote the 8087 emulator for the product. Thanks for the nostalgic trip down memory lane.
did it compile basic into a more native program instead of the more usual of adding the basic intepreter and the bas itself into an exe? I remember trying to coax borland c/c++ 3.1 into compiling as small executable as possible (I think I was like 11 back then). i think i got a 'flame' effect down to 8kb or so, which was still huge for what it was(just couple of loops really and calling bios to change the video mode and make the pallette).
@@lasskinn474 LOL! Yeah, these were genuine EXE files. The BASIC compilers BC 6 & BC 7 Professional were actually pretty good compilers for the time. BASIC code was lexed into 8086, 80286, or 80386 instructions. It even had math coprocessor options. statements like Print, Play, Open, Locate, REDIM, Strings, etc., generated calls into the runtime libraries just like C functions. We re-wrote the runtimes from *SCRATCH* in hand tuned assembly language (MASM). Because our libraries were so granular, the linker was able to bring in just the bare bones set of library routines that the programmer actually invoked. Nothing was just along for the ride. Since it was a library we were able to add extra functionality, such as the TSR capabilities, interrupt handlers, function pointers, pretty much anything a C coder could do, we put that power in the hands of the BASIC programmer. Then came Visual Basic for Windows. But that's a story for another day.
Just as a comment here, you don't want to do things like cmp al, 0. The reason is simple. You're comparing an immediate value and that value has to be stored in the instruction. Since you're only interested in the flags, you should test al, al whereby the assembly doesn't need to store the immediate value and it will set the flags for you to branch against.
@@InkboxSoftware Well when you've done as much assembly as I have and enter 4k/64k demo competitions at Revision and other parties, every byte does indeed matter! Good video though.
Maybe an interesting honorable mention: "fr-016c: fuenf (in your face)" by the demo group Farbrausch. It's a COM file that just produces an endless stream of garbage in the terminal, but because sometimes includes ascii bell symbols, it's technically an application with visuals and audio in 5 bytes. (It also crashes the terminal window)
it's been about 10 years since I last wrote any code, and about 12 years since I worked with processor instructions. This video made me remember that I loved this before my professors at college made me hate this with my soul 😅
I once made an executable a mere 20 bytes in size. It ran in DOS, was a .com, not a .exe. It's use was to set text resolution to 120x50 characters (if I remember correctly).
Over they years, Windows developers give zero consideration of the size of their programs. I'm glad to see someone demonstrating it is possible to write functionally interesting programs in the
*fun fact:* Registers: EDI, ESP, EBI and EFL could be used to power a RNG with some multiplications and/or additions. Some of them change to a pretty unique value only at start-up and then increment as the program is running, but others could be used to get multiple equally dispersed pseudo-random numbers.
I rember hearing this exact same fact somewhere else and that it exists and i think that there is a reason tl it but i dont rember Maybe an easter egg ???? Or testing idk
There used to be a full chess game for the Sinclair ZX81 (Timex 1000) which fitted in the 1k (1024 bytes) of static RAM which came with the system, from which some was used for the screen....
In practical terms, the smallest .exe file you can have is the cluster size the FAT uses on the storage device you're using. Assuming you saved your work that is.
The 2GB limit for RAM with 32 bit systems was a massive headache until I realised using GCC I was compiling for 32 bit OS's, when my OS was 64 bit. Changed that, and solved
I dived into bitmaps years ago and found unseen riches, unused digits at the ends of non 32bit arrays. nice work . we all need to keep kicking the tyres.
Just a heads up, the link in description for the diagram at 1:00 is not the original source, nor does it link to the original source. The author is 'corkami' (look up 'corkami pe101')
Writing small programs under an OS is fun. Try writing a bootable game :) People managed to pack whole simulations or games into the bootsector of a floppy. A sector has 512 bytes but the last 2 bytes are required or the BIOS won't recognise it as bootable (55AA in hex). Though the bootsector (depending on the medium) also contains some volume information. That only leaves a little bit left for the program. People have written bootsector games or simulations and still have the disk to work properly under DOS or windows. Also there are no sys calls to any OS function since we ARE the OS if we boot from this medium. All we have are the BIOS interrupts and direct memory access. The BIOS only loads the first sector (boot sector) of the medium which usually just contains code to load the actual OS boot loader from other sectors.
back in the day booter games on dos machines were not that common but there were a bunch of them. if you had only 1 floppy drive you didn't need to swap from dos, you didn't need to worry if the customers config.sys was weird etc and could have more confidence in whatever tricks you used to make it harder to copy the floppy. i think that there were those titles is one reason for why booting from floppy was default for so long on pc's so you couldn't leave a floppy in the drive.
Yeah but the BIOS _is_ an operating system. A very primitive one, but an operating system nonetheless. And a BIOS interrupt call is only two bytes. There's a plethora of them; printing text to the screen, loading things from the disk, setting timers, etc. When you programmed on DOS you made use of some them as well. Because DOS ran in real mode, it could share the interrupt table with the BIOS, so some interrupts where DOS system calls, while others were handled by the BIOS.
My game is 53 KB, runs even before/without Windows directly when the PC turn on power, and funny enough, it has the MZ header! I am quite happy with that 😇
Fun fact: I learned this from a Pascal 6 Book a friend read to me, at leas on MS-DOS it was 1 byte, the ╣B9h or Char 185, in a .exe file, which is the x86 Instruction RetF: Return Far, full qualified address: SEG:OFFSET is used to return to the Operating System. Great concise analysis, thank you
Decades ago, I came across an old DOS executable with just 64 bytes (!!), that produced a pretty impressive flight through a rotating cylindrical tunnel. It even had a simple blocky texture. I might still be able to find it, including the assembler code. It might even still run in Windows' CMD, even though it was a ".COM" executable, which had barely any header in contrast to ".exe" executables.
In my minimal work with exe compression in the demoscene I verified that if the PE header offset at 0x3C points to a value smaller than 0x40 then the microsoft signing tool will not recognize the image as an executable, which may be an indication towards future limitations. The space between 0x02 and 0x3C can still be used freely though... We use it to put a small vanity ASCII art header in there for our 64k intros ;)
That comment about every Windows executable having the same DOS header isn't true. You could specify an actual MS-DOS application as the header and create a dual-mode application. The same .EXE when run under DOS would run only the DOS "header" portion, while when run from Windows would skip over the DOS header and execute the Windows code. Many applications used to do this back in the day - and I've even created a few myself!
That's true, and that's why it has an MS-DOS header in the first place. It's just nowadays your standard compiler will fill in the string "This program cannot be run in DOS mode" automatically.
Essential code, the ASCII Art Of A Shrub (TM) stabilizes the program, protects against any form of cyberattack, and automatically generates cryptocurrency. It's the sequel to Sun Tzu's The Art Of War, even being more practical on the field of battle.
That sounds more like a COM file. EXE files *must* start with "MZ" to be valid; and then also contain a header (very similar to the one used by EXE files for Windows). COM files, on the other hand, were literally just something DOS loaded into a page of memory and then jumped to the first byte of. The smallest valid COM file is 1 byte, the encoding of the "RET" instruction (hex code C3). That would just immediately terminate and return control to DOS.
@@Drysart i know.. but I didn't say it was valid :D I discovered this as a kid who didn't have any compilers, but desperatly wanted to make my own exe. So, as all kids do, I just renamed a BAT to EXE to see what would happen 🤔 the first line was... .@echo off 😀 and the result was that it would restart when I run it. I progressively deleted stuff until I was left with just @e and it would still do the same.
linking an extra dll just for a rng is costing a lot of space. Ok not really a lot but you can do better. It would be smaller to implement a LFSR rng using a seed value from whatever source you can get that you're already have access to. A Console handle is probably good enough to use as a seed for a LFSR. The Handle values are likely going to be unique every time you run the exe though not guaranteed, the chances you'd get repeated Handle values are extremely unlikely. Or you can just go straight to using a cpu random instruction but you lose compatibility with old cpus.
it seems youtube deleted my reply so let's try it again... even though it's about elf32, I'd highly recommend the famous essay "A Whirlwind Tutorial on Creating Really Teensy ELF Executables for Linux"
Back in windows xp era I remember having a program that has no sections, only a header that also contains some code that restarts windows, IIRC its size was 97 bytes
Not sure if you already know about this, but you might be interested in Cosmopolitan libc, a project to compile c code to a polyglot binary running natively on Linux + Mac + Windows + FreeBSD + OpenBSD + NetBSD and even in BIOS
0:14 I use DUMMY,EXE files for my GPU's underclocking profiles. I make sure to TRIPPLE protect them, so only admin accounts can edit or change the exe. even the folder which they are located in is locked/write protected. and everything is in read only, even by default they are hidden files. so regular scrapers will overlook them
Books contain interesting exe files for copying and running in w3.1. These files mean, that expensive companies supported offices with such files. One is such, you can have your company name in w3.1 help menu, or just your signature. Other is such, you can use colour tiles as your paint in paintbrush, so a drawing can look like a colourful snake. Some others like radio buttons, etc. All show that hard wrking companies supported the Ms clients. It would be cool to introduce these secret programs in the new versions of Windows.
@@thewhitefalcon8539 He refers to all the DLLs, libraries, files, external to the Minecraft JAR or EXE file depending on the game edition (Java or Bedrock)
@@thewhitefalcon8539he means the entire game, not just the "executable" portion. (using quotation marks since it's java which is a VM and not a real executable)
![BLACK BAG - Official Trailer [HD] - Only in Theaters March 14](http://i.ytimg.com/vi/Du0Xp8WX_7I/mqdefault.jpg)
"I’m not sure typing simulator would become a best seller on Steam."
You'd be surprised what's taking over Steam tops these days...
banana
banana
terracotta
banana click simulator?
@@mjdevlog yeah, it basically just exists to make money off the item trading system on steam
Fun fact: That "MZ" required at the beginning of the header stands for Mark Zbikowski, who is credited with creating the .exe file format at Microsoft.
Yep, but sadly most IT/Is people never learn it unless they have the right class and right instructors.
@@RvnKnight Well I've taken no such official classes. Just an obsessive need to learn the 'why' as much as the 'how' and whatever experience comes from having to find my own way through this never-ending cave.
@@JJFX- Hence 'most*, not 'all'.
@@RvnKnight I don't think it's that sad that people aren't aware of the MZ standing for Mike Wazowski. It's pretty much the most useless trivia imaginable.
@@delphicdescant That's your opinion and you are entitled to it.
Just to be pedantic, the first file isn't actually 0 bytes. But its contents are!
But the file is nothing but its contents. Unless you're considering the directory entry of the file system part of the file, in which case you would also need to consider the filename and all the files metadata part of it. To my knowledge it isn't though. Directory structure is kept completely separate from file contents
@@HeyImSolaceFiles on your computer are stored with overhead. Some tools will tell you your file has a size of x bytes, and a size on disk of y bytes.
This difference will exist even if the contents are 0bytes, since usually the system reserves a section of disk for it.
This difference is op's 'being pedantic'
O Archon się tu zjawił.
@@mars5train601 I know that this overhead exists, but this overhead is not part of the file. This is like saying a Google search result is part of a Website. It takes up space, but it's not part of the website itself, it merely shows some data of it (Title and a little preview) and points to where it is (its web address).
@HeyImSolace nah, thats the difference in memory vs storage.
The file's metadata in the inode table has a minimum size. In windows ntfs (from googling), the equivalent looks to be the mft record, which clocks in at 1kb or 2 512B ssd sectors.
When I started watching, I somehow only expected the usual "overlapping the DOS and PE headers" and hello world message box, didn't expect you to enter the
Clearly the video failed hugely on its stated objective to create smallest possible .EXE. Have to give a grade of F for that reason.
@@TheSulross thats stupid, he found out how small it could be and then went on to, after answering the question, making something else related to the same topic, which again happened after answering the main question
@@128Gigabytes Sarcasm detection failure detected 👀😁
@@RoganGunn it very clearly isn't sarcasm based on context clues
F is the highest grade in the hexadecimal scale ⚖️ 😂
17:46 bad video for security. Now the number of executables is finite and we can brute force through all of the 2^2^34 executables to find all of the penetration vectors.
only for windows. unix/linux-based systems are still safe (but for how long?)
huh?
Good idea. I'm sure this calculation will complete before the heat death of the universe
While you're at it, you can figure out all the valid versions of Doom are in there ^^
@@samuelthecamel Well, it probably would. As technological advancements are constantly happening, once we have, per say, consumer-grade quantum computers, we just write the program, sit back, and enjoy.
"What is the largest possible EXE?" (Proceeds to develop an Electron application.)
I was exspecting this to just be a deep dive in to how small an exe can be, then he just made a game, this is amazing
Anticlickbait
@@egarcia1360 Yeah!!
@@SomethingCatchy2_is that twitter meme
@@NolanHOfficial ?
@@SomethingCatchy2_ "Yeah!!! 👍"
"Works twice as well as expected" is probably the best line i've ever heard. Great video!
The smallest piano having 555 Bytes, "like" THE one IC used for making simple sounds, the one and only NE555... That's just awesome and it cant be a coincidence
I thought the same thing lmao
I don’t understand your technical, so you mean like make a file that run beep/ small sound?
@@ping_th the NE555 is a small IC (integrated Circuit) / Chip used in electronic devices for timing/clock signal (and more) purposes. This chip (and its clones) are probably the most used IC for such stuff. It is so easy to use and cheap and widely used it became synonymous for this kind of usecase. Since any note you can play is nothing more than a sine wave with a specific frequency and the NE555 can easily produce such frequency you can use this chip to build a small piano (ignoring it not producing sine waves but square waves you then have to filter)
If he manages to shrink that by another 45 bytes, it'll fit into a boot sector and will clearly kick pac-man and doom from the most popular choices.
@@technikschaf1574 do modern computers still have this circuit, and if so, is it still possible to use it to make sounds/music that come straight from the computer?
I once talked to the kernel directly.
Now I got a lifetime supply of chicken wings.
Something is not right because it should have given you pop corn 🌽🍿😮
Why is your PFP Person... sideways...?
🏃♀️➡️🏃♀️➡️🏃♀️➡️➡️➡️➡️
@@computercoderawesome Gotta RunRunRunRunRunRRRRRRRRRRRRRR
𓊖 Emojitranslator.exe has stopped unexpectedly.
c
1:15 nice integer overflow lmao
idiots who dont know freaking a bit of coding arent allowed
Nearly 40 years ago, my first Computer was an Amstrad / Schneider CPC 464. I bought a monthly magazine which featured type in programs, back then called listings and there was always the 1kb program section, containing fantastic tools and full games with sound, colour graphics and joystick controls. I remember several Space Invaders clones, Shoot'em Ups, Racing and Puzzle games and an absolutely awesome Fractal generator, everything less than 1kb in Locomotive Basic and less than one page in the magazine.
Back in the day before magazines had download codes. Or DVDs. Or CDs. Or floppies.
No, they printed the code directly and you had to type it in manually.
11:58 „And that works twice as well as I expected“ Best line every
Great video! I never considered what the *largest* would be! Now I know :-)
Thanks for giving it a watch, it means a lot coming from you.
I was confused when I saw the video in my feed, thinking it would be basically the same as your video, but I love how this one explains everything visually! Kudos to you both
@@InkboxSoftware for a guy who scammed non educated ppl by selling fake cleaning software LOL
I remember having an exe file on my Amiga which was only 56 bytes in size: "reboot". This in fact rebooted the Amiga computer when you executed it.
I don't checked size of linux "reboot" program...
@@stalker32041its a symlink to systemctl if on systemd
@@stalker32041 Fifteen characters, 'shutdown -r now' is enough to do that, as root, from the Command Line.
@@stalker32041i am not sure if it even has one, as i saw one being a symlink to systemd, which checks its filename on start and if it was started via the reboot link recommends the main systemd to reboot.
@@RWBHere that is a command which executes the program known as "shutdown" with the arguements "-r now" and has nothing to do with the size of the shutdown program, which includes reboot functionality
Fun video.
ANSI (usually spoken as "ahn-see") control sequence originated with the Digital Equipment Corporation VT100. The SGR (Set Graphic Rendition) sequence was extended in "ANSI.SYS" on DOS to allow setting foreground and background colors. The glyphs the video shows beyond the ASCII characters are the VT100 graphics set, so the same program game would work using a serial port connected to a VT100+ terminal emulator that supported the color extension.
The flickering could be reduced by only redrawing the bottom row except as the meteors shift down. Using save cursor and restore cursor sequences could improve keeping the little house's position from flickering. This will take more code code, of course.
Yeah and similar usage of ANSI is also great for anyone who enjoys torturing themselves making unnecessarily complicated batch files. Using it to clear lines and restore cursor positions instead of simply clearing it is much more pleasant and opens up some cool possibilities that wouldn't be possible (or practical) otherwise.
*usually pronounced an-see
@@JJFX- as someone who has made unnecessarily complicated batch files, I appreciate all methods that make it less complicated, because that allows me to add more stuff.
@@HappyBeezerStudios It can simplify things but can become a headache to decipher later unless you mask repeated commands as vars. Aside from totally necessary colors and animations I mostly find it useful for advanced menu navigation, the ability to display information underneath an input line, temporarily replacing an invalid input with text, etc...
If I remember correctly, if you disable ASLR, you can simply jump to Win32 API function addresses. This will eliminate some strings in IMPORTS section of a PE file.
I haven't tried it myself.
Is that portable between Windows versions though?
@@shinyhappyrem8728 Nope.
5:17 wait a sec, wait a sec, does every windows executable really contain ASCII art of a shrub by default? I'm tempted to assume that was a joke so you would have something to rhyme, but honestly I'm hoping that it's true because that would be a hilarious and genuinely beautiful absurdity.
Just a joke :-)
This would be a software equivalent of those dudes etching Milhouse in the corners of their microchips.
@@avantesma1 one of the code bases my teams work with has a big old dragon ASCII. It isn't super uncommon. Though that is just made with comments so not technically the same
@@zpa89 If those etched Milhouses weren't functional parts of the chips, I'd say it's indeed, technically the same.
Comments are the non-functional part of code, right? 🙂
14:04 big thumbs up for CISC, without it we wouldn't have such gems as "AAA" and "PUNPCKHQDQ"
And the lovely EIEIO
One thing about compressors like crinkler is that you'll always have the potential hazard of the output executable being detected as possible malware. Most AV is leery of anything that looks like it's compressed, since it's a common way to hide telling features of the executable.
Ah, that's why windows defender detects it as a trojan.
@@RainDownpours I got that too
With the parameters now all in place, we can build Jorge Luis Borges' library of Babel, but for exe files. Since there's a limit to the number of bytes that can go in an exe and a finite number of values each byte can have, we can imagine a filesystem the contains every possible exe file. But it couldn't be a Windows filesystem.
Here you go: github.com/InkboxSoftware/DatabaseOfDamascus
10:56 those aren't actually ASCII characters (except for the @)
ASCII is only a 7-bit encoding, and thus only contains 127 characters. anything beyond that into the 8-bit realm is a vendor-specific extension beyond ASCII
Where's the 128th value?
@@sttate the 128th value is the number 127. the 1st value is the number 0
@@shallex5744 not only the value (not number) 0, but also the NULL byte. Sometimes used as a string terminator so not exactly a character, but if you think of the other control characters as a character, it can be argued that NULL is also one, which makes a total of 128 characters. Also, back in the day the remaining 128 values used to be known as high-ASCII which were more or less vendor or machine specific. Later they became code pages which were more or less standardised.
If you actually look at the image that you complain about, you first notice there are 16 rows with 8 columns, which excluding the NULL character makes for 127 characters. Then you notice that columns have 3 bits and rows 4 bits. Which is exactly 7 bits. And then you notice that those are precisely the ASCII characters.
@@marsovac i wasn't focused on the image, i was referring to him suggesting that the house and the horseshoe-looking thing were ASCII characters
Dude... Dealing with microsoft's documentation practoces, that's the challenge I saw. Amazing work of art
Most useful usage of chatgpt for me when coding is simply to have it return documentation info...... and then hope it doesn't decide to just make something up which ends up wasting more time than simply digging into it myself :/
@@JJFX-just tell it DO NOT HALLUCINATE . That’s what Apple does on their prompts
@@spht9ng lol "DON'T F'IN LIE TO ME!"
2:47 was a BRILLIANT sponsor cut?
After a certain time I can cut them out for a better viewing experience
@@InkboxSoftwareUnbelievably based
@@InkboxSoftware so swag
Lol I thought it was a sponsorblock skip, but then was confused why I didn't see the popup asking if I wanted to unskip lol
Demoscene members being chads as always. Crinkler is one of my favorite tools, so cursed yet so elegant.
Another demoscene-friendly language is JS. In my experience, it's quite versatile. Stuff like fitting a game that was 30,000 bytes into 3,081 bytes. People have also made 140-character JS programs back in the olden days of X/Twitter when there was a hard 140 character limit.
@@stgigamovement sounds more like code golf.
does the demoscene want or care about JS?
@@blarghblargh there's a section on Demozoo and Pouet for Javascript demoscene works, which I've used.
@@stgigamovement i mean you need the runtime to go with it lol so not really demosccene
@@stgigamovementThat’s code golf, you still need an interpreter for JS
Back in the day, we called it "Anne See" (like "antsy" with a silent "t"). not "Ay En Ess Eye"
it's the superior way to pronounce it. less syllables and easier to say. I've heard some people say ASCII one letter at a time, one of the worst experiences of my life
@@yourcomputerhasdied you poor soul. Stay strong!
@@yourcomputerhasdied Laugh out loud, what a waste of time.
In fairness, he might have never heard anyone say it 🤷🏻♂️
I pronounce it "ayensi"
for anyone wondering, the character at 4:06 is "嗯", which (according to Wiktionary) is a Chinese character that can mean doubt, curiosity, surprise, agreement, "pledge", or just onomatopoeia for groaning.
So in general the reaction of the programmer after the program starts and displays the message.
All of that.
1:26 the background music is the “if there is no CCP, there is no new China” wow very patriotic
2:55 - Who knew Plato was a Linux fan?
You brought me back in time to my computer science days, making games on console applications.
.
Those were the simplest of days.
.
Thank God I still have them.
.
.
Back in the days of DOS, we (Crescent Software) sold a library for BASIC (called PDQ) which would allow for the quick compilation & linking of EXEs less then 512 Bytes. Why 'so big' (LOL)? Because that was the size of *ONE* sector on a *FLOPPY DISK*. Anything less was moot.
It could also create TSR (Terminate & Stay Resident) programs, which were available with a keystroke or interrupt. And we could shove them up into Extended, or of you remember it, "Expanded" memory.
Since this was in the days prior to math coprocessors, I wrote the 8087 emulator for the product.
Thanks for the nostalgic trip down memory lane.
did it compile basic into a more native program instead of the more usual of adding the basic intepreter and the bas itself into an exe?
I remember trying to coax borland c/c++ 3.1 into compiling as small executable as possible (I think I was like 11 back then). i think i got a 'flame' effect down to 8kb or so, which was still huge for what it was(just couple of loops really and calling bios to change the video mode and make the pallette).
@@lasskinn474 LOL! Yeah, these were genuine EXE files. The BASIC compilers BC 6 & BC 7 Professional were actually pretty good compilers for the time. BASIC code was lexed into 8086, 80286, or 80386 instructions. It even had math coprocessor options. statements like Print, Play, Open, Locate, REDIM, Strings, etc., generated calls into the runtime libraries just like C functions.
We re-wrote the runtimes from *SCRATCH* in hand tuned assembly language (MASM). Because our libraries were so granular, the linker was able to bring in just the bare bones set of library routines that the programmer actually invoked. Nothing was just along for the ride.
Since it was a library we were able to add extra functionality, such as the TSR capabilities, interrupt handlers, function pointers, pretty much anything a C coder could do, we put that power in the hands of the BASIC programmer.
Then came Visual Basic for Windows. But that's a story for another day.
@@PaulPassarelli that's pretty cool, would've been cool to get hold of back in the day.
Just as a comment here, you don't want to do things like cmp al, 0. The reason is simple. You're comparing an immediate value and that value has to be stored in the instruction. Since you're only interested in the flags, you should test al, al whereby the assembly doesn't need to store the immediate value and it will set the flags for you to branch against.
Thanks, I had forgot about TEST, looks like I can shave off a few more bytes
@@InkboxSoftware Well when you've done as much assembly as I have and enter 4k/64k demo competitions at Revision and other parties, every byte does indeed matter!
Good video though.
Do not be embarrassed by your mistakes. Nothing can teach us better than our understanding of them. This is one of the best ways of self-education.
Maybe an interesting honorable mention: "fr-016c: fuenf (in your face)" by the demo group Farbrausch. It's a COM file that just produces an endless stream of garbage in the terminal, but because sometimes includes ascii bell symbols, it's technically an application with visuals and audio in 5 bytes.
(It also crashes the terminal window)
it's been about 10 years since I last wrote any code, and about 12 years since I worked with processor instructions. This video made me remember that I loved this before my professors at college made me hate this with my soul 😅
11:59 I laughed more than i should have on this part lol. Amazing video
I once made an executable a mere 20 bytes in size.
It ran in DOS, was a .com, not a .exe. It's use was to set text resolution to 120x50 characters (if I remember correctly).
Someone mention MattKC his QR code game could get an upgrade
that 嗯 got me laughing so hard in the binary selection part
Over they years, Windows developers give zero consideration of the size of their programs. I'm glad to see someone demonstrating it is possible to write functionally interesting programs in the
2:53 Subliminal ad placement
2:51
*fun fact:* Registers: EDI, ESP, EBI and EFL could be used to power a RNG with some multiplications and/or additions. Some of them change to a pretty unique value only at start-up and then increment as the program is running, but others could be used to get multiple equally dispersed pseudo-random numbers.
I know that 4k is comparatively huge to these tiny minimal-size exes, but I've seen some pretty cool demoscene exes that are only 4kb.
EXE doesn't actually indicate a PE file.
It may contain an early MS-DOS MZ, an MS-DOS 4.0 New Executable or a modern Portable Executable.
and also the LE and LX from Windows VxD services and Microsoft OS/2 2.0
4:22 ASCII art of a shrub??
Yeah wtf, why is there a shrub
I rember hearing this exact same fact somewhere else and that it exists and i think that there is a reason tl it but i dont rember
Maybe an easter egg ???? Or testing idk
0:26 sneaking in your mailing list is a genius move lol
There used to be a full chess game for the Sinclair ZX81 (Timex 1000) which fitted in the 1k (1024 bytes) of static RAM which came with the system, from which some was used for the screen....
In practical terms, the smallest .exe file you can have is the cluster size the FAT uses on the storage device you're using. Assuming you saved your work that is.
15:09 I braced myself for a Rickroll, but was glad that that wasn't the case.
The 2GB limit for RAM with 32 bit systems was a massive headache until I realised using GCC I was compiling for 32 bit OS's, when my OS was 64 bit. Changed that, and solved
watching videos like these gives me such a desire to c r e a t e something in assembly
but, yknow, then I remember, oh yea, I hate coding in assembly
I dived into bitmaps years ago and found unseen riches, unused digits at the ends of non 32bit arrays. nice work . we all need to keep kicking the tyres.
I mean, you can make a smaller executable for Windows by taking advantage of the NTVDM and making a COM-file
Yeah, you can get it down to one byte, a single RET instruction.
Wouldn't that be a 16-bit x86 file? Which wouldn't run on 64-bit Windows versions...
@@shinyhappyrem8728 Sure, but 32-bit Windows is still supported for about a year more
I never expected to see a video game that you can print on a single sheet of paper.
He who conquers others is strong; He who conquers himself is mighty.
Just discovered your channel. At first wasn't sure if I was watching clickbait. Then at 1:00 - 1:15 I knew you were the real deal!
Just a heads up, the link in description for the diagram at 1:00 is not the original source, nor does it link to the original source. The author is 'corkami' (look up 'corkami pe101')
I've been working on my own compiler, and diagrams from that person have been a huge help
Thank you for bringing that up, I've corrected it to reflect the original author. It it a great chart.
Writing small programs under an OS is fun. Try writing a bootable game :) People managed to pack whole simulations or games into the bootsector of a floppy. A sector has 512 bytes but the last 2 bytes are required or the BIOS won't recognise it as bootable (55AA in hex). Though the bootsector (depending on the medium) also contains some volume information. That only leaves a little bit left for the program. People have written bootsector games or simulations and still have the disk to work properly under DOS or windows. Also there are no sys calls to any OS function since we ARE the OS if we boot from this medium. All we have are the BIOS interrupts and direct memory access. The BIOS only loads the first sector (boot sector) of the medium which usually just contains code to load the actual OS boot loader from other sectors.
back in the day booter games on dos machines were not that common but there were a bunch of them. if you had only 1 floppy drive you didn't need to swap from dos, you didn't need to worry if the customers config.sys was weird etc and could have more confidence in whatever tricks you used to make it harder to copy the floppy.
i think that there were those titles is one reason for why booting from floppy was default for so long on pc's so you couldn't leave a floppy in the drive.
Yeah but the BIOS _is_ an operating system. A very primitive one, but an operating system nonetheless. And a BIOS interrupt call is only two bytes. There's a plethora of them; printing text to the screen, loading things from the disk, setting timers, etc. When you programmed on DOS you made use of some them as well. Because DOS ran in real mode, it could share the interrupt table with the BIOS, so some interrupts where DOS system calls, while others were handled by the BIOS.
My game is 53 KB, runs even before/without Windows directly when the PC turn on power, and funny enough, it has the MZ header! I am quite happy with that 😇
I looked at it, great video, the MZ header is used by the bootsector, very clever!
thats quite extreme compression you got. blew my mind!
Fun fact: I learned this from a Pascal 6 Book a friend read to me, at leas on MS-DOS it was 1 byte, the ╣B9h or Char 185, in a .exe file, which is the x86 Instruction RetF: Return Far, full qualified address: SEG:OFFSET is used to return to the Operating System.
Great concise analysis, thank you
ANSI you can just say "an-see"
Un-see🙈
I've got ANSI in my pantsi
This is actually just amazing, amazing work!
11:00 none of those characters are a part of ASCII... 😄
If memory serves, isn't this CP437?
@@timseguine2To my knolwedge with ANSI art, yes
The "@" sign isn't part of ASCII-7?
@ is ASCII at least
Decades ago, I came across an old DOS executable with just 64 bytes (!!), that produced a pretty impressive flight through a rotating cylindrical tunnel. It even had a simple blocky texture.
I might still be able to find it, including the assembler code. It might even still run in Windows' CMD, even though it was a ".COM" executable, which had barely any header in contrast to ".exe" executables.
Cool. Next time do a boot sector program so that way you don’t need the pesky bloat of an OS to execute.
In my minimal work with exe compression in the demoscene I verified that if the PE header offset at 0x3C points to a value smaller than 0x40 then the microsoft signing tool will not recognize the image as an executable, which may be an indication towards future limitations. The space between 0x02 and 0x3C can still be used freely though... We use it to put a small vanity ASCII art header in there for our 64k intros ;)
That comment about every Windows executable having the same DOS header isn't true. You could specify an actual MS-DOS application as the header and create a dual-mode application. The same .EXE when run under DOS would run only the DOS "header" portion, while when run from Windows would skip over the DOS header and execute the Windows code.
Many applications used to do this back in the day - and I've even created a few myself!
That's true, and that's why it has an MS-DOS header in the first place. It's just nowadays your standard compiler will fill in the string "This program cannot be run in DOS mode" automatically.
Older compiler uses slightly different string "This program requires Microsoft Windows".
@@rashidisw And certain versions of Delphi had a DOS stub that printed "This program must be run under Win32."
This is hands down the best channel on youtube right now. Interesting and concise!
5:15 the what now?
Essential code, the ASCII Art Of A Shrub (TM) stabilizes the program, protects against any form of cyberattack, and automatically generates cryptocurrency. It's the sequel to Sun Tzu's The Art Of War, even being more practical on the field of battle.
Markus Persson, in a
In DOS, the smallest EXE that will do something is 2 bytes. If an EXE contains "@e" it will reboot the computer when run.
That sounds more like a COM file. EXE files *must* start with "MZ" to be valid; and then also contain a header (very similar to the one used by EXE files for Windows).
COM files, on the other hand, were literally just something DOS loaded into a page of memory and then jumped to the first byte of. The smallest valid COM file is 1 byte, the encoding of the "RET" instruction (hex code C3). That would just immediately terminate and return control to DOS.
Last time I used DOS/Windows (c. late 1994), you could rename a .COM file to .EXE (and vice versa) with no effect on how it would run.
@@pdr. Back then they do checks for header within the file content instead of fully reliant on file extension.
@@Drysart i know.. but I didn't say it was valid :D I discovered this as a kid who didn't have any compilers, but desperatly wanted to make my own exe. So, as all kids do, I just renamed a BAT to EXE to see what would happen 🤔 the first line was... .@echo off 😀 and the result was that it would restart when I run it. I progressively deleted stuff until I was left with just @e and it would still do the same.
A DOS executable com file with a nop instructions = 1 byte file size.😂
17:41 This is close to 1.99 Gigabytes for those who don't know.
linking an extra dll just for a rng is costing a lot of space. Ok not really a lot but you can do better. It would be smaller to implement a LFSR rng using a seed value from whatever source you can get that you're already have access to. A Console handle is probably good enough to use as a seed for a LFSR. The Handle values are likely going to be unique every time you run the exe though not guaranteed, the chances you'd get repeated Handle values are extremely unlikely. Or you can just go straight to using a cpu random instruction but you lose compatibility with old cpus.
15:59 theoretically (based on size alone) it could run on the Apollo guidance computer. Congrats!
nice mix of humor and knowledge, good stuff
Getting into low-level programming is like trying to get what's going on in a soap opera that began 3 decades before you were born
would you save some space if you used RDRAND or RDSEED instead for the RNG?
Sometimes by losing a battle you find a new way to win the war.
Silence bot
@@InkboxSoftware LMAO
We really need Dave on this one
Yes, let's bring in the tone-deaf boomer millionaire. He'll solve our problems.
@@graealex Well, he kinda knows Windows, so he may be a good fit to explain Windows executables
This is not some esoteric boomer multimillionaire knowledge @@Saru-Dono
@@Saru-Dono He knows task manager. There are far more competent people still around in regards to WINAPI.
@@graealexhe developed windows... he made fucking task manager and zip... *_he probably knows a lot about executable files._*
8:14 you just offended a whole generation of people that played DOS games on Windows 9x lol
5:17 Okay, but WHY? WHYYYY in the world would Windows put a shrub there? What purpose does this serve?
what man, you ain't shrubbin? it's the new craze!
Even an exe file whose size has been reduced in this way will become a 1GB exe file with just one touch from a designer.
what is the smallest possible elf64?
(btw the result turns out to be 45 bytes)
it seems youtube deleted my reply so let's try it again...
even though it's about elf32, I'd highly recommend the famous essay "A Whirlwind Tutorial on Creating Really Teensy ELF Executables for Linux"
i think using a header someone i know made and assembling asm -> elf64 it was 40 or smth
@@aspectreishauntingeurope that's way smaller than I expected!
Back in windows xp era I remember having a program that has no sections, only a header that also contains some code that restarts windows, IIRC its size was 97 bytes
0:40 +1000 soical credit for the music
I thought no one else noticed
goof
no it’s -9999999999 social credit. it’s Taiwan’s music
Not sure if you already know about this, but you might be interested in Cosmopolitan libc, a project to compile c code to a polyglot binary running natively on Linux + Mac + Windows + FreeBSD + OpenBSD + NetBSD and even in BIOS
But can it run Doom?
0:14
I use DUMMY,EXE files for my GPU's underclocking profiles.
I make sure to TRIPPLE protect them, so only admin accounts can edit or change the exe.
even the folder which they are located in is locked/write protected. and everything is in read only,
even by default they are hidden files. so regular scrapers will overlook them
0:29 is that a CCP song lol
Yes. It's called "without the communist party, there will be no new china" or something like that
Congratulations on making the world's most basic bullet hell game, be sure to watch out for ZUN knocking on your door at 3 AM. Lock the door.
Wow, it took until 2016 for windows to support ANSI escape sequences? Yikes!
Books contain interesting exe files for copying and running in w3.1. These files mean, that expensive companies supported offices with such files. One is such, you can have your company name in w3.1 help menu, or just your signature. Other is such, you can use colour tiles as your paint in paintbrush, so a drawing can look like a colourful snake.
Some others like radio buttons, etc. All show that hard wrking companies supported the Ms clients.
It would be cool to introduce these secret programs in the new versions of Windows.
...do you have a homemade smartwatch?
16:37 lol i thought you were just going to say the results; animating the binary search was super funny
You could fit the entirety of Minecraft into a exe lol
yes? most games can be fit in an exe. in fact that's how games work. they are exes.
@@thewhitefalcon8539 you dont get my point...
@@thewhitefalcon8539 He refers to all the DLLs, libraries, files, external to the Minecraft JAR or EXE file depending on the game edition (Java or Bedrock)
@@thewhitefalcon8539he means the entire game, not just the "executable" portion. (using quotation marks since it's java which is a VM and not a real executable)
Yes, but saves
Meanwhile, bare metal developers:
"What do you mean you need more than ONE WORD?"
正津津有味看视频的时候突然蹦出来一个汉字，给我整恍惚了
看起来老哥你是真喜欢简中，关注了，加油！
useful video, enjoyed it 💚
keep it up
Next step: waste money to make the smallest digitally signed executable
Goodbye Rich Header, Goodbye DOS Stub,
Goodbye Debug Table and the ASCII Art of a Shrub.
i see what you were doing there :)