this post was submitted on 04 Jun 2025
134 points (97.9% liked)

Programmer Humor

36167 readers
372 users here now

Post funny things about programming here! (Or just rant about your favourite programming language.)

Rules:

founded 5 years ago
MODERATORS
 
you are viewing a single comment's thread
view the rest of the comments
[–] [email protected] 1 points 1 day ago* (last edited 1 day ago) (10 children)

I thank god every day people who make these comics are too stupid to open gcc's sha1.c because they'd see shit like:

#define M(I) ( tm =   x[I&0x0f] ^ x[(I-14)&0x0f] \
		    ^ x[(I-8)&0x0f] ^ x[(I-3)&0x0f] \
	       , (x[I&0x0f] = rol(tm, 1)) )

#define R(A,B,C,D,E,F,K,M)  do { E += rol( A, 5 )     \
				      + F( B, C, D )  \
				      + K	      \
				      + M;	      \
				 B = rol( B, 30 );    \
			       } while(0)

      R( a, b, c, d, e, F1, K1, x[ 0] );
      R( e, a, b, c, d, F1, K1, x[ 1] );
      R( d, e, a, b, c, F1, K1, x[ 2] );
      R( c, d, e, a, b, F1, K1, x[ 3] );
      R( b, c, d, e, a, F1, K1, x[ 4] );
      R( a, b, c, d, e, F1, K1, x[ 5] );
      R( e, a, b, c, d, F1, K1, x[ 6] );
      R( d, e, a, b, c, F1, K1, x[ 7] );
      R( c, d, e, a, b, F1, K1, x[ 8] );
      R( b, c, d, e, a, F1, K1, x[ 9] );
      R( a, b, c, d, e, F1, K1, x[10] );
      R( e, a, b, c, d, F1, K1, x[11] );
      R( d, e, a, b, c, F1, K1, x[12] );
      R( c, d, e, a, b, F1, K1, x[13] );
      R( b, c, d, e, a, F1, K1, x[14] );
      R( a, b, c, d, e, F1, K1, x[15] );
      R( dee, dee, dee, baa, dee, F1, K1, x[16] );
      R( bee, do, do, dee, baa, F1, K1, x[17] );
      R( dee, bee, do, dee, dee, F1, K1, x[18] );
      R( dee, dee, dee, ba, dee, F1, K1, x[19] );
      R( d, a, y, d, o, F1, K1, x[20] );

And think, yeah this is real programming. Remember the difference between being smart and incredibly stupid is what language you write it in. Using seemingly nonsensical coercion and operator overloaded is cringe, making your own nonsensical coercion and operator overloads is based.

That's why you should never subtract things from 0x5F3759DF in any language other than C.

[–] [email protected] 0 points 1 day ago* (last edited 1 day ago) (9 children)

sha1.c

Yeah, a hash function actually just looks like that intrinsically, though. Being impenetrable is the point.

[–] [email protected] 3 points 1 day ago* (last edited 1 day ago) (3 children)

There are plenty of sha1 implementations that are more readable and sensible and less readable and sensible. This portion is simply an manually unrolled loop (lmao these gcc nerds haven't even heard of Gentoo) of the hash chunk computation rounds. Hash functions aren't "impenetrable" they're just math. You can write math programmatically in a way that explains the math.

The point of this post is actually things like x[(I-3)&0x0f]. It's entirely the same concept as coercion to manipulate index values this way. What's funny is that void pointer math, function pointer math, void pointers and function pointers in general are typically seen as "beyond the pale" for whatever reason.

Beyond that if you know C you know why this is written this way with the parens. It's because C has fucked up order of operations. For example a + b == 7 is literally "does adding a + b equal 7", but if you write a & b == 7 you would think it means "does a AND b equal 7", but you'd be wrong. It actually means does b equal 7 AND a.

Furthermore a & (b ==7) makes no sense because b == 7 is a boolean value. Bitwise ANDing a boolean value should not work because the width of the boolean is 1 bit and the width of the int is 8 bits. ANDing should fail because there's 7 void bits between the two types. However the standard coerces booleans in these cases to fit the full width, coercing the void bits to 0's to make bitwise ANDing make sense.

Beyond that asking what the memory size of a variable in C is a fools errand because the real answer is "it depends" and "it also depends if someone decided to ignore what it typically depends on (compiler and platform) with some preprocessor fun". Remember how I said "void pointers" are beyond the pale? Yeah the typical "why" of that is because they don't have a known size, but remember the size of something for C is "it depends". 🤷

Almost every language has idiosyncratic stuff like this, but some let you make up your own shit on top of that. These kinda low hanging fruit jokes are just people virtue signaling their nerddom (JS bad am rite guis, use a real language like C), when in reality this stuff is everywhere in imperative languages and typically doesn't matter too much in practice. This isn't even getting into idiosyncracies based on how computers understand numbers which is what subtracting from 0x5F3759DF (fast inverse square root) references.

[–] [email protected] 0 points 1 day ago* (last edited 1 day ago) (1 children)

Hash functions aren’t “impenetrable” they’re just math.

I mean, they're both. At a high level it's math, but the individual operations are carefully designed to resist any symbolic manipulation (aka thought) - because that's one step away from an attack.

The point of this post is actually things like x[(I-3)&0x0f]. It’s entirely the same concept as coercion to manipulate index values this way. What’s funny is that void pointer math, function pointer math, void pointers and function pointers in general are typically seen as “beyond the pale” for whatever reason.

If it compiles pretty directly to register pointer math that way, I think it could be justified here. I can't say if it does, or if an alternate approach would too, though.

This portion is simply an manually unrolled loop

Lol, I didn't notice it's a perfect shift. Yeah, that could theoretically be done better. Presumably the justification is because it's a leaf function, and it's hard to guarantee every compilation will unroll it properly.

(lmao these gcc nerds haven’t even heard of Gentoo)

The flag -O3 exists. Or just -funroll-loops. You shouldn't even need -funroll-all-loops in this case, since hashes have a fixed size.

I sound way more competent with the flags than I am here, haha. Does Gentoo use an alternate compiler by default?

Beyond that asking what the memory size of a variable in C is a fools errand because the real answer is “it depends” and “it also depends if someone decided to ignore what it typically depends on (compiler and platform) with some preprocessor fun”.

As I understand it, that's pretty unavoidable if you want C to both compile onto multiple processors and work at a high level the same way on all of them. JavaScript catches shit for doing funny things purely because it was hastily built.

Ditto for fast inverse square root. It's absolutely cursed, but when you're at a certain low level you can't afford pretty anymore. You're feeling the constraint of limited die space and manufacturing steps not too far down the layers of abstraction. Browser scripting, on the other hand, is not low-level.

[–] [email protected] 1 points 1 day ago* (last edited 1 day ago) (1 children)

The flag -O3 exists. Or just -funroll-loops. You shouldn’t even need -funroll-all-loops in this case, since hashes have a fixed size.

I sound way more competent with the flags than I am here, haha. Does Gentoo use an alternate compiler by default?

This is in reference to an ancient linux meme cw: slur

[–] [email protected] 0 points 5 hours ago (1 children)

Lol, good to know the guys with souped up shitty cars have been around since the early internet.

Now I'm super curious about Gentoo and Portage. You don't hear so much about compiling your own stuff anymore (probably because there's less architectures around).

[–] [email protected] 1 points 5 hours ago* (last edited 5 hours ago) (1 children)

Now I’m super curious about Gentoo and Portage. You don’t hear so much about compiling your own stuff anymore (probably because there’s less architectures around).

"Nobody" runs Gentoo anymore because most distros have taken the 80% optimizations you can do and just mainlined them. This was back in 2000's where some distros weren't even by default compiling with -O2. Gentoo usage just proved out that the underlying code was effectively -O3 safe in the 80% case and nobody was sneakily relying on C/C++ vagaries.

I have much less time to tinker, but my favorite new bag is Fedora Atomic (currently using Bazzite on my main desktop). I'm incredibly interested in figuring out Nix though, but I haven't had the time. Immutable distros are honestly something incredibly useful for both power users and normies. The main issues I've had with Fedora Atomic have really been around vagueness in the "standard" but they're still figuring things out as far as I can tell.

[–] [email protected] 1 points 4 hours ago

Nix is cool. My next daily driver is probably going to be Qubes, which goes in a whole other direction. I bet you can run Nix within Qubes, though.

Guix and it's accompanying system are interesting to me for basically being Nix with Portage-style user compilation. The system is also GNU Hurd compatible and has a different approach to organising files, among other things.

[–] [email protected] 1 points 1 day ago* (last edited 1 day ago)

Precisely. This exact situation from the comic would happen in Java, too, except it complains when you subtract an int from a string. JavaScript was merely designed to minimize errors (since a web browser takes the place of the compiler, and random strangers visiting your site shouldn't get interpretation errors) so instead of throwing up it just does its best at interpreting what you meant.

[–] [email protected] 3 points 1 day ago (1 children)

b == 7 is a boolean value

Citation needed. I'm pretty sure it's an int.

[–] [email protected] 2 points 1 day ago* (last edited 1 day ago)

Yeah you're actually right, it's an int in C since K&R C didn't have bool, however it's a bool in C++. I forget my standards sometimes, because like I said this doesn't really matter. It's just nerd trivia.

https://en.cppreference.com/w/cpp/types/type_info/operator_cmp.html

load more comments (5 replies)
load more comments (5 replies)