It finally hit me the other day.
I'm working on an IoT project for a late-stage ALS patient. The setup is that he has a tablet he controls with his eye movements, and he wants to be able to control furnishings in his room without relying on anyone else.
I set up a socket connection between his tablet and the Raspberry Pi. From there it was a simple matter of using GPIO to turn a lamp or fan on or off. I did the whole thing in C, even the socket programming on the Pi.
As I was finishing up the main control of the program on the Pi I realized that I need to be more certain of this than anything I've ever done before.
If something breaks, the client may be forced to go days without being able to turn his room light on, or his fan off.
Understand he is totally trapped in his own body so it's not like he can simply turn the fan off. The nursing staff are not particularly helpful and his wife is tied up a lot with work and their two small children so she can't spend all day every day doting on him.
Think of how annoying it is when you're trying to sleep and someone turns the light on in your room; now imagine you can't turn it off yourself, and it would take you about twenty minutes to tell someone to turn it off -- that is once you get their attention, again without being able to move any part of your body except your eyes.
As programmers and devs, it's a skill to do thorough testing and iron-out all the bugs. It is an entirely different experience when your client will be depending on what you're doing to drastically improve his quality of life, by being able to control his comfort level directly without relying on others -- that is, to do the simplest of tasks that we all take for granted.
Giving this man some independence back to his life is a huge honor; however, it carries the burden of knowing that I need to be damned confident in what I am doing, and that I have designed the system to recover from any catastrophe as quickly as possible.

In case you were wondering how I did it all: The Pi launches a wrapper for the socket connection on boot.
The wrapper launches the actual socket connection in a child process, then waits for it to exit. When the socket connection exits, the wrapper analyzes the cause for the exit.
If the socket connection exited safely -- by passing a special command from the tablet to the Pi -- then the wrapper exits the main function, which allows updating the Pi. If the socket connection exited unexpectedly, then the Pi reboots automatically -- which is the fastest way to return functionality and to safeguard against any resource leaks.
The socket program itself launches its own child process, which is an executable on the Pi. The data sent by the tablet is the name of the executable on the Pi. This allows a dynamic number of programs that can be controlled from the tablet, without having to reprogram the Pi, except for loding the executable onto it. If this child of the socket program fails, it will not disrupt its parent process, which is the socket program itself.

Underlying pure C library and NodeJS wrapper

Binding a C-library for a higher-level language:
10% Actually binding Functions and Types to their higher-level equivalents
90% Writing a fuckton of wrapper-code to secure that dangling pile of C that could collapse at any time

¡rant|rant

Nice to do some refactoring of the whole data access layer of our core logistics software, let me tell an story.

The project is around 80k lines of code, with a lot of integrations with an ERP system and an sql database.

The ERP system is old, shitty api for it also, only static methods through an wrapper to an c++ library

imagine an order table.

To access an order, you would first need to open the database by calling Api.Open(...file paths) (yes, it's an fucking flat file type database)

Now the database is open, now you would open the orders table with method Api.Table(int tableId) and in return you would get an integer value, the pointer.

Now for the actual order. first you need to search for it by setting the search parameter to the column ID of the order number while checking all calls for some BS error code

Api.SetInt(int pointer, int column, int query Value)
Then call the find method.

Api.Find(int pointer)

Then to top this shitcake of an api of: if it doesn't find your shit it will use the "close enough" method of search.

And now to read a singe string 😑

First you will look in the outdated and incorrect documentation given to you from the devil himself and look for the column ID to find the length of the column.

Then you create a string variable with ALL FUCKING SPACES.

Now you call the Api.GetStr(int pointer, int column, ref string emptyString, int length)

Now you have passed your poor string to the api's demon orgy by reference.

Then some more BS error code checking.

Now you have read an string value 😀

Now keep in mind to repeat these steps for all 300+ columns in the order table.

News from the creators: SQL server? yes, sql is good so everything will be better?

Now imagine the poor developers that got tasked to convert this shitcake to use a MS SQL server, that they did.

Now I can honestly say that I found the best SQL server benchmark tool. This sucker creams out just above ~105K sql statements per second on peak and ~15K per second for 1.5 second to read an order. 1.5 second to read less than 4 fucking kilobytes!

Right at that moment I released that our software would grind to an fucking halt before even thinking about starting it. And that me & myself and I would be tasked to fix it.

4 months later and two weeks until functional beta, here I am. We created our own api with the SQL server 😀

And the outcome of all this...

Fixes bugs older than a year, Forces rewriting part of code base. Forces removal of dirty fixes. allows proper unit and integration testing and even database testing with snapshot feature.

The whole ERP system could be replaced with ~10 lines of code (provided same relational structure) on the application while adding it to our own API library.

Best part is probably the performance improvements 😀. Up to 4500 times faster and 60 times less memory usage also with only managed memory.

Wrote a c# wrapper around a library exposing the exact same functions.

I want to play a game with 2 mods, but the problem is you can only play with one of the mods at a time.
So I wrote a wrapper in C and now I can play with any number mod I want. 😎

Im building an android library that basically does file compression/ image compression etc its just a wrapper library around android's native c++ code.

I wonder how this is going to behave on different devices, as you all know when you do low level stuff, each manufacturer has his own way of doing things in their version of the Android OS.

Once i put this lib up on github please use the lib on your devices and tell me if you get any issues :D

It is great feeling, to leave company and leave all your crap code to others :D

500 lines bash generic wrapper to curl (just to catch and print errors, not just silently fail as most devs tell curl to do).

It was monster that used "function overload" and "subclasses" (based on dynamic source files). Also dynamically created inline AWK script to parse curl output. It kinda worked, but amount of high-level hacks I had to use was enormous.

Never use Bash when you do not have to. Even if you have experience with it. Others don't have it and will fail miserably trying to patch your code. Just leave bash for fast bridging between programs, leave python/java/c#/go or any other proper OOP language for a job. Please ?

The other day I was working with libpulse in a Qt project. Because I didn't like to deal with a C library from C++ code I copied wrapper code from a well-proven project. Suddenly the wildest compile errors occurred when I only tried to include a header from the new code. After hours of frustration towards me, C++, the compiler and everything involved I discovered that one of the headers I copied had the same include guard as one of the existing classes.

I'm still not over how LINQ is defined as a thin wrapper over both IQueryable (which can be efficiently queried) and IEnumerable (which can only be iterated), but IQueryable extends IEnumerable, so if you execute one unserializable operation anywhere in a query issued to your database it'll merrily pump whatever temporary collection happens to reside on the boundary through the C# program to execute that call on each row and process the rest as an IEnumerable.

I needed to create a c# wrapper class for an activex element which has functions for sudoku.

Then create a wpf interface to play it.
So far so good. Ive had c# wpf, so no problem.

Finished that

"Make another interface in asp.net mvc"

I thought to myself, sure how difficult can asp.net be?

Very difficult apparently.

At least, if you want to make a grid which you can change per cell with an x and y.

I just went to bed after i gave up, when I suddenly got an idea. It ain't pretty, but just might work

I am making a GUI wrapper in C# for a CLI tool written in Python. Obvioisly, the python exe is launched with the Process class and the output streams are redirected so I can process the console output. The problem is that some of that output is only printed if sys.stdout.isatty() returns True. Is there a hack that would allow me to launch the process in a way that python thinks that there is a console/tty attached?

I really don't want to touch the python source files, because that would be a messy solution. I also don't want the process to spawn an actual console window.

I'm at a bit of a loose end here, I'll do my best to explain and I hope it all makes sense.

I'm trying to find a way to wrap a C++ Console App in a C# Class so I can use the C++ App as a library within the C# app rather than triggering the C++ exe and providing a command to it.

The reason why I wish to do it this way is for maintainability, so I can make changes to the C++ app without affecting the C# App.

I've been looking at tutorials and stack overflow to see if's its possible, but for someone with learning difficulties, I'm struggling to find the right path to take as I'm seeing conflicting info.

Any help would be greatly appreciated, Thanks in advance

I'm doing a project for uni in Omnet (C++ framework that should facilitate working with networks of queues, simulating and displaying statistics).

I needed to retrieve a random value from an exponential distribution, and the function to do so requires a random number generator as input. The framework has 2 implementations of the RNG and I picked the first one.

I spent 3 hours trying every possible thing, using both the exponential() function and its class wrapper (both provided by the framework), it was always returning 0 or NaN.

The RNG was spitting out values correctly, so I thought it was okay.

When I was almost ready to give up, I figured I could try and change to the second implementation of RNG, expecting nothing to change. And it fucking worked.

Zero reports on this behavior on Google, no apparent reason why it would work with one and not with the other when the two RNGs literally implement the same abstract class and spit out the same exact numbers... Just black magic...

Oh and cherry on top, it works with the raw function but not with the class wrapper on that same function... IF YOU GOTTA IMPLEMENT SOMETHING IN YOUR DAMN FRAMEWORK THAT DOESN'T WORK, FUCKING DON'T! 1 combination working out of 4 is not good! Or at least document it!

Sorry just had to share my pain

One of the worst practices in programming is misusing exceptions to send messages.

This from the node manual for example:

> fsPromises.access(path[, mode])

> fsPromises.access('/etc/passwd', fs.constants.R_OK | fs.constants.W_OK)
> .then(() => console.log('can access'))
> .catch(() => console.error('cannot access'));

I keep seeing people doing this and it's exceptionally bad API design, excusing the pun.

This spec makes assumptions that not being able to access something is an error condition.

This is a mistaken assumption. It should return either true or false unless a genuine IO exception occurred.

It's using an exception to return a result. This is commonly seen with booleans and things that may or may not exist (using an exception instead of null or undefined).

If it returned a boolean then it would be up to me whether or not to throw an exception. They could also add a wrapper such as requireAccess for consistent error exceptions.

If I want to check that a file isn't accessible, for example for security then I need to wrap what would be a simple if statement with try catch all over the place. If I turn on my debugger and try to track any throw exception then they are false positives everywhere.

If I want to check ten files and only fail if none of them are accessible then again this function isn't suited.

I see this everywhere although it coming from a major library is a bit sad.

This may be because the underlying libraries are C which is a bit funky with error handling, there's at least a reason to sometimes squash errors and results together (IE, optimisation). I suspect the exception is being used because under the hood error codes are also used and it's trying to use throwing an exception to give the different codes but doesn't exist and bad permissions might not be an error condition or one requiring an exception.

Yet this is still the bane of my existence. Bad error handling everywhere including the other way around (things that should always be errors being warnings), in legacy code it's horrendous.

rant.author != this

Christ people. This is just sh*t.

The conflict I get is due to stupid new gcc header file crap. But what
makes me upset is that the crap is for completely bogus reasons.

This is the old code in net/ipv6/ip6_output.c:

mtu -= hlen + sizeof(struct frag_hdr);

and this is the new "improved" code that uses fancy stuff that wants
magical built-in compiler support and has silly wrapper functions for
when it doesn't exist:

if (overflow_usub(mtu, hlen + sizeof(struct frag_hdr), &mtu) ||
mtu <= 7)
goto fail_toobig;

and anybody who thinks that the above is

(a) legible
(b) efficient (even with the magical compiler support)
(c) particularly safe

is just incompetent and out to lunch.

The above code is sh*t, and it generates shit code. It looks bad, and
there's no reason for it.

The code could *easily* have been done with just a single and
understandable conditional, and the compiler would actually have
generated better code, and the code would look better and more
understandable. Why is this not

if (mtu < hlen + sizeof(struct frag_hdr) + 8)
goto fail_toobig;
mtu -= hlen + sizeof(struct frag_hdr);

which is the same number of lines, doesn't use crazy helper functions
that nobody knows what they do, and is much more obvious what it
actually does.

I guarantee that the second more obvious version is easier to read and
understand. Does anybody really want to dispute this?

Really. Give me *one* reason why it was written in that idiotic way
with two different conditionals, and a shiny new nonstandard function
that wants particular compiler support to generate even half-way sane
code, and even then generates worse code? A shiny function that we
have never ever needed anywhere else, and that is just
compiler-masturbation.

And yes, you still could have overflow issues if the whole "hlen +
xyz" expression overflows, but quite frankly, the "overflow_usub()"
code had that too. So if you worry about that, then you damn well
didn't do the right thing to begin with.

So I really see no reason for this kind of complete idiotic crap.

Tell me why. Because I'm not pulling this kind of completely insane
stuff that generates conflicts at rc7 time, and that seems to have
absolutely no reason for being anm idiotic unreadable mess.

The code seems *designed* to use that new "overflow_usub()" code. It
seems to be an excuse to use that function.

And it's a f*cking bad excuse for that braindamage.

I'm sorry, but we don't add idiotic new interfaces like this for
idiotic new code like that.

Yes, yes, if this had stayed inside the network layer I would never
have noticed. But since I *did* notice, I really don't want to pull
this. In fact, I want to make it clear to *everybody* that code like
this is completely unacceptable. Anybody who thinks that code like
this is "safe" and "secure" because it uses fancy overflow detection
functions is so far out to lunch that it's not even funny. All this
kind of crap does is to make the code a unreadable mess with code that
no sane person will ever really understand what it actually does.

Get rid of it. And I don't *ever* want to see that shit again.

I'm currently creating a Qt Wrapper for libpulse by extracting it from an application. The mix of C and C++ code gives me crap and I really want to punch the guy who spread templates all over the place