Allocated 4.21GB and did not loose a single byte. This day isn't going to get better.

When valgrind (C Memory allocation error detection tool) aborts due to a memory allocation error...

When you Valgrind your program for the first time for memory leaks and get "85000127 allocs, 85000127 deallocs, no memory leaks possible"

Achievement unlocked: malloc failed
😨

(The system wasn't out of memory, I was just an idiot and allocated size*sizeof(int) to an int**)

I'd like to thank myself for this delightful exercise in debugging, the GNU debugger, Julian Seward and the rest of the valgrind team for providing the necessary tools.

But most of all, I'd like that three hours of my life back 😩

New developers. Tip: There is no silver bullet.

If you like Python, please understand GIL's behavior before making a system that handles thousands of requests.

If you like Java, know that "Write once, run anywhere" is a fallacy. Even application servers don't like the same WAR.

If you like PHP, understand the life cycle of a request before connecting to the database from all corners.

If you like C#, don't make it a small command-line application that will be used on FreeBSD.

If you like C, meet valgrind.

If you like C++, templates are cool, but don't overdo it. And take the opportunity to meet valgrind.

Never use the same tool to do everything. Elect the language and framework for the given need with rationality.

Every time I see a "Java Man", a "C++ Chad" or anything like that, it comes to mind that if he were a carpenter, he would be tightening screws with hammers.

Every lock-in is bad.

I am working on an open source game project, and the most common way to draw things is using a class named ManagedSurface. The class is otherwise awesome, but it has a method called getBasePtr(x, y), which gives you a pointer to the requested coordinates. Fair enough (this is C++ without STL by the way).

But WHY THE HELL CAN I REQUEST ANY POINTER THAT I WANT, EVEN IF IT'S OUTSIDE THE SURFACE? Other cointainers have sanity checks, asserts and such, and the surface KEEPS TRACK OF IT'S WIDTH AND HEIGHT.

WAS IT SO FUCKING HARD TO ADD assert(x <= w); assert(y <= h);???

I spent 3 days on valgrind trying to find a heap corruption that manifested at random points in the code.

FUUUUCK!

On the bright side, I learned how to use valgrind (which is awesomely awesome).

Even Valgrind hates me now!

I love how my University CSSE courses make it hard for Windows users.
Student: Is it possible use VALGRIND in Windows CMD?
prof: I don't know, I only use Linux command line. Next question.

All programs must compile and run on Linux using g++. So everyone using windows has strip the IDE bloat and transfer the project files over to the remote Linux lab every single time they compile and run.

Benefits of being a Linux user!

Okay, friends. I have a challenge. Who can make the sneakiest memory leak example? I need to stump a class of students with something only valgrind can find and I'm having a hard time.

Just found out that the CI for valgrind on my personal project wasn't working properly, I'm leaking memory all over the fucking place...

Fuck this shit

Valgrind is awesome. Today I fixed a lot of memory leakage / overflow bugs thanks to it. An guess what? Now, everything works!

Is there something you find genuinely cool and would recommend ? Some webpage, program, OS, library or anything ?

I mean hey. There are SO MANY reaaaally cool things I didn't know until last few months.. Things I'd be so grateful for if I knew them earlier. I'll list some of them and I just know you have few of yours too. Feel free to educate the rest!

Processing - Program so fun to code in + CodingTrain(YTB channel)
Microcorruption.com - so freaking awesome if you wanna learn hacking / assembly (not x86 necessarily)
LiveOverflow - cool hacking channel
Radare - cool cmd Linux disassembler
vim-adventures.com - LEARN VIM (not just how to quit it) LITERALLY by playing a game!!!!!!!!!!!!!!
slashdot - stay updated , like really
"BEST-WEBSITES-A-PROGRAMMER-SHOULD-VISIT" - GUYS THIS! Sorry for caps but search this on GitHub and you will fucking die of happiness of how freaking useful links there are and no bullshit to dig through , just pure awesomeness. REALLY
HandBrake - Top media converter without bullshit and bloat stuff in it
Calibre - Best eBook management software capable of literally everything ebooks related. Kindle is a bloated joke compared to this
QubesOS - You know you can have every OS running at once - you have a Linux but are playing win games. Yup. It's there. Free
Computerphile - You all know it, it's just for completeness
Khan Academy - Same
VulnHub - download vulnerable VMs and hack them, or learn by reading writeup on how to do it!
Valgrind - MUST HAVE for C/C++ programmers
Computer Science crash course videos

That's all I can think of from top of my head but hey, there's more to it so definitely add your 2 cents!

Last thing, if nothing, just check the websites on GitHub, that's lifechanger

Looking forward to see some cool links & recommendations!

Other peoples' code... (in C++)
I am finding what some people consider good code is not as described. I found a class that provides strings. Great it gives me paths and stuff. I incorporated it in a new project.

segfaults

Hmmm, it must have an init function... It does, but not in the class. It has a friended init function:
friend init_function(). If this function is not created and called external to the class then the class will segfault...

okay...

I implement this. I use code from another project that implements this correctly. The friend class allows the private constructor to be called to create the main instance of the class. So its a fucking cryptic ass singleton. I look at this class. It uses a macro to decide what to function call in the class. The class already has function names for each call it needs to make. The class is literally a string lookup table. I vow to redo this shitty code, someday...

I start to wonder what other fragile code I will find. Not long later I keep getting errors on malloc. Like any malloc that is called results in a segfault. The malloc is not at fault though. I run valgrind and find a websocket library is returning an object a different size than the header file describes.

WTF...

Somebody has left an old ass highly modified definition of the websocket header in a location in that I include headers (partly my fault). I eliminate that from my include path. All is well, everything behaves. I will be making sure this fucking header is not used and it is going to die. Wasted a bunch of time.

Lessons learned: some code is just fucked and don't leave old ass shit you tried laying around.

That moment when you realize that the bug (my fault, oc) was a struct declared just outside a function instead of inside it... And it "only" took you >4h of GDB and valgrind.
I really C/GTK+ when it comes to debug it...

Valgrind! What are you?!!

Knowledge of C is complete only when you come out of Valgrind with 0 errors in 0 contexts.
Definitely lost I am.

In case you are confused, reaffirming that this is a rant and not a statement.

VP last week: No you can't have that equipment that fakes out the gps. It's very expensive and not in the budget. Just run valgrind and push your code so we can deploy it.
VP in today's all hands: Guys, if you need test equipment, come ask for it and we'll get you what you need. Not having equipment is not a valid excuse for skipping integration testing.

When C devoper creates a memory leak standard practice valgrind time.
When a webdeveloper creates a memory leak is the day they start to learn javascript.

GLFW is the cleanest, well documented, most convinient API for creating and handling windows in Linux and Windows I've ever used.
The only thing that bugs me is that valgrind detects memory leaks on it.

My university's IT department can't even install debuggers on the computers, so if we're on linux and need to debug something, we need to save the code to an usb stick, reboot to windows, boot a VM and install valgrind there (or manually install the needed .deb files, which ends up being even more of a hassle than just rebooting)

Okay, so I had an object consisting of tables (basically classes) and structs (classes with only scalars as their properties).

I was about to serialize the object with vectors of classes and structs and wrote some nice tests for it wondering why they fail to validate the data after deserialization and why I only got garbage for the vectors of structs whilst the tables worked just fine.

Turns out there is an undocumented function called CreateVectorOfStructs which shall be used for structs instead of the regular CreateVector ...

There go three hours of blaming memory issues and running Valgrind over and over again ...

Thank you Valgrind. That is so much of errors to mock my inability to handle them.

Spend like 3 weeks in mem-checking with valgrind and ASAN, because there seemed to be some leaks. So painful and scary. You loose all confidence in your software, the checking tool, your own sanity.
Some spurious result prevailed, could only move it around. Boss could not reproduce the problem on his machine; Ubuntu 18 with GCC 7, mine was Debian 9 with GCC 6, so I tried older Ubuntu with GCC 5. Also no problem.
Fuck it, I'm switching to clang.