C++ Exceptions for Smaller Firmware - Khalil Estell - CppCon 2024

[@AndreKostur]

This is going to be an impactful talk. Goes directly against the "common wisdom", and has a lot of evidence to support it!

[@danamyrocket]

Nice to see a working software engineer discussing pitfalls and challenges of embedded software

[@ruimvfigueira]

Very brave of you to tackle this topic and trying to improve the current state of things.

[@AxWarhawk]

Very interesting talk, thank you. I was quite happy with 1:12:23 mentioning the elephant in the room. In my opinion, this problem persists until compilers can provide programmers with guarantees that all possible exceptions are handled adequately. If such guarantees can be made with Exception Insights Tool, that's great, but I'd say, it needs to be part of the compiler's tool chain. One more thing regarding centralized error handling (1:02:39) when an exception is thrown at a "low-level" and centralized error handling is employed at some levels above that, one typically has to attach additional information to the exception while it's bubbling up the stack. Otherwise you'd just a get some generic "low-level" error in your "high-level" handling code. In general, the result-based approach does this by converting between different error types automatically. In Rust you commonly see low-level errors get wrapped into higher level errors while a result bubbles up the stack. With exceptions this means we'd need try-catch blocks where we re-throw in the catch block.

[@provod_]

This talk (and its earlier incarnation) is easily one of the best talks made in C++ community in the past several years. Brilliant stuff, amazing and immediately practically useful findings. And absolutely breakthrough performance in delivering these results!

[@colindewolfe3647]

As an embedded guy, thank you so much for this talk. It has backed up my suspicions, I've always thought the cost of testing for the unhappy path outweighs the cost of exceptions, but didn't know how to tackle what these actual costs are.

[@chaquator]

revolutionary talk, cant understate that. c++ needs less expensive exception handling for when you don't use all parts of it, this is great work towards this

[@archieatkinson7502]

Great talk. If expectations are good for Bjarne, they are good enough for me! I would love to see these changes to the expectation handling code added to ARM GCC.

[@UtahCppProgrammers]

What a great deep dive into exceptions!

[@N....]

I love that you're giving this talk at more conventions, keep up the great work! And I'm really looking forward to your other talks 👀

[@jeffersoncarpenter2]

Great talk! I'm sure C++ has different problems than Java, but knowing that Java has checked exceptions, I bet you can get quite far with the Exception Insights tool using static analysis.

[@Voltra_]

Amazing talk and speaker!

[@zach2good]

Interesting talk, fantastic speaker!

[@TheArech]

Superb talk! Thanks a ton!

[@kgnet8831]

Top Talk 👍

[@alextalker5979]

What a delivery! 🎉

[@VincentZalzal]

Great talk, excellent speaker, and very interesting results. Would have loved those x86-Windows results ;) I hope to watch more talks from Khalil!

[@DerHirni]

Amazing talk and amazing presenter! Learned a lot, exactly the kind of talk I like to see here.

[@alexeydmitrievich5970]

Damn, the exception insights tool sounds Rad

[@JackLilhammers]

Am I the only one that felt he was speaking too fast, then played it at 0.75x and it seemed normal?

[@imnirajan]

lol, meanwhile i didn't notice even tho i watched it in 2x

[@KhalilEstell]

Apologies 🙏 ill try and slow down for my next two talks 😅

[@ratgr]

@@KhalilEstell No, please don't, watched at 1.75x and didn't feel it bad

[@avask17]

Very nice talk! Thanks)

[@taraslitvinenko8651]

thank you. best talk! best speaker! interesting im use c++ in bare metal, this my hobby

[@raymundhofmann7661]

I look forward to the arm-none-eabi-gcc or other architectures pre compiled toolsets generally available with exception handling enabled and exceptions for embedded use getting more polished support from the toolchains. But beware of the Rustifa!

[@leaving5891]

Great talk.👍

[@FisherGrubb]

Hmm.... how does Zephyr & FreeRTOS handle errors? How would this addition benefit them, would be great to know as that's what is used in industry 😉

[@no111u3]

No one wants use exception in industrial real code. If they want it, please ask that happened if exception occurred while they travel in car from 50 to 1 floor with speed more than 2 m/s.