|
IMDEA Software researchers Facundo Molina, Juan Manuel Copia and Alessandra Gorla present FIXCHECK, a novel approach to improve patch fix analysis that combines static analysis, randomized testing and large language models. Why just get the AI to write the code when you can get it to write the tests to prove it works?
What could possibly go right?
|
|
|
|
|
Heh. Haven't read just yet.
I'm aware of some folks who would definitely not see it this way, but I think auto-generation of tests is a pretty good idea. A big reason though (maybe irony?) is that we cannot get this bit wrong. When we mess up a test it's like we did worse than any other kind of mess up and we did it on code we don't even have to have.
What if when I clicked Post Message, I held down a magic key combo which caused a branching on the repo with a new commit that contained tests across all logic that click caused to run?
|
|
|
|
|
Proprietary code goes unpublished – but no FOSS package ever dies If they ain't broke, rewrite them anyway?
|
|
|
|
|
One of the biggest time sinks in software development is code review and approval. Unless you count the "should this be InitializeWorkflow or StartWorkflow?" discussions as finding bugs
BeginStartInitializeWorkflow, of course
|
|
|
|
|
I always try my best to read every line of changed code in a review. I have found obvious bugs and requested changes. Sometimes you can even tell the developer didn't even bother to try and compile the code before submitting.
Looking at you M****s. (Someone I used to work with.)
I’ve given up trying to be calm. However, I am open to feeling slightly less agitated.
I’m begging you for the benefit of everyone, don’t be STUPID.
|
|
|
|
|
Hard to argue with the points even if the thesis put a cap in my knee.
"First, the yield on finding bugs in reviews was about 15% of code review comments defects despite it being the top motivation."
But I don't think that's the chief motivation. Rather, "meta" is. Was the approach a good one? Is there way better way? Are there fundamental oversights to the implementation? Are there code features that the reviewer learns from the review or that the reviewer can pass to the reviewed on account of the code in question being exhibition? Code style/staying within 'the lines'? Were the tests written? If they wrote tests, and you review those, there is a sort of comfort that the code doesn't have bugs. It's not 100% but...
|
|
|
|
|
Well, bugs usually involve the complex interplay of code from various different source files.
I doubt that any code reviewer could spot that looking at one file at a time.
The difficult we do right away...
...the impossible takes slightly longer.
|
|
|
|
|
At my last job, code reviews were very effective. They covered more than bugs; they helped sanity check code design.
|
|
|
|
|
We’re bringing over 350 additions to Uno Platform in this release, with the standout being official support for JetBrains Rider. Today I learned that there's more than just Visual Studio out there
|
|
|
|
|
I could probably learn to really like Rider.
The problem with that is I don't know if future employers would like Rider.
I suppose that's somewhat non-issue. It can't be that much to license it for yourself. But then there's maybe some code stylization sorts of things which may differ from VS (like both may have 'cleanup code' but one does it slightly different).
|
|
|
|
|
The 405B parameter model beats rivals at math, coding and multilingual tasks, Meta claimed. It now whips Winamp's butt
|
|
|
|
|
Now, Intel plans to address the processors’ troublesome ‘elevated operating voltage’ with a microcode patch. It seems you can have too much power
|
|
|
|
|
The GPT Builder ceased to exist as of July 10, with all custom GPTs now deleted by Microsoft. RIP: March 2024 -July 2024
Killed before the download was complete
|
|
|
|
|
The advertising industry can heave a sigh of relief. That's the way the cookie doesn't crumble
|
|
|
|
|
Freelancer, one of the biggest freelancing websites, reported that even with AI growth, freelancing jobs are increasing. Sadly, all the jobs are to use AI
|
|
|
|
|
The companies try AI, see that the results don't match the hype, and go back to "traditional" freelancing methods.
Freedom is the freedom to say that two plus two make four. If that is granted, all else follows.
-- 6079 Smith W.
|
|
|
|
|
|
A report sponsored by DevOps company JFrog suggests that executives over-estimate the extent to which developers within their organization defend against vulnerable or malicious packages in the software supply chain. If you're shocked, you might just be executive material
|
|
|
|
|
Once, Azure was a cloud platform dedicated to Windows. It's The Year of Linux on Azure!
|
|
|
|
|
It's the millennium of the Years of (talking about) Linux!
|
|
|
|
|
It's the age of aquariums!
|
|
|
|
|
That sounds a little fishy to me. 
"the debugger doesn't tell me anything because this code compiles just fine" - random QA comment
"Facebook is where you tell lies to your friends. Twitter is where you tell the truth to strangers." - chriselst
"I don't drink any more... then again, I don't drink any less." - Mike Mullikins uncle
|
|
|
|
|
What if the Python programming language itself was malicious? Fortunately it can be defeated with a single tab character
|
|
|
|
|
Windows Subsystem for Android (WSA) was supposed to be a game-changer. So it will be finished and finished
|
|
|
|
|
Our team has made changes to the C/C++ extension and the GitHub Copilot extension in VS Code to ensure that other relevant C++ context — like available types and methods — are also provided to Copilot completions. Now it always fills in that 'i++' in for loops
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
