KDevelop With Clang Greatly Improves Its C++ Handling

[Michael]

I love how I give Michael a shout-out to that blogpost last night and then first comment on the story I get quoted and vindicated... thank you Erendorn for making me smile haha

I actually wrote about it last night but as I've been traveling when I actually make an article go live is usually shifted based on amount in publishing queue, but thanks anyhow!

[the303]

GCC is compiler, not static analyser of image processor or web server. Compiler get human readable text, output binary target machine codes.

Static analyser get text, output human readable related things. Maybe replace each sentence of input file by perfect poetry variant, who know.

JIT emulate some abstract hardware on-fly.

Why GGC team MUST inclute unrelated features? Just because some one else do same? To entertain audience? Why not tetris game, of hot chicks pictures? imagine, you start compling and same time screen loads of dosens of kittens!

Unrelated features? Do you know what is a compiler at all and how it works? GCC already has them, it does not expose them and leads to code duplication.

Because GCC compiles the code. It already does the static analysis and spur out numerous warnings while compiling. Why have some other external static analyzer and make a playground for numerous bugs (Eclipse's Shit++ parser cannot properly enable C++11 fetaures, it uses GCC for static analysis but the it must be a bad designed interface to interact with I assume)? It also requires extra effort to keep up with C++xy features of the X? compiler (extra effort to specialize the static analyzer for different compilers that has different support of features, some less some more <- this is not necessary though, yet better), while GCC already has one but does not expose it.

It can also be seen as unnecessary duplication in hardware level whilst it is not important in modern computers I assume.

You are wrong. Your argument is baseless and childish anyway yet I will explain further. Compiler is not a huge application that magically compiles text directly into machine code. The process has various sub-processes like Micket said yet GCC wants to take the code and do not let go until at least the object file is created, while keeping the steps until there. It is not nice that internal steps are not exposed to outside whilst they could be used for good things, i.e. the things that clang allowed.

GCC not allowing the linking process to be separated would be stupid. So is GCC not allowing the static analysis process to be separated.

[storm_st]

You are wrong. Your argument is baseless and childish anyway yet I will explain further.

Why GCC team must get your childish idea? Why i must read that flood of nonsence about completely unrelated things? Why you continue to stretch LLVM into GCC? GCC team can decide to use parsers internally, or make NP-like search over perfect optimised goals, or make link to astral body of Tau Ceti to produce code. First GCC was very based on LISP interpretator, lately ws rewritten more to pure C parser, then GCC was almost completely rewritten to GIMPLE and abstract graph operations. LLVM what? No one was care what LLVM doing.

Keep your LLVM promotion out of GCC team and infrastructure, that is all. GCC task always was making robust tool to produce quality code. LLVM tast was hi-scool framework for students to mess with parsers and graphs. Great. Untill thay was choosed by Appple to allow stoling FLOSS ppl work and jumpo over GPL3+. Now stop forcing GCC team to rewrite base FLOSS compiler to unrelated joke for kids. You need good text parser for your project? Google parser, not compiler. Need Jit? Google GIT, not GCC. If you choose LLVM framework for that i only will be happy. Thanks.

[Sergio]

Shit storm

Shit storm about how Clang is the devil and GCC is the best thing since sliced bread in 3, 2... Oh, it already began...

[Micket]

FRIENDS! I KNOW WHAT YOU ARE THINKING! You want to reply to storm_st.

But wait, he's an idiot who confuses modularity with feature creep. He doesn't have a single clue what the fuck he's talking about and won't listen to reason.
After only a few replies, his ramblings have gone from just plain ignorant, to incoherent, to paranoid delusions.

Lets all just ignore him.

[FLHerne]

FRIENDS! I KNOW WHAT YOU ARE THINKING! You want to reply to storm_st.

But wait, he's an idiot who confuses modularity with feature creep. He doesn't have a single clue what the fuck he's talking about and won't listen to reason.
After only a few replies, his ramblings have gone from just plain ignorant, to incoherent, to paranoid delusions.

Lets all just ignore him.

Ayup. He's had answers to both of his ridiculous points, no need to carry on with that.
So...does anyone actually have anything to say about KDevelop and/or Clang?

[Luke_Wolf]

Ayup. He's had answers to both of his ridiculous points, no need to carry on with that.
So...does anyone actually have anything to say about KDevelop and/or Clang?

ooh ooh I do!

This is awesome and really how it should be done, as opposed to IDE authors hacking together their own parsers it makes far far more sense for them to actually integrate with the compiler and understand the language itself, not only from a standpoint of removing reinventing the wheel for every IDE out there but it also allows a lot of cool features in the future, because it inherently understands the language. An extreme feature of this is an example that has been shown with Roslyn where you can copy code in one language and paste it in another language 100% correctly translated.

[Ericg]

I actually wrote about it last night but as I've been traveling when I actually make an article go live is usually shifted based on amount in publishing queue, but thanks anyhow!

Ah, oh well. either way I just wanted to make sure you saw it since I found it quite interesting and figured you would too

[zxy_thf]

I don't understand why GNU's GCC doesn't allow such features easily. After all, shouldn't GCC have embraced "do one thing and do it well" approach and should have been modular and therefore easily expose its code validation to outside? I know GCC has a code testing without actual compilation but apparently it is not really what it should be since all these good news about clang. Maybe phoronix is just feeding FUD.

Hoping they make GCC to have a better infrastructure in short time.

GCC‘s frontend is designed to be a compiler's frontent. Clang is designed to be a parser that can emit code for some compiler's backend.

[Skrapion]

Why must people get so up-in-arms about Clang? GCC did an amazing thing for the open source world, but it wasn't perfect, and over the decades it grew brittle. Clang learned from GCC's successes and failures, took up the baton, and ran it even further. One day the same thing will happen to Clang, but for now, the FOSS world has an even better compiler than it did before, and we're better for it.