The Lounge is rated Safe For Work. If you're about to post something inappropriate for a shared office environment, then don't post it. No ads, no abuse, and no programming questions. Trolling, (political, climate, religious or whatever) will result in your account being removed.
While the MSVC compiler has improved significantly in recent years it still lags:
Note all the asterisks next to the C++20 library and language features.
The bottom line is those are partial or otherwise "supported with qualifications" which at the end of the day means your standard C++20 code might not compile for MSVC without conditional compiles.
Like I said, it has gotten better, but so did internet explorer before it was replaced.
I prefer to be able to use -std=C++20 and be confident that my standard code will compile, or if it won't, then it's *my fault* - metaprogramming is difficult enough when you know what you're using is fully supported. If it's not, tracking down the error can be a nightmare - I speak from unfortunate experience.
Additionally, the fact that MSVC consistently lags behind in standards support by years (i'm not saying lags behind other compilers, I'm saying lags behind the standard itself) means if I want my code to compile across compilers I have to dumb it down for the lowest common denominator.
And between the major compilers, that lowest common denominator comes up as MSVC. I guess some compiler had to fill that role, but they were going to be my target whoever they were, because they're going to be the reason that I can't have nice things.
Most of those asterisks seem to just show a different kind of version number.
20 years ago, MSVC lagged seriously behind the standard (I'm talkin' to you MSVC 6). Today, conformance is excellent, though for a given feature it may take longer to get into MSVC than the open-source compiler in which the feature was prototyped (duh). I'm comparing compilers for a book on optimization, and so far I haven't encountered a situation in which the three compilers didn't all support a feature, though I accept that they exist.
There are in fact differences in the way the compilers interpret the standard, but this has always been more a bug in the standard than in any of the compilers, or a situation in which one compiler accepts a more generous (nonconforming) syntax than another.
I'm guilty of glossing over most of that link. My fault. Serves me right for digging it up while in the middle of other things..
I admit a lot of my distaste for MSVC is historical. Pre 2017 it wasn't great, but since then I haven't used it as much as I used to because I do cross compiling mostly these days for MCUs and such.
That it didn't support SFINAE was just unacceptable, IMO.
At the end of the day, I don't trust their compiler products, and frankly, they earned that. It will take some time - a consistent showing of compliance over several standards versions for me to want to rely on them again.
It's possible to do portable executables (assuming I've picked the context right) with GCC , but clunky (haven't tried to do it with clang). You have to use windres to build a .res file, then you can link that as a source file like any other .c or .cpp file. Looking at an old makefile I think the "-mconsole" and "-mwindows" flags on the GCC command are important, but it's been long enough since I touched this that I can't really remember exactly how it works.
clang and gcc can compile PE files just fine. The only thing is that debug information is in DWARF format, not PDB, but that just means you use GDB rather than a Microsoft debugger.
The company I work for has used GCC to produce Windows executables from Ada, C and Fortran for decades (20 years counts as 'decades', right?). The gcc linker can be slow in some circumstances, but aside from that, it works just fine.
What frustrates you with msvc, BTW? It's improved immensely since the dark days of the late 90s/early 2000s, when it felt like they'd never keep up with the C++ Standard...
Java, Basic, who cares - it's all a bunch of tree-hugging hippy cr*p
Well, trying to use it to do SFINAE it failed, and worse, because it didn't support it, it also didn't support std::enable-if.
Now I don't know if they do now. *googles* - looks like they do finally, but frankly, I don't want to have to guess about which parts of the elephanting standard they feel are important while I'm coding. Or wonder if the code I'm writing is problematic or if it's the compiler being well... MSVC. It's a time sink.
And I hate their language extensions. everything is a #$&# pragma
Yeah but look at their C++20 support compared to GCC.
Someone should remind MS it's 2021.
I don't want to have to find out they don't support something the hard way when I'm in the middle of coding.
I'm not using C++20 at the moment, but I plan on switching over soon. If MS has been consistent about anything, it's being years behind the standards.
So I stand by the title of the post
Also if a codebase won't work with GCC I'd argue the problem is almost certainly with the codebase, not gcc.
1. GCC is a modern, popular compiler that is very standards compliant compared to other offerings. If your source can't work with it you're shooting yourself in the foot
2. What about your code isn't actually C++, but really a variant of C++ involving vendor specific extensions? It's not always the issue, but I find it usually is.
I don't typically use MSVC. In fact, I don't even test my code to see if it compiles on MSVC anymore - just clang and gcc. If MSVC can't compile it, MSVC is broken, as far as I'm concerned. That's how I treat it.
I think you misunderstand me. All compilers have bugs. The question is whether standards compliant code will compile with them or not.
Sure, gcc has some bugs, and some of those bugs might impact the above, but they have a lot of catching up to do if they want to ride the Microsoft fail train. This was especially true on previous versions of MSVC, but even today they're still lagging behind clang and GCC in terms of C++ standards support.
At the end of the day, that's what I care about.
Like, as an analogy (but i don't do webdev) I don't care that Chrome may have a rendering bug. I care that it supports HTML5, so that my HTML5 code will theoretically render capably under that system. If there's a specific bug I can deal with that, but that's knowing that Google knows it's a bug and is treating it like a bug.
Consider the different approach MS took with IE. It mirrors their treatment of C++ in terms of standards.
Microsoft's lack of standards compliance isn't treated as a bug. At least historically (which I can speak to with confidence, as opposed to their specific culture and attitude *today* WRT to C++) has been that it's by design, and if you code in the Microsoft ecosystem, you just have to suck it up. Maybe they'll get around to it on the next major version, maybe not.
That's a rather marked difference in terms of what I can expect and what I can rely on when I go to design my code.
In the first case, worst case I have a conditional for Chrome, to work around the bug, and make a note of it in the source.
In the second case, I basically have to fork, because not only is it not going to be fixed, but it's probably not the only thing they *decided* to do differently.