next »

[_Tommo_]

I find providing both 32 and 64 bit linux builds really annoying. Can't get cross-compiling to work (Seems that I also have to cross-compile all the dependencies that I could otherwise get from the package repository)
I've finally decided to install another linux inside virtualbox. Is it really that hard or am I missing something?
I guess the easiest option would be to just provide the sources, but the code is too ugly and would be embarrasing.

The current state of cross-arch support on Ubuntu is just sad right now, they are actually doing big steps in the right direction, so that "most" of runtimes come in different binary paths (eg. linux-i386 and linux-amd64) but SDKs are still in a terrible state, and many useful packages aren't still there.

For the last projects I've just built two Ubuntu 10.04 chroots, inside my 64 bit Ubuntu, both to ensure binary compatibility with old stuff (ie. libstdc++ 16 is "too new") and to compile 32 bit.
It's really easy actually, you just need to follow this guide here.
Still, it is a solution to problems that shouldn't have been there in the first place.

[QuadrupleA]

What's the reason for targeting 64bit at all? The advantage is usually access to >2-4GB RAM, but that's more than needed for most applications. Larger instruction operands can make a slight improvement to performance but there are usually a lot of higher-leverage options you can exhaust before worrying about that.

[rivon]

Can't get a 32-bit Ubuntu installation to actually boot in VirtualBox.

Wut? I've been using VirtualBox for a long time and it always just works out of the box.

What's the reason for targeting 64bit at all? The advantage is usually access to >2-4GB RAM, but that's more than needed for most applications. Larger instruction operands can make a slight improvement to performance but there are usually a lot of higher-leverage options you can exhaust before worrying about that.

Most people today use 64bit Linux and 32bit libs aren't installed by default. And also a lot of people don't want them. Not providing a 64bit version is like a slap in the face.

[Evan Balster]

Hey C/C++ naming convention authorities,

You can blame most of the major ALLCAPS stuff on Microsoft, and the bizarro abbreviated functions on C.  The C++ standard API, other than its backwards-compatibility bits, uses decent nomenclature.  Where it goes horribly awry is its templates.  D:

[ThemsAllTook]

Can't get a 32-bit Ubuntu installation to actually boot in VirtualBox.

Wut? I've been using VirtualBox for a long time and it always just works out of the box.

I've tried a few times now, and each time the installer has given me an uninformative error or two and then appeared to complete, but what it installs doesn't boot. It just sits at a blank screen after login and I can't do anything. Using Ubuntu 12.10 on a 64-bit Windows 7 host. I also have a bootable 64-bit Ubuntu host, which I haven't tried yet...maybe that would work better? I'll give it a shot later today.

[QuadrupleA]

Most people today use 64bit Linux and 32bit libs aren't installed by default. And also a lot of people don't want them. Not providing a 64bit version is like a slap in the face.

Good to know - been out of the Linux world for a while now. Used RedHat exclusively for a few months in the early 2000's but one too many conf files and source code build procedures broke my spirit

[epcc]

I've been struggling with the same thing recently. Can't get a 32-bit Ubuntu installation to actually boot in VirtualBox.

Please let us know if you get it all figured it out!

Can't say anything about Ubuntu, but linux mint installed as good as on a real computer. Performance is awful, however. On a Pentium 4 it took so long to even launch the installer that I decided to hijack my brothers computer. On a recent high-end CPU (i7?) it ran as good as on Linux Mint runs on my Pentium 4.

[Schrompf]

Current VMs should be at ~70% of the host performance, thanks to virtualization support from the current CPUS. I personally had no problems installing some Linux flavour in a VM. But to me it was just "mount ISO and let boot", so I can't assist here.

[ThemsAllTook]

Just switched to a new OS version and browser (Snow Leopard -> Mountain Lion, Safari -> Chrome). Everything is uncomfortable while I get used to the changes. Of course, none of my code compiles anymore either... I also switched to 64-bit Windows and Linux recently which also broke things, so I actually don't have a platform where I can build and run anything right at the moment.

[RalenHlaalo]

I just spent about an hour trying to figure out why my quad wasn't drawing only to realise I'd set its colour to the same as the background

[QuadrupleA]

I just spent about an hour trying to figure out why my quad wasn't drawing only to realise I'd set its colour to the same as the background

Sounds like the many stupid graphics bugs I've created over the years

[Geti]

Not sure if happy or grumpy. Swap from vector<bool> to bitset, huge performance increase in debug build. Surely operator[] for vector<bool> isn't that different than the bitset one?!

[QuadrupleA]

Not sure if happy or grumpy. Swap from vector<bool> to bitset, huge performance increase in debug build. Surely operator[] for vector<bool> isn't that different than the bitset one?!

Weird - not sure if you need it or not, but if you're in Visual Studio, the debug C/C++ runtime libraries slow things waaaay down - I switched them in my Debug build (Properties, C++ Code Generation, /M* switches and "Basic Runtime Checks") and got a 16-22x performance improvement in BOOST_FOREACH and STL iterator-based for loops (my game loop went from 12ms to 0.4ms).

Have since scrapped BOOST_FOREACH and STL iterators in most places for good-ol' integer index for loops - they're lightning fast in both debug and optimized builds (0.05ms for above-mentioned loop), wheras STL/Boost seem klunky as hell unless you turn all the optimizations on.

[RalenHlaalo]

I just fixed a bug that had me completely baffled for hours. The moment I finally discovered the problem was one of the biggest coding mind-fucks I've experienced 

My morale was already pretty low from the previous bug (see above).

[Geti]

Have since scrapped BOOST_FOREACH and STL iterators in most places for good-ol' integer index for loops

Welcome to our world Currently best practice / 3rd party coder friendly code comes waaaay further down the list than fast to write and fast to execute code that helps ship this game.
Cheers for the tip on turning off some of the debugging crap though. I feel like blaming a shocking implementation of std::vector<bool> cause it was seriously the use of operator[] that was causing the slowdown.
Obviously both were fast enough in release builds, but we spend most of our time _not_ in release builds

The moment I finally discovered the problem was one of the biggest coding mind-fucks I've experienced 

Do we get to know what this crazy problem is?

[RalenHlaalo]

Do we get to know what this crazy problem is?

It's a little complicated to explain exactly, but essentially I was performing a static down-cast on a pointer to the wrong type, but the type that I was wrongly casting to still contained the data I needed, which coincidentally was located in the same place as it would have been in the other class (the one I should have been casting to), and therefore the code worked fine, and had been working fine for months.

Then, last night I added a new data member to the class (EDIT: actually it was to a class that the class contains an instance of), thereby changing its layout, and that's when I noticed some strange behaviour. Every time the player entered a 'gravity region' (see old prototype), he was relocated to position (0,0). I've never been so confused in my life  I couldn't believe such a serious error had been silently lurking in the code all that time.

[Graham-]

You better believe. Shit happens :|.

[QuadrupleA]

Much as I'm loving STL and the Boost libraries, its authors may be among most pedantic bastards to ever walk the earth - reading through their documentation and references is like decoding mathematical proofs.

Tooltips get a little template-happy as well:   

std::Tree<std::_Tmap_traits<std::string, boost::property_tree::ptree, std::less<std::string>, std::allocator<std::pair<const std::string, boost::property_tree::ptree>>, false>>::const_iterator begin() const

[doihaveto]

Much as I'm loving STL and the Boost libraries, its authors may be among most pedantic bastards to ever walk the earth - reading through their documentation and references is like decoding mathematical proofs.

Funny you should say that. I've been reading "Elements of Programming" by Stepanov (STL author) and McJones. It's really, really good - but it's incredibly mathematical, to a point where it doubles as an introduction to abstract algebra. Which I thought was really unique for a CS book.

[pixhead]

Windows API. That is all.