I have a makefile that works in Windows using MinGW as well as on Linux. I want to tidy it up using some shell scripting in the recipe. But for that I need to add MSYS to my environment.
I am not sure if the users of MinGW use MSYS always/often/rarely/never?
Any feedback please.
Related
Question
As I understand from boost.org, MinGW in not supported, that is, if Boost library (or program using Boost) is compiled with MinGW, then it is unstable.
Is it correct?
If it is, then what is going on with bitcoin (it is compiled on Windows with MinGW)?
Link
http://www.boost.org/doc/libs/1_55_0/more/getting_started/windows.html#pch
Excerpt from the link
"A note to Cygwin and MinGW users
If you plan to use your tools from the Windows command prompt, you're in the right place. If you plan to build from the Cygwin bash shell, you're actually running on a POSIX platform and should follow the instructions for getting started on Unix variants. Other command shells, such as MinGW's MSYS, are not supported—they may or may not work."
Thanks in advance,
--- Kosta
Only MinGW's MSYS command shell is not supported.
MSYS is a collection of GNU utilities such as bash, make, gawk and grep to allow building of applications and programs which depend on traditionally UNIX tools to be present. It is intended to supplement MinGW and the "deficiencies" of the cmd shell.
BUT you can build Boost library with MinGW and b2 under the standard Windows cmd.exe command prompt.
With MinGW installed and paths correctly set, usually these two commands are enough:
bootstrap.bat mingw
b2 threading=multi toolset=gcc
Once you have the libraries you can use them under the MSYS environment.
When I use the MinGW shell, I can use all the UNIX dev paths (such as /usr/local/include and /usr/local/lib) in my GCC commands. However, if I try to tell Code::Blocks to include /usr/local/include, it does not recognize it. It seems that when using Code::Blocks, I am not tapping into the msys 'environment'. Is there a way I can do that? For example, I'd love to be able to commit Code::Blocks project files that target those paths and just work regardless of where someone installed MinGW.
I have a C project with a Makefile, I want to build it on windows. I downloaded GNU Make installed. And I have VS2010. The make file has UNIX commands like cc, nvcc and g++. I have to replace them with MS C, CUDA and C++ compilers right? What are the command line commands I have to replace them with? Is there anything I need to install?
I want to do it without Cygwin (it's installation is taking forever), otherwise Cygwin is the fallback method.
You don't actually have to replace any of those commands with the Windows-variant. A working compiler will still compile source code just fine even if it isn't the defacto compiler for an operating system.
I hear ya on the Cygwin bit--you can just install a Minimal GNU environment using MinGW and that'll give you everything that you need to be able to the utilize the Unix makefile.
I am using Cygwin and have GCC (version 4.3.4 20090804 (release) 1 ) installed as Cygwin package.
When I built C code using GCC under Cygwin shell, the generated executable output file is a executable of type (PE32 executable for MS Windows (console) Intel 80386 32-bit) and it can only be executed/run under Cygwin shell, not as standalone .exe on Windows shell/command prompt. If I try to run it standalone on Windows command prompt it gives an error window saying "The program can't run because cygwin.dll is missing from your computer".
How can one make this .exe standalone, which can be executed on a command prompt of any other system or even in my own system?
I thought GCC under Cygwin would build a Linux executable (ELF 32-bit LSB executable), but it's not so. How can I use the gcc-cygwin combination to generate a *.out kind of Linux executable file?
Also, I cannot run a Linux executable generated on a Linux-gcc combination to execute under Cygwin.
Any pointers would be helpful.
Despite widespread rumours, Cygwin is not a Linux emulator, i.e. it doesn't produce or run Linux executables. For that you'll need a virtual machine or coLinux.
Instead, Cygwin is a compatibility layer, which aims to implement as much as possible of the POSIX and Linux APIs within Windows. This means that programs have to be compiled specifically for Cygwin, but it also means that it's better integrated with Windows, e.g. it lets you mix POSIX and Windows APIs in the same program.
It's the cygwin1.dll that provides that compatibility layer. If you build non-Cygwin executables using gcc-3's -mno-cygwin switch or one of the MinGW compilers, then of course you can't use any POSIX/Linux-specific APIs, and any such calls will need to be replaced with Windows equivalents.
Cygwin is an emulation layer. It allows UNIX code to run on Windows, if you compile it to use the emulation layer. To Windows it looks like any normal DLL and makes OS calls as normal. But when you compile against it, it shows the same functions as UNIX (well, POSIX technically, but UNIX it is)
1) When you build with cygwin, it automatically brings in this cygwin1.dll. This is the code that you need to make it look like UNIX. There are flags to make it not use this cygwin dll, meaning ignore the UNIX stuff and use native Windows calls. the -mno-cygwin flag will make the binary a native Windows binary, not using cygwin emulation. As others have said, you also need the -mwindows to make actual GUI apps using cygwin gcc.
2) To compile on one platform to run on another platform, you need what's called a cross compiler. You also need the libraries and headers for the other system. For what you want you'd need a cygwin-Linux cross compiler, Linux headers and libraries. It would probably be much easier to run Linux in a virtual machine than to set this up.
3) Remember that Cygwin just looks like UNIX to UNIX source code, not to Linux binaries. Once you compile things, the function calls are windows DLL calls. A cygwin program is still a Windows program (which is why you need a Windows style DLL to run it). Cygwin provides code for UNIX functions/system calls such as fork(), but even the method of calling them is now different. fork() now calls into a Windows DLL, not into a Linux kernel. It's still a very different system.
Not a complete answer, but I think I am able to give some pointers.
1) http://www.cygwin.com/cygwin-ug-net/programming.html says you should use -mswindows parameter. Take a look of MinGW.
2) You need a cross gcc to do this. By default cygwin gcc produces binaries linked against cygwin.dll.
3) That it because it is a linux binary. Easiest way is to recompile software in cygwin.
ad 1) There are currently three different mingw cross-compilers available for cygwin:
The old gcc3 -mno-cygwin is deprecated.
There are new mingw64 cross-compilers for 32 bit (mingw64-i686-gcc) and 64 bit windows targets (mingw64-x86_64-gcc).
There's no mingw-i686-gcc matching the official cygwin gcc4 compiler yet.
ad 2) There's no linux cross-compiler as cygwin package yet available. But people report success building such a cross-compiler by themselves.
ad 3) There's no cygwin cross-compiler as linux package available, but many mingw cross-compilers. Those mingw executables can also be executed under cygwin, though they cannot use cygwin features, just the simple windows runtime.
Correcting errors in others people posts:
-mswindows is not valid, -mwindows tells the linker to generate a GUI app without console.
-mno-cygwin is only valid for the old deprecated gcc3 compiler and is not supported anymore. Don't use it. With cygwin you should use ordinary host and target triples as with every other cross-compiler.
You need to have cygwin.dll in your path.
Or just use MinGW to compile native windows code without dependencies.
I've installed cygwin environment on Windows.
There is gcc 4.3. How to install gcc 4.4 in this environment?
The alternative is to build one yourself, but it will be extremely slow on cygwin. If you are a Linux user, you can build a native compiler (which runs in cmd.exe on windows, and produces native win32 binaries) on Linux. It is much faster (order of magnitude faster in my experience).
Otherwise, mingw is a good solution, although in my experience, the binary from equation.com work better for gcc 4.4 (there is no official gcc 4.4 from MinGW yet)
If you don't need the cygwin environment for special reasons, I'd go for Mingw with the gcc 4.4. You can find builds on: http://www.tdragon.net/recentgcc/.
Reading the MinGW FAQ, reveals that its possible to pass requests to use the MinGW binaries instead of the ones shipped with Cygwin.
For those who would like to use the Cygwin environment for development, yet generate non-Cygwin-dependant executables, a much easier option to "-mno-cygwin" does exist. Simply install Cygwin and the MinGW distribution in separate directories (i.e. "C:\CYGWIN" and "C:\MINGW"), and make sure that the "/bin" subdirectory beneath your MinGW installation comes before Cygwin's "/bin" subdirectory in your PATH environment variable (i.e. "PATH=%PATH%;C:\MINGW\BIN;C:\CYGWIN\BIN"). This will allow you access to all the UNIX tools you want, while ensuring that the instance of GCC used is the MinGW version. %%%
Bear in mind that within the "/etc/profile" file, Cygwin by default places "/usr/local/bin", "/usr/bin", and "/bin" ahead of your system-level PATH. Therefore, it is not enough to have the MinGW's "/bin" ahead of Cygwin in your Windows path... it must also be set to come first within the Cygwin environment (either by modifying "/etc/profile" or setting it manually).