Updated OS type identification

[interkosmos]

As noted in the source code and discussed in #144, the OS detection of get_os_type() in fpm/src/environment.f90 is not accurate. Please review this request, which checks environment variables and the existance of some files in order to determine the OS more precisely.

[LKedward]

Thanks @interkosmos, this looks like an improvement. Do note that you'll need to update any functions that use get_os_type such as list_files in fpm.f90 to consider all different OS behaviours.

I'm hoping #155 will be merged soon, and this further increase the usage of get_os_type (mainly for filesystem/command_line interaction) so it may be best to wait for this to be merged?

I have no experience with Solaris or BSD so I don't know what quirks need to be specifically addressed, but we may run into the problem that we cannot test them in the CI.

[awvwgk]

@LKedward BSD could be tested using Cirrus CI, which provides FreeBSD images. Grouping it together with OSX and Linux will probably be fine for most use cases right now, to avoid another CI provider for this project.

Testing cygwin is not (easily) possible on the GH actions windows image as well, which means another CI provider would be required for this purpose (like Appveyor).

No idea where to find a CI solution for Solaris.

[interkosmos]

In most cases, the behaviour of OS_LINUX, OS_MACOS, OS_CYGWIN, OS_SOLARIS, and OS_FREEBSD should be identical, if we stick to POSIX standards. I can’t say if there are any CI providers that feature OpenSolaris-based images (OpenIndiana, illumos, Tribblix, …).

[certik]

This looks good. I'll let @LKedward merge #155 first, and then this.

[certik]

Suggested change

	!! names. If this fails too, check the existance of files that can be
	!! names. If this fails too, check the existence of files that can be

[milancurcic]

Thank you!

[LKedward]

@interkosmos, would you mind also updating the following uses of get_os_type() in order to prevent the select case statements from silently falling through?

fpm/fpm/src/fpm_filesystem.f90

Lines 51 to 56 in e02171d

	select case (get_os_type())
	case (OS_LINUX,OS_MACOS)
	filesep = '/'
	case (OS_WINDOWS)
	filesep = '\'
	end select

fpm/fpm/src/fpm_filesystem.f90

Lines 117 to 124 in e02171d

	select case (get_os_type())
	case (OS_LINUX,OS_MACOS)
	call execute_command_line("mkdir -p " // dir , exitstat=stat)
	write(*,*) "mkdir -p " // dir
	case (OS_WINDOWS)
	call execute_command_line("mkdir " // windows_path(dir), exitstat=stat)
	write(*,*) "mkdir " // windows_path(dir)
	end select

fpm/fpm/src/fpm_filesystem.f90

Lines 148 to 158 in e02171d

	select case (get_os_type())
	case (OS_LINUX)
	call execute_command_line("ls " // dir // " > "//temp_file, &
	exitstat=stat)
	case (OS_MACOS)
	call execute_command_line("ls " // dir // " > "//temp_file, &
	exitstat=stat)
	case (OS_WINDOWS)
	call execute_command_line("dir /b " // windows_path(dir) // " > "//temp_file, &
	exitstat=stat)
	end select

fpm/fpm/src/fpm_command_line.f90

Lines 66 to 73 in e02171d

	select case (get_os_type())
	case (OS_LINUX)
	print *, "OS Type: Linux"
	case (OS_MACOS)
	print *, "OS Type: macOS"
	case (OS_WINDOWS)
	print *, "OS Type: Windows"
	end select

[certik]

The current failure is (on Windows):

Command line: gfortran -c -Jbuild\gfortran_debug\fpm -Ibuild\gfortran_debug\toml-f -Wall -Wextra -Wimplicit-interface -fPIC -fmax-errors=1 -g -fbounds-check -fcheck-array-temporaries -fbacktrace -o build/gfortran_debug/fpm/fpm_filesystem.o src\fpm_filesystem.f90
Exit code: 1
Stderr:
src\fpm_filesystem.f90:2:10:

     use :: fpm_environment, only: get_os_type, &
          1
Fatal Error: Can't open module file 'fpm_environment.mod' for reading at (1): No such file or directory
compilation terminated.

So that needs to be fixed before we can merge it.

[interkosmos]

@certik: does this happend because fpm_filesystem.f90 is compiled before fpm_environment.f90 by fpm?

[LKedward]

@interkosmos can you try reverting to the non colon form of the use statement - this may be a bug in Haskell fpm.

[LKedward]

Great work @interkosmos. CI is now passing so this looks good to go! Only left one minor comment regarding a remaining colon use statement.

[certik]

Looks good.

Next time I recommend to do changes like " -> ' (and white space formatting) in a separate pull request, as those are unrelated to the "OS type identification" improvement. It makes it easier to review of what actually changed.

[interkosmos]

Please excuse the inconvenience.