Newer
Older
- Installing Cygwin
- Download and Unpack
- Semi-Optional apps/ Package
- Installation Directories with Spaces in the Path
- Instantiating "Canned" Configurations
- NuttX Configuration Tool
- Incompatibilities with Older Configurations
- Cross-Development Toolchains
- NuttX Buildroot Toolchain
o Shells
- Build Targets and Options
patacongo
committed
- Native Windows Build
- Installing GNUWin32
patacongo
committed
o Cygwin Build Problems
- Strange Path Problems
- Window Native Toolchain Issues
NuttX may be installed and built on a Linux system or on a Windows
system if Cygwin is installed. Installing Cygwin on your Windows PC
is simple, but time consuming. See http://www.cygwin.com/ for
installation instructions. Basically you just need to download a
tiny setup.exe program and it does the real, internet installation
for you.
NOTE: NuttX can also be installed and built on a native Windows
system, but with some potential tool-related issues (see the
1. Install at C:\cygwin
2. Install EVERYTHING: "Only the minimal base packages from the
Cygwin distribution are installed by default. Clicking on categories
and packages in the setup.exe package installation screen will
provide you with the ability to control what is installed or updated.
Clicking on the "Default" field next to the "All" category will
provide you with the opportunity to install every Cygwin package.
Be advised that this will download and install hundreds of megabytes
to your computer."
If you use the "default" installation, you will be missing many
of the Cygwin utilities that you will need to build NuttX. The
build will fail in numerous places because of missing packages.
NOTE: You don't really have to install EVERYTHING but I cannot
answer the question "Then what should I install?" I don't know
the answer to that and so will continue to recommend installing
EVERYTHING.
After installing Cygwin, you will get lots of links for installed
tools and shells. I use the RXVT native shell. It is fast and reliable
and does not require you to run the Cygwin X server (which is neither
fast nor reliable). Unless otherwise noted, the rest of these
instructions assume that you are at a bash command line prompt in
either Linux or in Cygwin shell.
UPDATE: The last time I installed EVERTHING, the download was
about 5GiB. The server I selected was also very slow so it took
over a day to do the whole install!
Download and Unpack
-------------------
patacongo
committed
Download and unpack the NuttX tarball. If you are reading this, then
you have probably already done that. After unpacking, you will end
up with a directory called nuttx-version (where version is the NuttX
version number). You might want to rename that directory nuttx to
match the various instructions in the documentation and some scripts
in the source tree.
Semi-Optional apps/ Package
---------------------------
All NuttX libraries and example code used to be in included within
the NuttX source tree. As of NuttX-6.0, this application code was
moved into a separate tarball, the apps tarball. If you are just
beginning with NuttX, then you will want to download the versioned
apps tarball along with the NuttX tarball. If you already have your
own product application directory, then you may not need the apps
tarball.
It is call "Semi-optional" because if you don't have some apps/
directory, NuttX will *fail* to build!
Download the unpack the apps tarball in the same directly where you
unpacked the NuttX tarball. After you unpack the apps tarball, you
will have a new directory called apps-version (where the version
should exactly match the version of the NuttX tarball). Again, you
might want to rename the directory to simply apps/ to match what
you read in the documentation
After unpacking the apps tarball, you will have two directories side
by side like this:
|
+----+----+
| |
nuttx/ apps/
This is important because the NuttX build will expect to find the
apps directory in that (default) location. That default location
can be changed by editing your NuttX configuration file, but that
Installation Directories with Spaces in the Path
------------------------------------------------
The nuttx build directory should reside in a path that contains no
spaces in any higher level directory name. For example, under
patacongo
committed
Cygwin, your home directory might be formed from your first and last
names like: "/home/First Last". That will cause strange errors when
the make system tries to build.
[Actually, that problem is probably not to difficult to fix. Some
I work around spaces in the home directory name, by creating a
new directory that does not contain any spaces, such as /home/nuttx.
Then I install NuttX in /home/nuttx and always build from
patacongo
committed
Downloading from Repositories
-----------------------------
Gregory Nutt
committed
Cloning the Repository
Gregory Nutt
committed
The current NuttX du jour is available in from a GIT repository. Cloning
instructions are available here:
Gregory Nutt
committed
https://sourceforge.net/p/nuttx/git
Configuring the Cone
Set your identity:
git config --global user.name "My Name"
git config --global user.email my.name@example.com
Colorized diffs are much easier to read:
git config --global color.branch auto
git config --global color.diff auto
git config --global color.interactive auto
git config --global color.status auto
Checkout other settings
git config --list
Gregory Nutt
committed
Cloning NuttX Inside Cygwin
Gregory Nutt
committed
If you are cloning the NuttX repository, it is recommended to avoid
automatic end of lines conversions by git. These conversions may break
some scripts like configure.sh. Before cloning, do the following:
Gregory Nutt
committed
git config --global core.autocrlf false
Notes about Header Files
------------------------
patacongo
committed
patacongo
committed
Some toolchains are built with header files extracted from a C-library
distribution (such as newlib). These header files must *not* be used
with NuttX because NuttX provides its own, built-in C-Library. For
toolchains that do include built-in header files from a foreign C-
Library, NuttX must be compiled without using the standard header files
that are distributed with your toolchain. This prevents including
conflicting, incompatible header files (such as stdio.h).
patacongo
committed
Header Files Provided by Your Toolchain.
Certain header files, such as setjmp.h, stdarg.h, and math.h, may still
be needed from your toolchain and your compiler may not, however, be able
to find these if you compile NuttX without using standard header file.
If that is the case, one solution is to copy those header file from
your toolchain into the NuttX include directory.
Duplicated Header Files.
There are also a few header files that can be found in the nuttx/include
directory which are duplicated by the header files from your toolchain.
stdint.h and stdbool.h are examples. If you prefer to use the stdint.h
and stdbool.h header files from your toolchain, those could be copied
into the nuttx/include/ directory. Using most other header files from
your toolchain would probably cause errors.
math.h
Even though you should not use a foreign C-Library, you may still need
to use other, external libraries with NuttX. In particular, you may
need to use the math library, libm.a. NuttX supports a generic, built-in
math library that can be enabled using CONFIG_LIBM=y. However, you may
still want to use a higher performance external math library that has
been tuned for your CPU. Sometimes such such tuned math libraries are
bundled with your toolchain.
The math libary header file, math.h, is a then special case. If you do
nothing, the standard math.h header file that is provided with your
toolchain will be used.
If you have a custom, architecture specific math.h header file, then
that header file should be placed at arch/<cpu>/include/math.h. There
is a stub math.h header file located at include/nuttx/math.h. This stub
header file can be used to "redirect" the inclusion to an architecture-
specific math.h header file. If you add an architecture specific math.h
header file then you should also define CONFIG_ARCH_MATH_H=y in your
NuttX Configuration file. If CONFIG_ARCH_MATH_H is selected, then the
top-level Makefile will copy the stub math.h header file from
include/nuttx/math.h to include/math.h where it will become the system
math.h header file. The stub math.h header file does nothing other
than to include that architecture-specific math.h header file as the
float.h
If you enable the generic, built-in math library, then that math library
will expect your toolchain to provide the standard float.h header file.
The float.h header file defines the properties of your floating point
implementation. It would always be best to use your toolchain's float.h
header file but if none is available, a default float.h header file will
provided if this option is selected. However, there is no assurance that
the settings in this float.h are actually correct for your platform!
In most cases, the correct version of stdarg.h is the version provided
with your toolchain. However, sometimes there are issues with with
using your toolchains stdarg.h. For example, it may attempt to draw in
header files that do not exist in NuttX or perhaps the header files that
is uses are not compatible with the NuttX header files. In those cases,
you can use an architecture-specific stdarg.h header file by defining
CONFIG_ARCH_STDARG_H=y.
See the discussion above for the math.h header. This setting works
exactly the same for the stdarg.h header file.
Instantiating "Canned" Configurations
-------------------------------------
"Canned" NuttX configuration files are retained in:
configs/<board-name>/<config-dir>
Where <board-name> is the name of your development board and <config-dir>.
Configuring NuttX requires only copying three files from the <config-dir>
to the directory where you installed NuttX (TOPDIR) (and sometimes one
additional file to the directory the NuttX application package (APPSDIR)):
Copy configs/<board-name>/<config-dir>/Make.def to ${TOPDIR}/Make.defs
Make.defs describes the rules needed by you tool chain to compile
and link code. You may need to modify this file to match the
specific needs of your toolchain.
Copy configs/<board-name>/<config-dir>/setenv.sh to ${TOPDIR}/setenv.sh
setenv.sh is an optional convenience file that I use to set
the PATH variable to the toolchain binaries. You may chose to
use setenv.sh or not. If you use it, then it may need to be
modified to include the path to your toolchain binaries.
Copy configs/<board-name>/<config-dir>/defconfig to ${TOPDIR}/.config
The defconfig file holds the actual build configuration. This
file is included by all other make files to determine what is
included in the build and what is not. This file is also used
to generate a C configuration header at include/nuttx/config.h.
General information about configuring NuttX can be found in:
${TOPDIR}/configs/README.txt
${TOPDIR}/configs/<board-name>/README.txt
There is a configuration script in the tools/ directory that makes this
easier. It is used as follows:
cd ${TOPDIR}/tools
./configure.sh <board-name>/<config-dir>
There is an alternative Windows batch file that can be used in the
windows native environment like:
cd ${TOPDIR}\tools
configure.bat <board-name>\<config-dir>
See tools/README.txt for more information about these scripts.
Refreshing Configurations
-------------------------
Configurations can get out of data. It is a good practice to "refresh"
each configuration before making. To refresh the configuration, use the
NuttX Configuration Tool like this:
make oldconfig
If you configuration is out of date, you will be prompted to resolve the
issues detected by the configuration tool. Doing this can save you a lot
of problems done the road due to a bad configuration. The NuttX
NuttX Configuration Tool
------------------------
An automated tool is under development to support re-configuration
of NuttX. This tool, however, is not yet quite ready for general
usage.
This automated tool is based on the kconfig-frontends application
available at http://ymorin.is-a-geek.org/projects/kconfig-frontends
(A snapshot of this tool is also available at ../misc/tools). This
application provides a tool called 'kconfig-mconf' that is used by
the NuttX top-level Makefile. The following make target is provided:
make menuconfig
This make target will bring up NuttX configuration menus.
WARNING: Never do 'make menuconfig' on a configuration that has
not been converted to use the kconfig-frontends tools! This will
http://www.nuttx.org/doku.php?id=wiki:howtos:convertconfig).
The 'menuconfig' make target depends on two things:
1. The Kconfig configuration data files that appear in almost all
NuttX directories. These data files are the part that is still
under development (patches are welcome!). The Kconfig files
contain configuration information for the configuration settings
relevant to the directory in which the Kconfig file resides.
NOTE: For a description of the syntax of this configuration file,
see ../misc/tools/kconfig-language.txt.
2. The 'kconfig-mconf' tool. 'kconfig-mconf' is part of the
kconfig-frontends package. You can download that package from
the website http://ymorin.is-a-geek.org/projects/kconfig-frontends
or you can use the snapshot in ../misc/tools.
Building kconfig-frontends under Linux may be as simple as
'configure; make; make install' but there may be some build
complexities, especially if you are building under Cygwin. See
the more detailed build instructions at ../misc/tools/README.txt
The 'make install' step will, by default, install the 'kconfig-mconf'
tool at /usr/local/bin/mconf. Where ever you choose to
install 'kconfig-mconf', make certain that your PATH variable includes
a path to that installation directory.
The kconfig-frontends tools will not build in a native Windows
environment directly "out-of-the-box". For the Windows native
case, you should should the modified version of kconfig-frontends
that can be found at
http://uvc.de/posts/linux-kernel-configuration-tool-mconf-under-windows.html
The basic configuration order is "bottom-up":
- Select the build environment,
- Select the processor,
- Select the board,
- Select the supported peripherals
- Configure the device drivers,
- Configure the application options on top of this.
This is pretty straight forward for creating new configurations
but may be less intuitive for modifying existing configurations.
If you have an environment that supports the Qt or GTK graphical systems
(probably KDE or gnome, respectively), then you can also build the
graphical kconfig-frontends, kconfig-qconf and kconfig-gconf. In
these case, you can start the graphical configurator with either:
make qconfig
or
make gconfig
Refreshing Configurations with 'make oldconfig'
-----------------------------------------------
Whenever you use a configuration, you really should always do
the following *before* you make NuttX:
make oldconfig
This will make sure that the configuration is up-to-date in
the event that it has lapsed behind the current NuttX development.
WARNING: Never do 'make oldconfig' (OR 'make menuconfig') on a
configuration that has not been converted to use the kconfig-frontends
tools! This will damage your configuration (see
http://www.nuttx.org/doku.php?id=wiki:howtos:convertconfig).
Incompatibilities with Older Configurations
-------------------------------------------
***** WARNING *****
Gregory Nutt
committed
The current NuttX build system supports *only* the new configuration
files generated using the kconfig-frontends tools. The older, legacy,
manual configurations and the new kconfig-frontends configurations are
not compatible. Old legacy configurations can *not* be used with the
kconfig-frontends tool and, hence, cannot be used with recent releases
Gregory Nutt
committed
of NuttX:
If you run 'make menuconfig' with a legacy configuration the resulting
configuration will probably not be functional.
Q: How can I tell if a configuration is a new kconfig-frontends
configuration or an older, manual configuration?
Gregory Nutt
committed
A: Only old, manual configurations will have an appconfig file
Q: How can I convert a older, manual configuration into a new,
kconfig-frontends toolchain.
Gregory Nutt
committed
A: Refer to http://www.nuttx.org/doku.php?id=wiki:howtos:convertconfig
NuttX Configuration Tool under DOS
----------------------------------
Recent versions of NuttX support building NuttX from a native Windows
patacongo
committed
console window (see "Native Windows Build" below). But kconfig-frontends
is a Linux tool. At one time this was a problem for Windows users, but
now there is a specially modified version of the kconfig-frontends tools
that can be used:
http://uvc.de/posts/linux-kernel-configuration-tool-mconf-under-windows.html
It is also possible to use the version of kconfig-frontends built
under Cygwin outside of the Cygwin "sandbox" in a native Windows
environment:
1. You can run the configuration tool using Cygwin. However, the
Cygwin Makefile.win will complain so to do this will, you have
to manually edit the .config file:
a. Delete the line: CONFIG_WINDOWS_NATIVE=y
b. Change the apps/ directory path, CONFIG_APPS_DIR to use Unix
style delimiters. For example, change "..\apps" to "../apps"
And of course, after you use the configuration tool you need to
restore CONFIG_WINDOWS_NATIVE=y and the correct CONFIG_APPS_DIR.
2) You can, with some effort, run the Cygwin kconfig-mconf tool
patacongo
committed
directly in the Windows console window. In this case, you do not
have to modify the .config file, but there are other complexities:
a. You need to temporarily set the Cgywin directories in the PATH
variable then run kconfig-mconf manually like:
kconfig-mconf Kconfig
There is a Windows batch file at tools/kconfig.bat that automates
these steps:
tools/kconfig menuconfig
b. There is an issue with accessing DOS environment variables from
patacongo
committed
the Cygwin kconfig-mconf running in the Windows console. The
following change to the top-level Kconfig file seems to work
around these problems:
config APPSDIR
string
- option env="APPSDIR"
+ default "../apps"
TOOLCHAINS
^^^^^^^^^^
Cross-Development Toolchains
----------------------------
In order to build NuttX for your board, you will have to obtain a cross-
compiler to generate code for your target CPU. For each board,
configuration, there is a README.txt file (at configs/<board-name>/README.txt).
That README file contains suggestions and information about appropriate
tools and development environments for use with your board.
In any case, the script, setenv.sh that was deposited in the top-
level directory when NuttX was configured should be edited to set
the path to where you installed the toolchain. The use of setenv.sh
is optional but can save a lot of confusion in the future.
NuttX Buildroot Toolchain
-------------------------
For many configurations, a DIY set of tools is available for NuttX. These
tools can be downloaded from the NuttX SourceForge file repository. After
unpacking the buildroot tarball, you can find instructions for building
the tools in the buildroot/configs/README.txt file.
Check the README.txt file in the configuration director for your board
to see if you can use the buildroot toolchain with your board (this
README.txt file is located in configs/<board-name>/README.txt).
This toolchain is available for both the Linux and Cygwin development
environments.
Advantages: (1) NuttX header files are built into the tool chain,
and (2) related support tools like NXFLAT tools, the ROMFS
genromfs tools, and the kconfig-frontends tools can be built into your
toolchain.
Disadvantages: This tool chain is not was well supported as some other
toolchains. GNU tools are not my priority and so the buildroot tools
often get behind. For example, until recently there was no EABI support
in the NuttX buildroot toolchain for ARM.
NOTE: For Cortex-M3/4, there are OABI and EABI versions of the buildroot
toolchains. If you are using the older OABI toolchain the prefix for
the tools will be arm-nuttx-elf-; for the EABI toolchain the prefix will
be arm-nuttx-eabi-. If you are using the older OABI toolchain with
an ARM Cortex-M3/4, you will need to set CONFIG_ARMV7M_OABI_TOOLCHAIN
in the .config file in order to pick the right tool prefix.
patacongo
committed
If the make system ever picks the wrong prefix for your toolchain, you
can always specify the prefix on the command to override the default
like:
make CROSSDEV=arm-nuttx-elf
SHELLS
^^^^^^
The NuttX build relies on some shell scripts. Some are inline in the
Makefiles and many are executable scripts in the tools/. directory. The
scripts were all developed using bash and many contain bash shell
dependencies.
Most of the scripts begin with #!/bin/bash to specifically select the
bash shell. Some still have #!/bin/sh but I haven't heard any complaints
so these must not have bash dependencies.
There are two shell issues that I have heard of:
1. Linux where /bin/sh refers to an incompatible shell (like ksh or csh).
In this case, bash is probably available and the #!/bin/bash at the
beginning of the file should do the job. If any scripts with #!/bin/sh
fail, try changing that to #!/bin/bash and let me know about the change.
2. FreeBSD with the Bourne Shell and no bash shell.
The other, reverse case has also been reported on FreeBSD setups that
have the Bourne shell, but not bash. In this base, #!/bin/bash fails
but #!/bin/sh works okay. My recommendation in this case is to create
a symbolic link at /bin/bash that refers to the Bourne shell.
There may still be issues, however, with certain the bash-centric scripts
that will require modifications.
NuttX builds in-place in the source tree. You do not need to create
any special build directories. Assuming that your Make.defs is setup
properly for your tool chain and that setenv.sh contains the path to where
your cross-development tools are installed, the following steps are all that
are required to build NuttX:
cd ${TOPDIR}
. ./setenv.sh
make
At least one configuration (eagle100) requires additional command line
arguments on the make command. Read ${TOPDIR}/configs/<board-name>/README.txt
to see if that applies to your target.
Re-building is normally simple -- just type make again.
But there are some things that can "get you" when you use the Cygwin
development environment with Windows native tools. The native Windows
tools do not understand Cygwin's symbolic links, so the NuttX make system
does something weird: It copies the configuration directories instead of
linking to them (it could, perhaps, use the NTFS 'mklink' command, but it
doesn't).
A consequence of this is that you can easily get confused when you edit
a file in one of the linked (i.e., copied) directories, re-build NuttX,
and then not see your changes when you run the program. That is because
build is still using the version of the file in the copied directory, not
your modified file! To work around this annoying behavior, do the
following when you re-build:
This 'make' command will remove of the copied directories, re-copy them,
Build Targets and Options
-------------------------
patacongo
committed
Build Targets:
patacongo
committed
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
Below is a summary of the build targets available in the top-level
NuttX Makefile:
all
The default target builds the NuttX executable in the selected output
formats.
clean
Removes derived object files, archives, executables, and temporary
files, but retains the configuration and context files and directories.
distclean
Does 'clean' then also removes all configuration and context files.
This essentially restores the directory structure to its original,
unconfigured stated.
Application housekeeping targets. The APPDIR variable refers to the user
application directory. A sample apps/ directory is included with NuttX,
however, this is not treated as part of NuttX and may be replaced with a
different application directory. For the most part, the application
directory is treated like any other build directory in the Makefile script.
However, as a convenience, the following targets are included to support
housekeeping functions in the user application directory from the NuttX
build directory.
apps_clean
Perform the clean operation only in the user application directory
apps_distclean
Perform the distclean operation only in the user application directory.
The apps/.config file is preserved so that this is not a "full" distclean
but more of a configuration "reset."
export
The export target will package the NuttX libraries and header files into
an exportable package. Caveats: (1) These needs some extension for the KERNEL
build. (2) The logic in tools/mkexport.sh only supports GCC and, for example,
explicitly assumes that the archiver is 'ar'
download
This is a helper target that will rebuild NuttX and download it to the target
system in one step. The operation of this target depends completely upon
implementation of the DOWNLOAD command in the user Make.defs file. It will
generate an error an error if the DOWNLOAD command is not defined.
The following targets are used internally by the make logic but can be invoked
from the command under certain conditions if necessary.
depend
Create build dependencies. (NOTE: There is currently no support for build
dependencies under Cygwin using Windows-native toolchains.)
context
The context target is invoked on each target build to assure that NuttX is
properly configured. The basic configuration steps include creation of the
the config.h and version.h header files in the include/nuttx directory and
the establishment of symbolic links to configured directories.
clean_context
This is part of the distclean target. It removes all of the header files
and symbolic links created by the context target.
Build Options:
Of course, the value any make variable an be overridden from the make command
line. However, there is one particular variable assignment option that may
be useful to you:
V=1
patacongo
committed
This is the build "verbosity flag." If you specify V=1 on the make command
line, you will see the exact commands used in the build. This can be very
useful when adding new boards or tracking down compile time errors and
warnings (Contributed by Richard Cochran).
patacongo
committed
patacongo
committed
Native Windows Build
--------------------
The beginnings of a Windows native build are in place but still not often
used as of this writing. The windows native build logic initiated
if CONFIG_WINDOWS_NATIVE=y is defined in the NuttX configuration file:
patacongo
committed
This build:
- Uses all Windows style paths
- Uses primarily Windows batch commands from cmd.exe, with
patacongo
committed
In this build, you cannot use a Cygwin or MSYS shell. Rather the build must
patacongo
committed
be performed in a Windows console window. Here is a better terminal than the
standard issue, CMD.exe terminal: ConEmu which can be downloaded from:
patacongo
committed
http://code.google.com/p/conemu-maximus5/
Build Tools. The build still relies on some Unix-like commands. I use
the GNUWin32 tools that can be downloaded from http://gnuwin32.sourceforge.net/.
patacongo
committed
Host Compiler: I use the MingGW GCC compiler which can be downloaded from
patacongo
committed
http://www.mingw.org/. If you are using GNUWin32, then it is recommended
the you not install the optional MSYS components as there may be conflicts.
This capability should still be considered a work in progress because:
(1) It has not been verified on all targets and tools, and
(2) it still lacks some of the creature-comforts of the more mature environments.
There is an alternative to the setenv.sh script available for the Windows
native environment: tools/configure.bat. See tools/README.txt for additional
information.
Installing GNUWin32
-------------------
The Windows native build will depend upon a few Unix-like tools that can be
provided either by MSYS or GNUWin32. The GNUWin32 are available from
http://gnuwin32.sourceforge.net/. GNUWin32 provides ports of tools with a
GPL or similar open source license to modern MS-Windows (Microsoft Windows
2000 / XP / 2003 / Vista / 2008 / 7). See
http://gnuwin32.sourceforge.net/packages.html for a list of all of the tools
available in the GNUWin32 package.
The SourceForge project is located here:
http://sourceforge.net/projects/gnuwin32/. The project is still being
actively supported (although some of the Windows ports have gotten very old).
Some commercial toolchains include a subset of the GNUWin32 tools in the
installation. My recommendation is that you download the GNUWin32 tools
directly from the sourceforge.net website so that you will know what you are
using and can reproduce your build environment.
The following steps will download and execute the GNUWin32 installer.
1. Download GetGNUWin32-x.x.x.exe from
http://sourceforge.net/projects/getgnuwin32/files/. This is the
installer. The current version as of this writing is 0.6.3.
4. Select the installation directory. My recommendation is the
directory that contains this README file (<this-directory>).
5. After running GetGNUWin32-0.x.x.exe, you will have a new directory
<this-directory>/GetGNUWin32
Note that the GNUWin32 installer didn't install GNUWin32. Instead, it
installed another, smarter downloader. That downloader is the GNUWin32
package management tool developed by the Open SSL project.
The following steps probably should be performed from inside a DOS shell.
6. Change to the directory created by GetGNUWin32-x.x.x.exe
7. Execute the download.bat script. The download.bat script will download
about 446 packages! Enough to have a very complete Linux-like environment
under the DOS shell. This will take awhile. This step only downloads
the packages and the next step will install the packages.
8. This step will install the downloaded packages. The argument of the
install.bat script is the installation location. C:\gnuwin32 is the
standard install location:
NOTE: This installation step will install *all* GNUWin32 packages... far
more than you will ever need. If disc space is a problem for you, you might
need to perform a manual installation of the individual ZIP files that you
will find in the <this directory>/GetGNUWin32/packages directory.
patacongo
committed
Strange Path Problems
---------------------
patacongo
committed
If you see strange behavior when building under Cygwin then you may have
a problem with your PATH variable. For example, if you see failures to
locate files that are clearly present, that may mean that you are using
the wrong version of a tool. For example, you may not be using Cygwin's
'make' program at /usr/bin/make. Try:
When you install some toolchains (such as Yargarto or CodeSourcery tools),
they may modify your PATH variable to include a path to their binaries.
At that location, they make have GNUWin32 versions of the tools. So you
might actually be using a version of make that does not understand Cygwin
paths.
The solution is either:
1. Edit your PATH to remove the path to the GNUWin32 tools, or
2. Put /usr/local/bin, /usr/bin, and /bin at the front of your path:
$ export PATH=/usr/local/bin:/usr/bin:/bin:$PATH
patacongo
committed
Window Native Toolchain Issues
------------------------------
patacongo
committed
There are many popular Windows native toolchains that may be used with NuttX.
Examples include CodeSourcery (for Windows), devkitARM, and several vendor-
provied toolchains. There are several limitations with using a and Windows
based toolchain in a Cygwin environment. The three biggest are:
1. The Windows toolchain cannot follow Cygwin paths. Path conversions are
performed automatically in the Cygwin makefiles using the 'cygpath' utility
but you might easily find some new path problems. If so, check out 'cygpath -w'
2. Windows toolchains cannot follow Cygwin symbolic links. Many symbolic links
are used in Nuttx (e.g., include/arch). The make system works around these
problems for the Windows tools by copying directories instead of linking them.
But this can also cause some confusion for you: For example, you may edit
a file in a "linked" directory and find that your changes had no effect.
That is because you are building the copy of the file in the "fake" symbolic
directory. If you use a Windows toolchain, you should get in the habit of
making like this:
make clean_context all
An alias in your .bashrc file might make that less painful. The rebuild
is not a long as you might think because there is no dependency checking
if you are using a native Windows toolchain. That bring us to #3:
3. Dependencies are not made when using Windows versions of the GCC on a POSIX
platform (i.e., Cygwin). This is because the dependencies are generated
using Windows paths which do not work with the Cygwin make.
patacongo
committed
MKDEP = $(TOPDIR)/tools/mknulldeps.sh
patacongo
committed
If you are building natively on Windows, then no such conflict exists
and the best selection is:
patacongo
committed
patacongo
committed
General Pre-built Toolchain Issues
To continue with the list of "Window Native Toolchain Issues" we can add
the following. These, however, are really just issues that you will have
if you use any pre-built toolchain (vs. building the NuttX toolchain from
the NuttX buildroot package):
There may be incompatibilities with header files, libraries, and compiler
built-in functions at detailed below. For the most part, these issues
are handled in the existing make logic. But if you are breaking new ground,
patacongo
committed
4. Header Files. Most pre-built toolchains will build with a foreign C
library (usually newlib, but maybe uClibc or glibc if you are using a
patacongo
committed
Linux toolchain). This means that the header files from the foreign
C library will be built into the toolchain. So if you "include <stdio.h>",
you will get the stdio.h from the incompatible, foreign C library and
not the nuttx stdio.h (at nuttx/include/stdio.h) that you wanted.
This can cause really confusion in the builds and you must always be
patacongo
committed
sure the -nostdinc is included in the CFLAGS. That will assure that
patacongo
committed
5. Libraries. What was said above header files applies to libraries.
You do not want to include code from the libraries of any foreign
C libraries built into your toolchain. If this happens you will get
perplexing errors about undefined symbols. To avoid these errors,
patacongo
committed
you will need to add -nostdlib to your CFLAGS flags to assure that
you only take code from the NuttX libraries.
This, however, may causes other issues for libraries in the toolchain
that you do want (like libgcc.a or libm.a). These are special-cased
in most Makefiles, but you could still run into issues of missing
libraries.
6. Built-Ins. Some compilers target a particular operating system.
Many people would, for example, like to use the same toolchain to
develop Linux and NuttX software. Compilers built for other
operating systems may generate incompatible built-in logic and,
for this reason, -fno-builtin should also be included in your
C flags
And finally you may not be able to use NXFLAT.
7. NXFLAT. If you use a pre-built toolchain, you will lose all support
for NXFLAT. NXFLAT is a binary format described in
patacongo
committed
Documentation/NuttXNxFlat.html. It may be possible to build
standalone versions of the NXFLAT tools; there are a few examples
of this in the misc/buildroot/configs directory. However, it
is possible that there could be interoperability issues with
your toolchain since they will be using different versions of
patacongo
committed
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
Building Original Linux Boards in Cygwin
Some default board configurations are set to build under Linux and others
to build under Windows with Cygwin. Various default toolchains may also
be used in each configuration. It is possible to change the default
setup. Here, for example, is what you must do in order to compile a
default Linux configuration in the Cygwin environment using the
CodeSourceery for Windows toolchain. After instantiating a "canned"
NuttX configuration, run the target 'menuconfig' and set the following
items:
Build Setup->Build Host Platform->Windows
Build Setup->Windows Build Environment->Cygwin
System Type->Toolchain Selection->CodeSourcery GNU Toolchain under Windows
In Windows 7 it may be required to open the Cygwin shell as Administrator
("Run As" option, right button) you find errors like "Permission denied".
Recovering from Bad Configurations
Many people make the mistake of configuring NuttX with the "canned"
configuration and then just typing 'make' with disastrous consequences;
the build may fail with mysterious, uninterpretable, and irrecoverable
build errors. If, for example, you do this with an unmodified Linux
configuration in a Windows/Cgwin environment, you will corrupt the
build environment. The environment will be corrupted because of POSIX vs
Windows path issues and with issues related to symbolic links. If you
make the mistake of doing this, the easiest way to recover is to just
start over: Do 'make distclean' to remove every trace of the corrupted
configuration, reconfigure from scratch, and make certain that the set
the configuration correctly for your platform before attempting to make
again.
Just fixing the configuration file after you have instantiated the bad
configuration with 'make' is not enough.
Additional information can be found in the Documentation/ directory and
also in README files that are scattered throughout the source tree. The
documentation is in HTML and can be access by loading the following file
into your Web browser:
Documentation/index.html
NuttX documentation is also available online at http://www.nuttx.org.
Below is a guide to the available README files in the NuttX source tree:
| | `- src
| | `- lpc214x/README.txt
| |- sh/
| | |- include/
| | | `-README.txt
| | |- src/
| | | `-README.txt
| |- x86/
| | |- include/
| | | `-README.txt
| | `- src/
| | `-README.txt