Clean up the no and unknown file extension info since only .h and .c
files supported. Just ignore them quietly.
Signed-off-by: liuhaitao <liuhaitao@xiaomi.com>
* tools/Config.mk - Add empty definition POSTBUILD
* tools/Makefile.unix/win - Replace CXD56xx specific logic with $(call POSTBUILD)
* boards/arm/cxd56xx/scripts/Config.mk - Add POSTBUILD definitions with logic removed from Makefile.unix/win
* boards/arm/cxd56xx/spresense/scripts/Make.defs - Include the CXD56xx Config.mk immediately after tools/Config.mk so that the empty POSTBUILD definition will be replaced with the correct one
NOTE: There is a precedent for this approach. This is the way that other architecture-specific build options are implemented. See, for example, tools/zds/Config.mk
If we are going to use this tool for things like
git pre-commit hook or CI workflow, it's critical to
avoid false alarms.
This reverts commit de764af9aa.
Squashed commit of the following:
commit 685951b5385062035ac558df1112353c9441c910
Author: Adam Feuer <adam@starcat.io>
Date: Mon Feb 24 16:33:00 2020 -0800
fixed typo in readme
commit ad1d3289b48af0de3095e3f365429017e57278b3
Author: Adam Feuer <adam@starcat.io>
Date: Mon Feb 24 16:07:48 2020 -0800
simulator host route helper script and docs
Squashed commit of the following:
commit cf5cddcf55a155303cb5abb1aa026f6dcaf369ca
Author: Adam Feuer <adam@starcat.io>
Date: Mon Feb 24 15:50:40 2020 -0800
syslog on console to fix compile error
commit def1bb73fcfbc8b115c55d9f4544b97d583807f2
Author: Adam Feuer <adam@starcat.io>
Date: Mon Feb 24 15:46:13 2020 -0800
formatting as per PR feedback; removed ping
commit b179fd8831b77fbbe85527a4ab3161e1f1ca1e43
Author: Adam Feuer <adam@starcat.io>
Date: Mon Feb 24 15:45:49 2020 -0800
formatting as per PR feedback
commit e3280bede9798d9a00b118e126c02ceab497e33a
Author: Adam Feuer <adam@starcat.io>
Date: Mon Feb 24 13:56:50 2020 -0800
add simhostroute.sh description to readme
commit 09a6b0ca3bacf005c2a79102141ae8cc3eb91849
Author: Adam Feuer <adam@starcat.io>
Date: Sun Feb 23 17:38:40 2020 -0800
fixed error in simhostroute.sh usage docs
commit d838582119d43ee2002ce3808051c82b23e98c58
Author: Adam Feuer <adam@starcat.io>
Date: Sun Feb 23 17:28:35 2020 -0800
added tcpblaster defconfig and updated docs
commit af3d2d6591f12d1127027fdd363858052094e624
Author: Adam Feuer <adam@starcat.io>
Date: Sun Feb 23 17:21:29 2020 -0800
added info about capabilities and running as root
commit 6359cfdfedefc818b169455401942d3d33a59f41
Author: Adam Feuer <adam@starcat.io>
Date: Sun Feb 23 17:20:42 2020 -0800
updated network linux readme
commit 47feb08aa74e4b1fc6c802567bc777d31e7c9a83
Author: Adam Feuer <adam@starcat.io>
Date: Sun Feb 23 17:09:04 2020 -0800
initial addition
This tools needs to CD to directory holding the library because it requires that the library be in the current working directory when objects are inserted. Previously, the tool did this by creating an absolute path to the object file itself. So instead of the object file being in the current directory and the library in a directory at an absolute path, this tool used an abolte path to the object file with the library in the current directory.
For the most part, that worked fine. However, in the apps directory the object file names are greatly extended in length. In the object file name was extended to a full absolute path and these huge file names were were exceeding some internal limit within the ZDS-II librarian. But useing shorter, relative paths, we were able to avoid hitting this limit and the entire build completed correctly.
* tools/zds/zds_Config.mk: Move boards/z80/ez80/scripts/eZ80_Config.mk to tools/zds/zds_Config.mk where it can be shared by other ZDS-II platforms.
* boards/z16/z16f: Duplicate changes for new ZDS-II support from ez80
* boards/z80/z8: Duplicate changes for new ZDS-II support from ez80
* arch/z16/src/z16f/Toolchain.defs: Create required Toolchain.defs file for Z16f
Eliminate the MOVEOBJ definition from tools/Config.mk, Makefiles, and other locations. The MOVEOBJ definition was used to work around the case where the object output of a compile or assemble is in a different directory than the source file. The ZDS-II tools have no command line option to perform this move; MOVEOBJ handled the move as a post-compiler operation.
MOVEOBJ is no longer needed because this the ez80 COMPILE and ASSEMBLE definitions now handle all of the object file movement cases.
tools/zds/zdsgen.c: Remove this program. It should have been a simpler way to manage ZDS-II compiler and assember differences. However, the compiler arguments needed to pass through too many layers of bash command line modifications and, as a result, it was not possible to retain quotes on critical strings. This approch was abandoned for an alternative approach.
boards/z80/ez80/scripts/eZ80_Config.mk: Add definitions to move the object files as necessary. This seems to work well and is implemented for both native and Cygwin Windows build -- but only for the Z20x.
And additional wrapper that is needed for the ZDS-II build is a wrapper for the compiler and assemble. This is needed because the ZDS-II cannot control the name or location of the output of the output object file: it is always in the same directory as the source file and with the same name as the source file except with the .obj extension.
This was handled in the past with the MOVEOBJ definition which was specifically called Makefiles to move the objects to the correct position. However, now there is a new behavior: Output object files may also be named differently with added decoration in the file name. This is done in the current apps/ directory build.
There is currently some ugly implementation that includes a long sequence of Bash code to handle the moving and/or renaming. There is nothing in place for the Windows native case.
This wrapper, when complete, will clean up the Makefiles, improve build performance, and provide a solution that is portable to both the Cywin/MSYS environment as well as to the Windows native environment.
This zdsar librarian wrapper is code complete and has no known bugs.
However, there are still some build instabilities so perhaps there
are still some bug to be identified. Or perhaps there are other
undiscovered bugs unrelated to zdsar which appears to be behaving
correctly in all cases.
It will take some additional testing know for sure.
A pointer to a string in a memory buffer was losing its value. The reason was that the buffer was occasionally being used for other purposes. The fix is to strdup() the string so that there is a private, protected copy.
tools/zds: A new tool sub-directory intended to hold tools for making life working with the ZDS-II toolchain less painful.
tools/zds/zdsar.c: This is a wrapper around the ZDS_II librarian. It simplifies the build scripts by replacing large sequences of complex Bash script that were added to the build files. Not only does this clean up the build files but it also improves performance and, more importantly, provides a common solution for the Windows native build case. This tool should work with all ZDS-II based platforms including z8, zNeo, and ez80.
tools/README.txt: Add a brief description about the zds sub-directory. Also re-ordered some tool descriptions. They are supposed to be in alphabetical order, but this seems to have fallen apart.
boards/z80/ez80/scripts/eZ80_Config.mk: Updated to use tools/zds/zdsar.exe.
This includes some fixes for the ez80 build including:
1. Ability to handle decorated object file names,
2. Ability to handle libraries described with absolute paths
This commit address only the POSIX build and needs to be updated
for the Windows native build.
to avoid shell complain that @echo can't find if the variable expand in the compound
command(e.g. ARLOCK) and remove $(Q) before $AR so the result is almost same as before.
To ensure size_t same as toolchain definition in the first place and rename CXX_NEWLONG to ARCH_SIZET_LONG. The change also check whether __SIZE_TYPE__ exist before CONFIG_ARCH_SIZET_LONG so our definition can align with toolchain(gcc/clang) definition automatically.
Make 'checkpatch.sh -c commits' and 'checkpatch.sh -p patches' do full
file check as default, while 'checkpatch.sh -r -c commits' and 'checkpatch.sh
-r -p patches' do ranges check only.
Eliminate warnings. Ignore mixed case identifies beginning with PRIx.
These most likely come from inttypes.h and we nxstyle must tolerate
those definitions even though they do not follow the coding style.
Namely this changes HOSTOS for macOS from "Other" to "Darwin".
Also, suppress the following harmless messages during a "make":
(Missing redirect in tools/Makefile.host)
uname: illegal option -- o
usage: uname [-amnprsv]
Consider junk.c for example:
/****************************************************************************
* xx
****************************************************************************/
/****************************************************************************
* Private Types
****************************************************************************/
struct foo_s
{
int bar;
};
/****************************************************************************
* Public Functions
****************************************************************************/
int dofoo(int barin)
{
barout = barin;
return barout;
}
nxstyle not detects these problems:
$ tools/nxstyle.exe junk.c
junk.c:11:0: error: Blank line before opening left brace
junk.c:21:0: error: Blank line before opening left brace
This capability was erroneously removed by a previous commit. This setting is necessary when the C++ code generated by the compiler uses the new and delete operators provided in libs/libxx.
Here is an example of an error scenario where this capability is required:
1. CONFIG_CXX_NEWLONG=y in the defconfig file
2. This causes libs/libxx to to generate a delete operator of this form:
void delete(void*, unsigned long);
3. But the compiler expects to link with a delete operator of this form:
void delete(void*, unsigned int);
That results in a link-time failure and the ONLY solution for avoiding that link time failure (short of some complete redesign) is to have a control over the CONFIG_CXX_NEWLONG setting and make sture that it is always disabled for that toolchain.
Other toolchains will require that CONFIG_CXX_NEWLONG=y in all cases. This cannot be predicted because it is dependndent on the definitions provided by C libraries and header files that were build into binary toolchain.
CI should know the build status of testbuild.sh, so use -k make flag option as default
instead of -i ignore-error option. Then use fail to record build result. In addition,
refresh .config if toolchain updated.
Change-Id: I182c2b2db489e6ccb0a79fdc664072d19974c3ca
Signed-off-by: liuhaitao <liuhaitao@xiaomi.com>
1.Scan the folder under nuttx/boards by prefix /
2.Skip the specified configuration by prefix -
Change-Id: Ibb67f2c6b96cd8e439d94bf71d098c1fd9e5a79d
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Make sure kconfig-tweak disables CONFIG_TOOLCHAIN_WINDOWS under linux host, or it would break the build when using testbuild.sh for some configs.
Signed-off-by: liuhaitao <liuhaitao@xiaomi.com>
1. Can't pipe uname stderr output to /dev/null. /dev/null does not exist in the Windows native enviornment.
2. Don't test if CONFIG_WINDOWS_NATIVE is defined. Tht leads to a "chicken'n'egg" problem: We need to build configure.c in order to configure the system, but we can't get the CONFIG_WINDOWS_NATIVE until after the system is configured.
3. The default name of the MinGW GCC compiler is mingw32-gcc.exe, not mingw-gcc.exe
Commit cf5d17f795 added logic to detect #define pre-processor definitions outside of the "Pre-processor Definitions" file section. That commit was verified against numerous .c source files. But not against any .h header files.
When run against a header file, that change causes a false alarm warning like:
file:line:pos: warning: #define outside of 'Pre-processor Definitions' section
That is caused the idempotence, guard definition that must appear in the header file BEFORE the first file section.
This commit adds logic to nxstyle to ignore pre-processor definitions in header files that occur before the first file section is encountered.
1. No .c file should include a "Public Types" section. Only a header file can define a public type. A .c file can only define a private type. Several files contained private type definitions. The section that they were defined in, however, was incorrectly named "Public Types." Those were easilty changed to "Private Types" which is what they are.
2. No .c file should include a "Public Function Prototypes" section. All global function prototypes should be provided via a header file and never declared with a .c file.
For No. 2, I corrected as many cases as was reasonable for the time that I had available. But there are still a dozen or so .c files that declare "Public Function Prototypes" within a .c file. This is bad programming style. These declarations should all be moved to the proper header files.
* tools/nxstyle: Added logic to parse section headers (like Included files, Pre-processor Definitions, etc.) and to assure that the section headers are correct for the file type. Also (1) verify that #include appears only in the 'Included Files' section and that (2) #define only occurs in the Pre-processor definition section.
Right now, there are several places where that rule is not followed. I think most of these are acceptable so these failures only generate warnings, not errors. The warning means that we need to go look at the offending #define or #include and decide if it is a acceptable usage or not.
