bpringlemeir at nbsps.com
Thu Jan 10 15:45:29 PST 2013
On 10 Jan 2013, bpringlemeir at nbsps.com wrote:
>> hpa at zytor.com wrote:
>> We should figure out why the output of "make headers_install" doesn't
>> work, and fix it.
> Here is another guy,
> Anyways, I found that passing a relative path for $(INSTALLROOT) is a
> disaster. The build doesn't fail if you do this...
> - $(Q)$(MAKE) -C $(KLIBCKERNELSRC) ARCH=$(KLIBCARCH)
> INSTALL_HDR_PATH=$(INSTALLROOT)$(INSTALLDIR)/$(KCROSS) headers_install
> + $(Q)$(MAKE) -C $(KLIBCKERNELSRC) ARCH=$(KLIBCARCH) INSTALL_HDR_PATH=$(abspath $(INSTALLROOT)$(INSTALLDIR)/$(KCROSS)) headers_install
Sorry, I hadn't read the whole list from November.
There are two sets of advice on how to compile,
1. 'make help'
2. 'README' -> '/usr/klibc/README.klibc'
I see advice to set 'KLIBCKERNELSRC' to a target of a 'make
headers_install' from linux on the mailing list. See the 'Compilation
errors' thread at,
Obviously the 'make -C <header_path>' is not going to work. Why are we
making a symlink to 'headers_install' and then going to run the same
make command again to generate the same files in a different place?
Can't we just do a 'cp' if the KLIBCKERNELSRC is already a
'headers_install' version? Maybe the Kbuild.install should check for a
'Makefile' in KLIBCKERNELSRC and if it exists, do the 'headers_install'
or else just do a 'cp'. At the very least a check for a Makefile and a
warning might be good to stop people pointing the build to the wrong
Currently the 'make help' says,
KLIBCKERNELSRC - Path to a configured linux tree
I think that text should change? This implies a kernel source tree as
does the make variable name.
More information about the klibc