Closed glaubitz closed 7 years ago
On 03/16/2017 09:01 AM, Laurent Vivier wrote:
Just removing the "inline" from the console.h is enough.
I discussed this with @karcherm and he said, the original reasoning of the inlining was to accelerate console_putstring() which it would still do with my current change. For external translation units, the function would be exported normally.
We are in the case when we are not on a m68k machine and we don't have a cross-compiler, so this is an error case, we really must exit.
I agree. However, the problem is that the machine detection is not bullet-proof and may fail, so we will just end up with a half-generated tools.mk which was the original reason we had problems building the documentation.
I think it's better to remove the constant NULL comparison.
Can we be 100% sure that the pointer passed is never NULL?
tools are supposed to be built and installed on another arch than m68k. Did you test your change in a cross-build environment?
Do you mean the changes regarding make install
in particular?
The problem here is that the binaries are built into $(TARGET), but make install
as it is currently used here does not support that.
This PR contains multiple changes that were required to make the package build in a current Debian unstable chroot. Tested with qemu-m68k-user.