On Sat, Aug 19, 2017 at 09:49:25AM -0400, Alan Ott wrote:
On 08/19/2017 06:37 AM, Leif Lindholm wrote:
(Adding linaro-uefi, since this is not official tooling.)
Ok, no problem.
On Fri, Aug 18, 2017 at 07:29:59PM -0400, Alan Ott wrote:
edk2-build.sh | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/edk2-build.sh b/edk2-build.sh index 60da4df..fb0cb84 100755 --- a/edk2-build.sh +++ b/edk2-build.sh @@ -73,11 +73,13 @@ function do_build import_openssl fi
- set_cross_compile
- CROSS_COMPILE="$TEMP_CROSS_COMPILE"
- if [ -z $CROSS_COMPILE ]; then
set_cross_compileCROSS_COMPILE="$TEMP_CROSS_COMPILE"- fi echo "Building $PLATFORM_NAME - $PLATFORM_ARCH"
- echo "CROSS_COMPILE="$TEMP_CROSS_COMPILE""
- echo "CROSS_COMPILE="$CROSS_COMPILE"" echo "$board"_BUILDFLAGS="'$PLATFORM_BUILDFLAGS'" if [ "$TARGETS" == "" ]; then
Sorry, can't accept this one.
I use this script to build multiple platforms across multiple architectures in one go, and this change breaks that.
I see now that it's in do_build() which is called for each platform built, so yes, my patch won't work. Sorry :(
But also, do you really need it? Cross compilers accessible on the PATH should be automatically detected.
Yes, I can do this.[1]
If you do need it, because you're using some non-standard toolchain (such as aarch64-none-eabi), I would like to see something like what exists (in semi-broken form) in uefi-build.sh: CROSS_COMPILE_32 vs CROSS_COMPILE_64. However, in order to be useful, that would need to be extended to do a per-architecture override: CROSS_COMPILE_AARCH64 CROSS_COMPILE_ARM CROSS_COMPILE_IA32 CROSS_COMPILE_X64
Yes, I was incorrectly thinking it was intended to be similar to uefi-build.sh and didn't look closely enough.
Sorry for the extra noise,
Alan.
[1] For what it's worth I've gotten into the habit of specifying the full path in CROSS_COMPILE when cross compiling things which build this way (kernels, u-boots, etc). It makes typos fail early and operates independently of my actual PATH, preventing a mistyped or missing path entry from causing it to use a compiler down the list in the system or user path.
Sure, it's a valid thing to do. Especially if testing across different versions of toolchains. Just, I tend to do that by flipping a symlink. If someone was to implement what I suggest above, I would be happy to take it.
And sorry for mistyping the address to my own mailing list :|
/ Leif