On 8/2/2024 1:46 AM, Ard Biesheuvel wrote:
On Fri, 2 Aug 2024 at 00:39, Oliver Smith-Denny
<o...@linux.microsoft.com> wrote:
CompilerIntrinsicsLib and ArmSoftFloatLib add ARM/AARCH64 compiler
intrinsics and floating point functions required by OpenSSL,
respectively. CompilerIntrinsicsLib is used almost in every DSC that
builds ARM/AARCH64 and ArmSoftFloatLib is used by every DSC that builds
logic from OpenSSL.
What's even worse: OpenSSL doesn't really use floating point at all -
the only reason we need this is because there is some global double_t
somewhere that records entropy or somesuch, but IIRC that code is not
actually used.
So perhaps we could add a hack in OpensslLib to work around this in
one way or another for ARM only, and get rid of SoftFloatLib
altogether.
I took a quick look at this. Unfortunately, I don't think this is
trivial. There are some uses of doubles within OpenSSL and it
doesn't look like we can just not build part of it, because
the first error I hit is that crypto/params.c uses fp operations:
https://github.com/openssl/openssl/blob/de90e54bbe82e5be4fb9608b6f5c308bb837d355/crypto/params.c#L762
and this provides a lot of definitions for other OpenSSL code. A
quick search turns up a number of other instances, so for now I will
go with moving the submodule, unless you want to take a look at
OpenSSL.
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#120237): https://edk2.groups.io/g/devel/message/120237
Mute This Topic: https://groups.io/mt/107675828/21656
Group Owner: devel+ow...@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-
