[Toybox] NDK 28-beta2 still doesn't understand -static-libasan

[enh]

On Mon, Dec 2, 2024 at 5:15 PM Rob Landley <rob at landley.net> wrote:

> On 12/2/24 14:18, enh wrote:
> > i asked an expert, and they said...
> >
> > ```
> > $ external/toolchain-utils/llvm_tools/git_llvm_rev.py --sha
> > dcf70e16684420ae211dad58dbfacc0430409ab4
> > r542442
> > ```
> > So r542442 is the merged SHA for that patch, but r28 is only using
> r530567
> > based on https://github.com/android/ndk/wiki/Changelog-r28
> >
> > back to me, i'll add that it's not in the "current next" toolchain
> either:
> >
> > ~/aosp-main-with-phones/prebuilts/clang/host/linux-x86$ ls -1d clang-*
> > clang-3289846
> > clang-r522817
> > clang-r530567
> > clang-r530567b
> > clang-r536225
> > clang-stable
> > ~/aosp-main-with-phones/prebuilts/clang/host/linux-x86$
>
> Alas, when I wired around it locally the static build still died with
> "ld.lld: error: undefined symbol: _DYNAMIC" so not immediately useful
> anyway.
>
> I had regression testing the NDK's ASAN on my todo list since
> https://landley.net/notes-2024.html#27-08-2024 but presumably you're
> already doing it in that context.


yeah, hwasan for arm64 devices and asan for x86-64 CI bots.


> And I have no idea how to fix debian's
> ASAN...
>

install clang instead of gcc? "works for me" on ubuntu at least :-)


> I need a test failure category that's "unhappy but sort of expected",
> but there's no useful automated information to extract from that one.
> Sort of a "still not king" indicator
> (https://www.ealasaid.com/misc/vsd/aragorn.html).
>
> Rob
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.landley.net/pipermail/toybox-landley.net/attachments/20241202/92e4c415/attachment.htm>