On 03.11.2019 18:43, Kamil Rytarowski wrote: > On 03.11.2019 18:35, Joerg Sonnenberger wrote: >> On Sun, Nov 03, 2019 at 03:21:51PM +0100, Kamil Rytarowski wrote: >>> On 24.10.2019 11:23, Kamil Rytarowski wrote: >>>> On 24.10.2019 01:20, Robert Swindells wrote: >>>>> >>>>> Kamil Rytarowski <n54%gmx.com@localhost> wrote: >>>>>> On 23.10.2019 19:28, Robert Swindells wrote: >>>>>>> >>>>>>> There is a restriction in our ld.elf_so that it can only load two >>>>>>> segments. >>>>>> >>>>>> Please test the patch from: >>>>>> >>>>>> http://mail-index.netbsd.org/tech-toolchain/2019/01/06/msg003368.html >>>>> >>>>> The patch works for me. >>>>> >>>>> Would something like this get into NetBSD-9 ? >>>>> >>>> >>>> Unless we will get a patch with a better version, I plan to land it >>>> after 2 weeks and pull-request to -9. >>>> >>>> BTW. Please share your rebased version. >>>> >>> >>> Ping? >> >> It's on my list once the LLVM update is done. ...which is primarily >> waiting on UBSAN now. >> >> Joerg >> > > Micro-UBSan is done, if there are still any incompatibilities please report. > For tests, our Clang+LLD patches (the ones for the starting point of LLD on NetBSD) are merged into pkgsrc/devel/lld an pkgsrc/lang/clang. While there, are there plans to upgrade libc++? If so, I have got a strong suggestion to import headers into canonical LLVM path "c++/v1/". It is now supported by the Clang NetBSD driver. The customized path caused a lot of pain and need for workarounds/hacks in dealing with external toolchain for no gain.
Attachment:
signature.asc
Description: OpenPGP digital signature