[ CCed all people who were involved in this thread ]
Hi Tom,
personally, I am interested to test the prebuilt-toolchains for
Ubuntu/xenial alias 16.04 LTS and Debian/Jessie v8.5.0 AMD64.
The available toolchains are incomplete and thus useless.
Just as a fact: There is still no compiler-rt source-tarball available.
I know of Linux distributions which refuse to build and upload new
packages when there are no official tarballs around.
Can you please - as the release-manager of LLVM v3.8.1 - care of
uploading missing and/or incomplete source-tarballs and/or
prebuilt-toolchains?
( "Care" in the sense of you know whom to poke. )
Thanks in advance.
Hi,
I am working on this now, I'll let you know when I'm finished.
I have seen there is a compiler-rt source-tarball available, now.
Not sure if all hash-files were updated with corresponding sig-files.
I compared all my selfmade source-tarballs with the official ones.
compiler-rt seems to get some updates (MIPS msan sanityzer) after
tagging, maybe that's the reason for the delay?
I compared all my selfmade source-tarballs with the official ones.
compiler-rt seems to get some updates (MIPS msan sanityzer) after
tagging, maybe that's the reason for the delay?
Thanks.
- Sedat -
The official tarball, the final tag, and the branch all have identical contents. llvm-branch-commits did receive some commit emails after the tag but their revision numbers are less than the revision that was tagged (r273407). I think the mailing list just delivered the emails late.
I am playing now with the optimization and speedup builds of a
selfmade llvm-toolchain.
In a 2nd build I could build a LLVMgold plugin (before I built a
binutils v2.26.1 and used all ar|nm|ld.bfd|ld.gold, placed LLVMgold.so
into /path/to/binutils/lib/bfd-plugins/ - with binutils v2.22 shipped
with Ubuntu/precise I could not build a LLVMgold.so).
My 3rd build with backported lto-flag from LLVM v3.9 is still running,
so this will be LTO-optimized.
Let's see PGO, optimized-TableGen and split-DWARF in my 4th build.