On Thu, Jul 11, 2019 at 12:39:52AM +0000, Hans Malissa via blfs-support wrote: > > Thanks a lot. Unfortunately I didn't log the tests, which would have been > really helpful. I will build it again, and log both the build and the tests. > I will also try llvm-8.0 and do the same. > I'm not sure if I understand the connection between LLVM and rustc; I have > already built and installed rustc on my system (following > http://linuxfromscratch.org/blfs/view/stable-systemd/general/rust.html); > neither rustc nor LLVM list each other as a dependency. > Greetings, > > Hans
As a general rule, if you are taking the time to run tests then you really need to log them. A few packages end up with 'All passed' if everythign is ok, but even there details of problems can be much earlier if there are any failures. More generally, packages spit the test results out in batches, e.g. run tests in one directory, report, then repeat in another directory. When I updated rustc to 1.32.0 we were using LLVM-7 and that version of rust was horribly flakey - for some people - with system LLVM. Also noted on Arch, eventually it became clear that it needed a newer LLVM (and 8.0.0 had not been released, so we used the version of LLVM which was included within the rust source). Now that firefox-68 needs a newer rust we've had to again update. But the system LLVM-8.0.0 seems to be good enough so we can revert to using system llvm when building rust (to save time and space, assuming people have already installed LLVM). I say "seems to be good enough" because I've only used the system version on a couple of machines where I have LLVM-8, the other machines with older systems are using LLVM-7 so I've again built rust with its shipped LLVM on those. And if I didn't need to build LLVM for the AMD part on my radeon|amdgpu machines, and with clang etc to use to build rust, I would do my best to avoid LLVM ;-) Specifically, clang's reported strengths differ from gcc's and it still lacks the ability to compile the linux kernel (on x86) and its strengthening features are considered to lag behind gcc, as well as its options being different (hit one the day before yesterday in testing, it didn't like a -malign-data option I was trying, fixed by telling that package to use gcc and g++). OTOH, mozilla go out of their way to favour clang, so in a future version of firefox we might again need to use it. ĸen -- This is magic for grown-ups; it has to be hard because we know there's no such thing as a free goblin. -- Pratchett, Stewart & Cohen - The Science of Discworld II -- http://lists.linuxfromscratch.org/listinfo/blfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
