On 2/21/19 1:03 AM, Pierre Labastie via blfs-dev wrote:
On 20/02/2019 23:07, Bruce Dubbs via blfs-dev wrote:
On 2/20/19 3:34 PM, Douglas R. Reno via blfs-dev wrote:
On 2/20/19 12:24 AM, Bruce Dubbs via blfs-dev wrote:
On 2/19/19 10:40 PM, Douglas R. Reno via blfs-dev wrote:
On 2/19/19 10:12 PM, Bruce Dubbs via blfs-dev wrote:
I've run into a problem trying to validate SM and FF for 8.4, First SM.
It appears that SM prefers clang over gcc. The build fails early because
it uses clang by default and the sanity check fails because the CFLAGS we
pass are incompatible with clang.
I added CC=gcc CXX=g++ and SM built. It seems to run OK, so I've added
those variables to the book with a note that says if you want to use
clang, don't set the flags.
The problem with FF is more serious. At the 29 minute point I get:
27:48.76 Compiling gkrust v0.1.0
(/tmp/firefox/firefox-65.0/toolkit/library/rust)
29:17.59 error: Could not compile `gkrust`.
29:17.60 Caused by:
... (long command line here)
(signal: 11, SIGSEGV: invalid memory reference)
We use clang for FF so I tried gcc and got the same error.
The funny thing is that I built FF a week ago with exactly the same
instructions (the same script) on my Skylake workstation (basically a
less complete 8.4 system) and it is fine. My development system is a
Haswell. Both are running rustc 1.32.0.
The only thing I could find on google is
https://aur.archlinux.org/packages/icecat/?comments=all
but it doesn't seem to have a solution.
Any ideas?
I'm going to have to build Firefox so that I can configure CUPS probably
tonight or tomorrow. I'll get back to you with the results off that,
especially if we have the same problem.
I'm beginning to think it might be a CPU HW issue, but it's mostly
speculation. Can you check it out on your Haswell?
-- Bruce
I can't confirm this with my Skylake.
0:20.37 make: Leaving directory
'/sources/firefox-65.0.1/firefox-65.0.1/firefox-build-dir'
0:20.38 /usr/bin/notify-send --app-name=Mozilla Build System Mozilla Build
System Install complete
'/usr/lib/firefox/browser/plugins' -> '../../mozilla/plugins'
'firefox.desktop' -> '/usr/share/applications/firefox.desktop'
'/usr/share/pixmaps/firefox.png' ->
'/usr/lib/firefox-65.0.1/browser/icons/mozicon128.png'
5792.9 Elasped Time - firefox-65.0.1
SBU=47.482
270332 /sources/firefox-65.0.1.source.tar.xz size (263.996 MB)
12421704 kilobytes build size (12130.570 MB)
md5sum : 38c01a14e58cce894dda513739bd15d3 /sources/firefox-65.0.1.source.tar.xz
sha1sum: 5b3a64d5c843f4e23cb3220a3211077bc58cba34
/sources/firefox-65.0.1.source.tar.xz
It works fine on it.
I got it to work. I rebuilt rustc and then FF built.
To really start over, I renamed ~/.cargo.
At first I used the wrong script and installed rust in /usr.
The original script had
[llvm]
targets = "X86"
link-shared = true
[build]
extended = true
[install]
prefix = "/usr"
docdir = "share/doc/rustc-$VER"
[rust]
channel = "stable"
rpath = false
codegen-tests = false
[target.x86_64-unknown-linux-gnu]
llvm-config = "/usr/bin/llvm-config"
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
This is what generated the error: You used system llvm,
which is broken. That's why it was removed.
-----
The working script has
[llvm]
ninja = true
targets = "X86"
experimental-targets = ""
[build]
docs = false
extended = true
[install]
prefix = "/opt/rustc-$VER"
docdir = "share/doc/rustc-$VER"
[rust]
channel = "stable"
rpath = false
codegen-tests = false
backtrace-on-ice = true
------
The old script has a sed for Cargo.lock and did not have the option
--exclude src/tools/miri
My best guess as to the problem is the original build did not use
ninja = true
or that it did use
link-shared = true
That could be, but the main reason is system llvm
Thank you for the reminder. I completely forgot. Curiously, using
system llvm was good enough for librsvg and cbinggen.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
