Re: CVS commit: src [llvm]
Le 16/11/2019 à 23:01, Joerg Sonnenberger a écrit : On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : Module Name:src Committed By: joerg Date: Mon Nov 11 22:45:32 UTC 2019 [...] Log Message: Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) Since this change, I cannot cross-compile GENERIC amd64 from Linux: Problem for GCC 7 in NetBSD itself should be fixed. Please check if it works for your version too or otherwise report your GCC version. cross-compilation with GCC 9.2.1 on Linux still doesn't work
Re: CVS commit: src [llvm]
On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: > Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : > > Module Name:src > > Committed By: joerg > > Date: Mon Nov 11 22:45:32 UTC 2019 > > > > [...] > > Log Message: > > Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) > > Since this change, I cannot cross-compile GENERIC amd64 from Linux: Problem for GCC 7 in NetBSD itself should be fixed. Please check if it works for your version too or otherwise report your GCC version. Joerg
Re: CVS commit: src [llvm]
On Thu, Nov 14, 2019 at 01:11:02PM +, Robert Swindells wrote: > The 18446744073709551615 value is . > > Have built with MKLLVMRT=no for now. I'm about to commit a workaround for that to unbreak that build (watiting for final test build to complete). Martin
Re: CVS commit: src [llvm]
Building aarch64 using gcc I get this: In file included from /u1/src/external/apache2/llvm/librt/libLLVMAnalysis/../../ lib/../dist/llvm/include/llvm/IR/Attributes.h:22, from /u1/src/external/apache2/llvm/librt/libLLVMAnalysis/../../ lib/../dist/llvm/include/llvm/IR/CallSite.h:31, from /u1/src/external/apache2/llvm/librt/libLLVMAnalysis/../../ lib/../dist/llvm/include/llvm/Analysis/ValueTracking.h:20, from /u1/src/external/apache2/llvm/librt/libLLVMAnalysis/../../ lib/../dist/llvm/lib/Analysis/ValueTracking.cpp:14: In member function 'size_t llvm::StringRef::find(char, size_t) const', inlined from 'bool llvm::getConstantStringInfo(const llvm::Value*, llvm::Str ingRef&, uint64_t, bool)' at /u1/src/external/apache2/llvm/librt/libLLVMAnalysis /../../lib/../dist/llvm/lib/Analysis/ValueTracking.cpp:3580:39: /u1/src/external/apache2/llvm/librt/libLLVMAnalysis/../../lib/../dist/llvm/inclu de/llvm/ADT/StringRef.h:309:37: error: 'void* memchr(const void*, int, size_t)' specified size 18446744073709551615 exceeds maximum object size 9223372036854775 807 [-Werror=stringop-overflow=] if (const void *P = ::memchr(Data + FindBegin, C, Length - FindBegin)) ^ The 18446744073709551615 value is . Have built with MKLLVMRT=no for now.
Re: CVS commit: src [llvm]
On Wed, Nov 13, 2019 at 01:25:19PM +0100, Thomas Klausner wrote: > On Wed, Nov 13, 2019 at 12:30:52PM +0100, Tobias Nygren wrote: > > On Wed, 13 Nov 2019 10:29:52 +0100 > > Maxime Villard wrote: > > > > > Am I the only one seeing this? It's hard to blame the configuration when > > > the > > > failure is seen on several hardware and several OSes. > > > > FWIW I'm typing this from an x86_64 machine running today's LLVM > > userland and kernel. I have some local patches but none that should be > > relevant. > > > > I have a different build problem: That's another bogus unspecific GCC warning. Most lovely is that it doesn't even have a flag for disabling it. Joerg
Re: CVS commit: src [llvm]
On Wed, Nov 13, 2019 at 10:29:52AM +0100, Maxime Villard wrote: > Le 12/11/2019 à 23:10, Joerg Sonnenberger a écrit : > > On Tue, Nov 12, 2019 at 04:31:00PM +0100, Maxime Villard wrote: > > > Le 12/11/2019 à 16:06, Maxime Villard a écrit : > > > > Le 12/11/2019 à 14:37, Joerg Sonnenberger a écrit : > > > > > On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: > > > > > > Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : > > > > > > > Module Name: src > > > > > > > Committed By: joerg > > > > > > > Date: Mon Nov 11 22:45:32 UTC 2019 > > > > > > > > > > > > > > [...] > > > > > > > Log Message: > > > > > > > Update LLVM to 10.0.0git > > > > > > > (01f3a59fb3e2542fce74c768718f594d0debd0da) > > > > > > > > > > > > Since this change, I cannot cross-compile GENERIC amd64 from Linux: > > > > > > > > > > What non-standard settings are you using? amd64 is certainly on my > > > > > list > > > > > of architectures to test. > > > > > > > > No particular settings. Just the latest src/ compiled unchanged with > > > > > > > > ./build.sh -j4 -U -O ../obj -V MKLLVM=yes -V MKGCC=no -V HAVE_LLVM=yes > > > > -m amd64 kernel=GENERIC > > > > > > I'm also having the same issue on a NetBSD-amd64 host, with again an > > > unchanged > > > src/ and the same command > > > > Is that a clean build? I certainly do full release builds for all LLVM > > architectures and I don't get this. > > Yes, a clean build. > > I've done two again just now, on different physical machines. One has NetBSD, > the other has Linux. At least on NetBSD 9-BETA, it is a miscompile by GCC 7.4.0. Don't know what your Linux version has. Joerg
Re: CVS commit: src [llvm]
On Tue, 12 Nov 2019 23:10:45 +0100 Joerg Sonnenberger wrote: > On Tue, Nov 12, 2019 at 04:31:00PM +0100, Maxime Villard wrote: > > Le 12/11/2019 à 16:06, Maxime Villard a écrit : > > > Le 12/11/2019 à 14:37, Joerg Sonnenberger a écrit : > > > > On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: > > > > > Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : > > > > > > Module Name: src > > > > > > Committed By: joerg > > > > > > Date: Mon Nov 11 22:45:32 UTC 2019 > > > > > > > > > > > > [...] > > > > > > Log Message: > > > > > > Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) > > > > > > > > > > Since this change, I cannot cross-compile GENERIC amd64 from Linux: > > > > > > > > What non-standard settings are you using? amd64 is certainly on my list > > > > of architectures to test. > > > > > > No particular settings. Just the latest src/ compiled unchanged with > > > > > > ./build.sh -j4 -U -O ../obj -V MKLLVM=yes -V MKGCC=no -V HAVE_LLVM=yes -m > > > amd64 kernel=GENERIC > > > > I'm also having the same issue on a NetBSD-amd64 host, with again an > > unchanged > > src/ and the same command > > Is that a clean build? I certainly do full release builds for all LLVM > architectures and I don't get this. I was able to reproduce the problem by building in a NetBSD-9 chroot with a GCC host toolchain, but not with my -current LLVM host toolchain. I wonder if GCC miscompiles something.
Re: CVS commit: src [llvm]
On Wed, Nov 13, 2019 at 12:30:52PM +0100, Tobias Nygren wrote: > On Wed, 13 Nov 2019 10:29:52 +0100 > Maxime Villard wrote: > > > Am I the only one seeing this? It's hard to blame the configuration when the > > failure is seen on several hardware and several OSes. > > FWIW I'm typing this from an x86_64 machine running today's LLVM > userland and kernel. I have some local patches but none that should be > relevant. > I have a different build problem: # build.sh -j 32 -x -V MKDEBUG=yes -V MKDEBUGLIB=yes -V MKLLVM=yes -T /usr/obj/tools.gcc -m amd64 -O /usr/obj/src.amd64 -D /usr/obj/ amd64.gcc.20191113 -R /usr/obj/amd64.gcc.20191113.release distribution In file included from /usr/src/external/apache2/llvm/lib/libclangParse/../../dist/clang/include/clang/Parse/Parser.h:25, from /usr/src/external/apache2/llvm/lib/libclangParse/../../dist/clang/lib/Parse/ParseCXXInlineMethods.cpp:13: /usr/src/external/apache2/llvm/lib/libclangParse/../../dist/clang/include/clang/Sema/Sema.h:2566:42: error: 'clang::Sema::DefaultedFunctionKind::Comparison' is too small to hold all values of 'enum class clang::Sema::DefaultedComparisonKind' [-Werror] DefaultedComparisonKind Comparison : 8; ^ Clean checkout from an hour or two ago. Thomas
Re: CVS commit: src [llvm]
On Wed, 13 Nov 2019 10:29:52 +0100 Maxime Villard wrote: > Am I the only one seeing this? It's hard to blame the configuration when the > failure is seen on several hardware and several OSes. FWIW I'm typing this from an x86_64 machine running today's LLVM userland and kernel. I have some local patches but none that should be relevant.
Re: CVS commit: src [llvm]
Le 12/11/2019 à 23:10, Joerg Sonnenberger a écrit : On Tue, Nov 12, 2019 at 04:31:00PM +0100, Maxime Villard wrote: Le 12/11/2019 à 16:06, Maxime Villard a écrit : Le 12/11/2019 à 14:37, Joerg Sonnenberger a écrit : On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : Module Name: src Committed By: joerg Date: Mon Nov 11 22:45:32 UTC 2019 [...] Log Message: Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) Since this change, I cannot cross-compile GENERIC amd64 from Linux: What non-standard settings are you using? amd64 is certainly on my list of architectures to test. No particular settings. Just the latest src/ compiled unchanged with ./build.sh -j4 -U -O ../obj -V MKLLVM=yes -V MKGCC=no -V HAVE_LLVM=yes -m amd64 kernel=GENERIC I'm also having the same issue on a NetBSD-amd64 host, with again an unchanged src/ and the same command Is that a clean build? I certainly do full release builds for all LLVM architectures and I don't get this. Yes, a clean build. I've done two again just now, on different physical machines. One has NetBSD, the other has Linux. I deleted the src and obj directories and started from scratch using github, with the following commands exactly as-is on both hosts from an empty folder: git clone https://github.com/NetBSD/src.git cd src ./build.sh -j 4 -U -O ../obj -V MKGCC=no -V MKLLVM=yes -V HAVE_LLVM=yes -m amd64 tools ./build.sh -j 4 -U -O ../obj -V MKGCC=no -V MKLLVM=yes -V HAVE_LLVM=yes -m amd64 kernel=GENERIC Both hosts fail the same way. My two tests from yesterday were on the same physical machine but with both NetBSD/Linux, using the same commands as above except that it was anoncvs instead of github, and both hosts failed the same way. Am I the only one seeing this? It's hard to blame the configuration when the failure is seen on several hardware and several OSes.
Re: CVS commit: src [llvm]
On Tue, Nov 12, 2019 at 04:31:00PM +0100, Maxime Villard wrote: > Le 12/11/2019 à 16:06, Maxime Villard a écrit : > > Le 12/11/2019 à 14:37, Joerg Sonnenberger a écrit : > > > On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: > > > > Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : > > > > > Module Name: src > > > > > Committed By: joerg > > > > > Date: Mon Nov 11 22:45:32 UTC 2019 > > > > > > > > > > [...] > > > > > Log Message: > > > > > Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) > > > > > > > > Since this change, I cannot cross-compile GENERIC amd64 from Linux: > > > > > > What non-standard settings are you using? amd64 is certainly on my list > > > of architectures to test. > > > > No particular settings. Just the latest src/ compiled unchanged with > > > > ./build.sh -j4 -U -O ../obj -V MKLLVM=yes -V MKGCC=no -V HAVE_LLVM=yes -m > > amd64 kernel=GENERIC > > I'm also having the same issue on a NetBSD-amd64 host, with again an unchanged > src/ and the same command Is that a clean build? I certainly do full release builds for all LLVM architectures and I don't get this. Joerg
Re: CVS commit: src [llvm]
Le 12/11/2019 à 16:06, Maxime Villard a écrit : Le 12/11/2019 à 14:37, Joerg Sonnenberger a écrit : On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : Module Name: src Committed By: joerg Date: Mon Nov 11 22:45:32 UTC 2019 [...] Log Message: Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) Since this change, I cannot cross-compile GENERIC amd64 from Linux: What non-standard settings are you using? amd64 is certainly on my list of architectures to test. No particular settings. Just the latest src/ compiled unchanged with ./build.sh -j4 -U -O ../obj -V MKLLVM=yes -V MKGCC=no -V HAVE_LLVM=yes -m amd64 kernel=GENERIC I'm also having the same issue on a NetBSD-amd64 host, with again an unchanged src/ and the same command
Re: CVS commit: src [llvm]
Le 12/11/2019 à 14:37, Joerg Sonnenberger a écrit : On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : Module Name:src Committed By: joerg Date: Mon Nov 11 22:45:32 UTC 2019 [...] Log Message: Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) Since this change, I cannot cross-compile GENERIC amd64 from Linux: What non-standard settings are you using? amd64 is certainly on my list of architectures to test. No particular settings. Just the latest src/ compiled unchanged with ./build.sh -j4 -U -O ../obj -V MKLLVM=yes -V MKGCC=no -V HAVE_LLVM=yes -m amd64 kernel=GENERIC
Re: CVS commit: src [llvm]
On Tue, Nov 12, 2019 at 11:39:09AM +0100, Maxime Villard wrote: > Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : > > Module Name:src > > Committed By: joerg > > Date: Mon Nov 11 22:45:32 UTC 2019 > > > > [...] > > Log Message: > > Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) > > Since this change, I cannot cross-compile GENERIC amd64 from Linux: What non-standard settings are you using? amd64 is certainly on my list of architectures to test. > Preprocessed source(s) and associated run script(s) are located at: > x86_64--netbsd-clang: note: diagnostic msg: /tmp/assym-5fcf44.c > x86_64--netbsd-clang: note: diagnostic msg: /tmp/assym-5fcf44.sh Both files would otherwise be a starting point... Joerg
Re: CVS commit: src [llvm]
Le 11/11/2019 à 23:45, Joerg Sonnenberger a écrit : Module Name:src Committed By: joerg Date: Mon Nov 11 22:45:32 UTC 2019 [...] Log Message: Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da) Since this change, I cannot cross-compile GENERIC amd64 from Linux: 1. /netbsd-llvm/src/sys/sys/time.h:112:14: current parser token ';' 2. /netbsd-llvm/src/sys/sys/time.h:109:1: parsing function body 'bintime_addx' 3. /netbsd-llvm/src/sys/sys/time.h:109:1: in compound statement ('{}') /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x3cdd568] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x3cdb774] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x3cdb8d5] /lib64/libpthread.so.0(+0x12e70)[0x7fe616827e70] /lib64/libc.so.6(gsignal+0x10f)[0x7fe61634e57f] /lib64/libc.so.6(abort+0x127)[0x7fe616338895] /lib64/libc.so.6(+0x22769)[0x7fe616338769] /lib64/libc.so.6(+0x30a26)[0x7fe616346a26] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1b8f672] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1b8e6c7] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1b8e89d] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1b8e95a] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1b8eba2] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x15f496d] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x15f549f] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x15a8848] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x15ab06f] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x15aec92] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x15af1a6] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11e6741] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11e1a80] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11e3d91] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11e3fd4] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11e5c49] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1221858] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1fe] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1222ae9] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x1227012] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x122941b] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11a74e3] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11c6487] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11a2a81] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11a3081] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11a7e48] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11a8ac8] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x11a058a] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0xcb3178] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x4e9c99] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x4ab006] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x48a72f] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x47d21a] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x4120c4] /lib64/libc.so.6(__libc_start_main+0xf3)[0x7fe61633a413] /netbsd-llvm/obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin/x86_64--netbsd-clang[0x47a10e] x86_64--netbsd-clang: error: unable to execute command: Aborted (core dumped) x86_64--netbsd-clang: error: clang frontend command failed due to signal (use -v to see invocation) clang version 9.0.0 (01f3a59fb3e2542fce74c768718f594d0debd0da) Target: x86_64-unknown-netbsd Thread model: posix InstalledDir: /netbsd-llvm/src/../obj/tooldir.Linux-5.3.6-100.fc29.x86_64-x86_64/bin x86_64--netbsd-clang: note: diagnostic msg: PLEASE submit a bug report to http://llvm.org/bugs/ and include the crash backtrace, preprocessed source, and as