Date: Fri, 16 May 2014 12:36:09 +0200 From: Dimitry Andric <dim@FreeBSD.org> To: jhellenthal@dataix.net Cc: hackers@freebsd.org Subject: Re: stable/10 r266159 base clang 3.4 Assertion on nul \0 Message-ID: <1BE05E62-27FE-4D51-B028-2F926BD4E5C8@FreeBSD.org> In-Reply-To: <20140516044950.GA37592@DataIX.net> References: <20140516044950.GA37592@DataIX.net>
next in thread | previous in thread | raw e-mail | index | archive | help
[-- Attachment #1 --] On 16 May 2014, at 06:49, jhellenthal@dataix.net wrote: > > On powerpc maybe others, I believe I am hitting a bug in clang34 that is part of base. After seearching around it has lead me to (llvm/tools/clang/lib/Frontend/TextDiagnostic.cpp) line 973ish code block where I believe it never checks for the condition of no new line at end of file but I am unfamiliar with the code and could use someone to take a closer look when they have the time. > > This is the link that lead me there: (Says resolved so I believe that to be another issue) > http://comments.gmane.org/gmane.comp.compilers.llvm.bugs/24021 > > This is the error I see repeated through compilation of different programs: > clang -c -I. -I../Src -I../Src -I../Src/Zle -I. -I/usr/local/include -DLIBICONV_PLUG -I/usr/local/include -DHAVE_CONFIG_H -O2 -pipe -DLIBICONV_PLUG -fno-strict-aliasing -o builtin.o builtin.c > UNREACHABLE executed! > Stack dump: > 0. Program arguments: /usr/bin/clang -cc1 -triple powerpc-unknown-freebsd10.0 -emit-obj -disable-free -disable-llvm-verifier -main-file-name builtin.c -mrelocation-model static -mdisable-fp-elim -relaxed-aliasing -masm-verbose -mconstructor-aliases -target-cpu ppc -coverage-file /export/usr/obj/export/usr/ports/shells/zsh/work/zsh-5.0.5/Src/builtin.o -resource-dir /usr/bin/../lib/clang/3.4 -D LIBICONV_PLUG -D HAVE_CONFIG_H -D LIBICONV_PLUG -I . -I ../Src -I ../Src -I ../Src/Zle -I . -I /usr/local/include -I /usr/local/include -O2 -fdebug-compilation-dir /export/usr/obj/export/usr/ports/shells/zsh/work/zsh-5.0.5/Src -ferror-limit 19 -fmessage-length 94 -mstackrealign -fno-signed-char -fobjc-runtime=gnustep -fdiagnostics-show-option -fcolor-diagnostics -vectorize-loops -vectorize-slp -o builtin.o -x c builtin.c > 1. <eof> parser at end of file > 2. Code generation > 3. Running pass 'Function Pass Manager' on module 'builtin.c'. > 4. Running pass 'PowerPC DAG->DAG Pattern Instruction Selection' on function '@bin_print' > clang: error: unable to execute command: Abort trap > clang: error: clang frontend command failed due to signal (use -v to see invocation) > FreeBSD clang version 3.4 (tags/RELEASE_34/final 197956) 20140216 > Target: powerpc-unknown-freebsd10.0 > Thread model: posix This is something completely different than http://llvm.org/PR16339 which you referred to. You seem to be running into a problem in the pass manager, which is part of the optimizer. Please try reproducing it with clang trunk, and if it still happens there, try reducing the testcase to a minimal form. Or submit the generated .sh and .c file in a new LLVM PR. If it doesn't reproduce on trunk, send me the generated .sh and .c file, CC'ing rdivacky@FreeBSD.org (as he is definitely more knowledgeable about PowerPC than me :). -Dimitry [-- Attachment #2 --] -----BEGIN PGP SIGNATURE----- Version: GnuPG/MacGPG2 v2.0.22 (Darwin) iEYEARECAAYFAlN16iIACgkQsF6jCi4glqNdgQCgzGl2BzU/wx7H05shJ0QnF4QV V7cAn0+YCgeoLjrbeOyidRgY6aY7mMPc =bs9b -----END PGP SIGNATURE-----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1BE05E62-27FE-4D51-B028-2F926BD4E5C8>