arm: 0.950
TCG: 0.928
performance: 0.927
i386: 0.918
architecture: 0.918
user-level: 0.833
PID: 0.831
debug: 0.789
permissions: 0.776
ppc: 0.775
device: 0.757
graphic: 0.755
register: 0.738
risc-v: 0.730
network: 0.722
peripherals: 0.707
semantic: 0.696
VMM: 0.683
x86: 0.663
files: 0.655
vnc: 0.621
hypervisor: 0.612
kernel: 0.612
KVM: 0.602
socket: 0.580
assembly: 0.570
boot: 0.558
mistranslation: 0.503
virtual: 0.494

Regressions on arm target with some GCC tests

Hi,

After trying qemu master:
commit 474f3938d79ab36b9231c9ad3b5a9314c2aeacde
Merge: 68d7ff0 14f5d87
Author: Peter Maydell <email address hidden>
Date:   Fri Jun 21 15:40:50 2019 +0100

I found several regressions compared to qemu-3.1 when running the GCC testsuite.
I'm attaching a tarball containing several GCC tests (binaries), needed shared libs, and a short script to run all the tests.

All tests used to pass w/o error (one of them is verbose), but with a recent qemu, all of them make qemu crash:

qemu: uncaught target signal 6 (Aborted) - core dumped

This was noticed with GCC master configured with
--target arm-none-linux-gnueabi
--with-mode arm
--with-cpu cortex-a9

and calling qemu with --cpu cortex-a9 (the script uses "any", this makes no difference).

I have noticed other failures with arm-v8 code, but this is probably the same root cause. Since it's a bit tedious to manually rebuild & extract the testcases, I'd prefer to start with this subset, and I can extract more if needed later.

Thanks



I bisected a chunk of the errors to:

  commit c6fb8c0cf704c4a1a48c3e99e995ad4c58150dab (refs/bisect/bad)
  Author: Richard Henderson <email address hidden>
  Date:   Mon Feb 25 11:42:35 2019 -0800

      tcg/i386: Support INDEX_op_extract2_{i32,i64}

      Signed-off-by: Richard Henderson <email address hidden>

Specifically I think when tcg_gen_deposit_i32 handles the if (ofs + len == 32) case.


Fixed by:

Subject: [PATCH for-4.1] tcg: Fix constant folding of INDEX_op_extract2_i32
Date: Tue,  9 Jul 2019 14:19:00 +0200
Message-Id: <email address hidden>


I confirm this patch fixes the problem I reported. Thanks!