Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 15 August 2017 at 08:56, Diana Picus via cfe-commits < cfe-commits@lists.llvm.org> wrote: > Ugh, I'm not sure that last email made it through. > > Here's the IR with -Xclang -disable-llvm-passes, at r310400 and r310401: > https://goo.gl/4n84SR > https://goo.gl/gxhxp9 > > The diff between them

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Ugh, I'm not sure that last email made it through. Here's the IR with -Xclang -disable-llvm-passes, at r310400 and r310401: https://goo.gl/4n84SR https://goo.gl/gxhxp9 The diff between them is small enough so I don't think there's much point in reducing it further. Hope that helps. On 15 August

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 15 August 2017 at 01:25, Richard Smith wrote: > On 14 August 2017 at 03:27, Diana Picus via cfe-commits > wrote: >> >> Hi, >> >> Strangely enough, it turns out that if I run >> Asan-armhf-with-calls-Noinst-Test on the command line it fails,

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 14 August 2017 at 03:27, Diana Picus via cfe-commits < cfe-commits@lists.llvm.org> wrote: > Hi, > > Strangely enough, it turns out that if I run > Asan-armhf-with-calls-Noinst-Test on the command line it fails, > although it doesn't fail when run with lit. Looks like the crash is within the

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi, Ok that's a progress. Could you figure out which optimizer pass breaks it. I bet it is the inliner. I assume we could run a reducer on the ll file. Cheers, Vassil On 14/08/17 17:21, Diana Picus wrote: Hi, I didn't manage to reproduce this at -O0. Yes, I think the version in 1.8.0,

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi, I didn't manage to reproduce this at -O0. Yes, I think the version in 1.8.0, slightly modified (see llvm/utils/unittest/googletest/README.LLVM) On 14 August 2017 at 17:06, Vassil Vassilev wrote: > On 14/08/17 15:59, Diana Picus wrote: >> >> No, the buildbots don't

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 14/08/17 15:59, Diana Picus wrote: No, the buildbots don't build with -O0 (at least not the ones that broke). The command line for that particular object is: build/./bin/clang -fPIC -fvisibility-inlines-hidden -Werror=date-time -std=c++11 -Wall -W -Wno-unused-parameter -Wwrite-strings

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 14 August 2017 at 16:59, Diana Picus wrote: > No, the buildbots don't build with -O0 (at least not the ones that broke). > > The command line for that particular object is: > build/./bin/clang -fPIC -fvisibility-inlines-hidden -Werror=date-time > -std=c++11 -Wall -W

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

No, the buildbots don't build with -O0 (at least not the ones that broke). The command line for that particular object is: build/./bin/clang -fPIC -fvisibility-inlines-hidden -Werror=date-time -std=c++11 -Wall -W -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wmissing-field-initializers

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 14/08/17 15:08, Vassil Vassilev wrote: On 14/08/17 13:04, Diana Picus wrote: See attached. Thanks! It looks like asan_test.i doesn't have gtest.cc which appears in the stack trace. Am I missing something? Could you paste the compiler invocation. Are we building with -O0 (I see quite a

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 14/08/17 13:04, Diana Picus wrote: See attached. Thanks! It looks like asan_test.i doesn't have gtest.cc which appears in the stack trace. Am I missing something? On 14 August 2017 at 13:30, Vassil Vassilev wrote: On 14/08/17 11:27, Diana Picus wrote: Hi,

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Also if you want the disassembly for the whole test executable (with just that test in it): https://goo.gl/pjULbN It's 177MB though. On 14 August 2017 at 14:04, Diana Picus wrote: > See attached. > > On 14 August 2017 at 13:30, Vassil Vassilev

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 14/08/17 11:27, Diana Picus wrote: Hi, Strangely enough, it turns out that if I run Asan-armhf-with-calls-Noinst-Test on the command line it fails, although it doesn't fail when run with lit. I've attached the stack trace from gdb. It looks like some trouble passing down va_arg parameters,

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi Diana, On 14/08/17 11:27, Diana Picus wrote: Hi, Strangely enough, it turns out that if I run Asan-armhf-with-calls-Noinst-Test on the command line it fails, although it doesn't fail when run with lit. I've attached the stack trace from gdb. It looks like some trouble passing down va_arg

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi, Strangely enough, it turns out that if I run Asan-armhf-with-calls-Noinst-Test on the command line it fails, although it doesn't fail when run with lit. I've attached the stack trace from gdb. It looks like some trouble passing down va_arg parameters, but I haven't looked into too much

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Well, these are ASAN tests, I'm not sure how that would interact with Valgrind. Anyway, I'll try to reproduce the environment, I'm guessing it would be best to catch this in gdb so I can actually see what's going on. On 11 August 2017 at 15:21, Vassil Vassilev wrote: >

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

That's really strange. It looks like some random behavior. Did you run some memory checker like valgrind? Do the environment provided by the test runner and yours match? Sent from my phone. Please excuse my brevity. > On 11 Aug 2017, at 15:58, Diana Picus wrote: > >

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi again, I finally got the debug build, but unfortunately the stack traces that the tests print look the same. My suspicion is that this is because the addresses printed by the tests are funny (i.e. odd numbers instead of divisible by 4). I tried to follow those addresses in an objdump of the

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi guys, I'm SO sorry about the delays. I've been having all sorts of trouble getting that debug build on the board (from ld running out of memory to the board just crashing on me, in which case I need to ask someone else to reboot it because I can't power cycle it remotely). I can assure you

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi Diana, Thanks for helping us out! Cheers, Vassil On 10/08/17 11:28, Diana Picus wrote: Hi Vassil, My build is in progress, but since it's a full build it's probably going to take another couple of hours to complete. I'll let you know when it's done. Thanks, Diana On 10 August 2017 at

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi Vassil, My build is in progress, but since it's a full build it's probably going to take another couple of hours to complete. I'll let you know when it's done. Thanks, Diana On 10 August 2017 at 10:09, Vassil Vassilev wrote: > It looks like I can not reproduce it on

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

It looks like I can not reproduce it on osx (non-arm)... :( On 09/08/17 22:54, Diana Picus wrote: Reverting this also fixed the selfhost bots: http://lab.llvm.org:8011/builders/clang-cmake-thumbv7-a15-full-sh/builds/2142

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

On 09/08/17 22:54, Diana Picus wrote: Reverting this also fixed the selfhost bots: http://lab.llvm.org:8011/builders/clang-cmake-thumbv7-a15-full-sh/builds/2142 http://lab.llvm.org:8011/builders/clang-cmake-armv7-a15-selfhost/builds/2309

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Reverting this also fixed the selfhost bots: http://lab.llvm.org:8011/builders/clang-cmake-thumbv7-a15-full-sh/builds/2142 http://lab.llvm.org:8011/builders/clang-cmake-armv7-a15-selfhost/builds/2309 http://lab.llvm.org:8011/builders/clang-cmake-armv7-a15-selfhost-neon/builds/1819 I'm afraid the

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi Diana, It seems the service is down. Could you send us the details of the failures (incl stack traces if any) Many thanks, Vassil On 09/08/17 15:27, Diana Picus via cfe-commits wrote: Hi Richard, I'm sorry but I've reverted this in r310464 because it was breaking some ASAN tests on

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

Hi Richard, I'm sorry but I've reverted this in r310464 because it was breaking some ASAN tests on this bot: http://lab.llvm.org:8011/builders/clang-cmake-armv7-a15-full/builds/9452 Please let me know if I can help debug this. Cheers, Diana On 8 August 2017 at 21:14, Richard Smith via

Re: r310401 - PR19668, PR23034: Fix handling of move constructors and deleted copy

I forgot to say: Based on a patch by Vassil Vassilev, which was based on a patch by Bernd Schmidt, which was based on a patch by Reid Kleckner. On 8 August 2017 at 12:12, Richard Smith via cfe-commits < cfe-commits@lists.llvm.org> wrote: > Author: rsmith > Date: Tue Aug 8 12:12:28 2017 > New