MSan issues when using lld |
||||||
Issue descriptionThis happened when webrtc tried to roll chromium/build past the lld switch: https://webrtc-review.googlesource.com/c/src/+/7180 I tracked it down to a bug in lld which is fixed as of LLVM r315119. We should disable lld when using msan in order to avoid any further false positives until we can roll past that revision.
,
Oct 6 2017
,
Oct 6 2017
Thank you!
,
Oct 18 2017
We rolled clang -- time to revert that CL?
,
Oct 18 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/855199c7ee2b2dce192aff80e97d8e6604726e67 commit 855199c7ee2b2dce192aff80e97d8e6604726e67 Author: Peter Collingbourne <pcc@chromium.org> Date: Wed Oct 18 13:24:43 2017 Revert "Disable lld when building with msan for now." This reverts commit 91f0b5cbdd5acfecd2184056c164b477a732aa9b. Reason for revert: lld bug is now fixed. Original change's description: > Disable lld when building with msan for now. > > Due to an lld bug which causes false positives in msan. > > Bug: 772559 > Change-Id: I27a735d838a773763879f293f5f6a6c4790de782 > Reviewed-on: https://chromium-review.googlesource.com/706339 > Reviewed-by: Nico Weber <thakis@chromium.org> > Reviewed-by: Dirk Pranke <dpranke@chromium.org> > Commit-Queue: Peter Collingbourne <pcc@chromium.org> > Cr-Commit-Position: refs/heads/master@{#507229} TBR=thakis@chromium.org,dpranke@chromium.org,pcc@chromium.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: 772559 Change-Id: I4d5c3dfa1499aba30f7116195fd9e1721d5d2bd7 Reviewed-on: https://chromium-review.googlesource.com/725239 Reviewed-by: Peter Collingbourne <pcc@chromium.org> Reviewed-by: Nico Weber <thakis@chromium.org> Cr-Commit-Position: refs/heads/master@{#509757} [modify] https://crrev.com/855199c7ee2b2dce192aff80e97d8e6604726e67/build/config/compiler/compiler.gni
,
Oct 18 2017
,
Oct 20 2017
The revert appears to have caused issue 776355 and issue 776628 , so I'll revert it for now.
,
Oct 20 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/e01826fa133837ba45e7730054f50bc06a5254e5 commit e01826fa133837ba45e7730054f50bc06a5254e5 Author: Peter Collingbourne <pcc@chromium.org> Date: Fri Oct 20 21:31:57 2017 Revert "Revert "Disable lld when building with msan for now."" This reverts commit 855199c7ee2b2dce192aff80e97d8e6604726e67. Reason for revert: Caused issue 776355 and issue 776628 Bug: 776355 , 776628 , 772559 Original change's description: > Revert "Disable lld when building with msan for now." > > This reverts commit 91f0b5cbdd5acfecd2184056c164b477a732aa9b. > > Reason for revert: lld bug is now fixed. > > Original change's description: > > Disable lld when building with msan for now. > > > > Due to an lld bug which causes false positives in msan. > > > > Bug: 772559 > > Change-Id: I27a735d838a773763879f293f5f6a6c4790de782 > > Reviewed-on: https://chromium-review.googlesource.com/706339 > > Reviewed-by: Nico Weber <thakis@chromium.org> > > Reviewed-by: Dirk Pranke <dpranke@chromium.org> > > Commit-Queue: Peter Collingbourne <pcc@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#507229} > > TBR=thakis@chromium.org,dpranke@chromium.org,pcc@chromium.org > > # Not skipping CQ checks because original CL landed > 1 day ago. > > Bug: 772559 > Change-Id: I4d5c3dfa1499aba30f7116195fd9e1721d5d2bd7 > Reviewed-on: https://chromium-review.googlesource.com/725239 > Reviewed-by: Peter Collingbourne <pcc@chromium.org> > Reviewed-by: Nico Weber <thakis@chromium.org> > Cr-Commit-Position: refs/heads/master@{#509757} TBR=thakis@chromium.org,dpranke@chromium.org,pcc@chromium.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: 772559 Change-Id: I165c1bafe60fd3826ad8ba781e96acadbe417b15 Reviewed-on: https://chromium-review.googlesource.com/730566 Reviewed-by: Peter Collingbourne <pcc@chromium.org> Reviewed-by: Nico Weber <thakis@chromium.org> Reviewed-by: Dirk Pranke <dpranke@chromium.org> Commit-Queue: Peter Collingbourne <pcc@chromium.org> Cr-Commit-Position: refs/heads/master@{#510569} [modify] https://crrev.com/e01826fa133837ba45e7730054f50bc06a5254e5/build/config/compiler/compiler.gni
,
Nov 3 2017
The "false positives" from clusterfuzz were in fact caused by stack traces that were incorrectly symbolized because of a bug in the glog symbolizer. As a result they failed to match other stack traces that were collected either side of the revert. I have a fix for the glog bug here: https://chromium-review.googlesource.com/c/chromium/src/+/752753 Once that passes CQ I'll send it upstream. Two other problems were mentioned on issue 777980 : undefined symbol errors from the linker and stack traces weren't being symbolized. Hopefully the latter is the same as the clusterfuzz issue; I'll see if I can reproduce the former.
,
Nov 4 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/6f903f3a228bc8e10cd7bda76342954feb3000c5 commit 6f903f3a228bc8e10cd7bda76342954feb3000c5 Author: Peter Collingbourne <pcc@chromium.org> Date: Fri Nov 03 23:38:51 2017 Compute base addresses from program headers while reading /proc/self/maps. This cherry picks this glog change: https://github.com/google/glog/pull/261 with a reimplementation of the program header reading logic for the sandboxed symbolizer. This should cause unsymbolized stack traces to contain the correct addresses for binaries linked with lld. Bug: 772559 Change-Id: Ief9cbb463b3b4c32149da893c89c2eefd76b05d4 Reviewed-on: https://chromium-review.googlesource.com/752753 Commit-Queue: Peter Collingbourne <pcc@chromium.org> Reviewed-by: Mark Mentovai <mark@chromium.org> Cr-Commit-Position: refs/heads/master@{#513956} [modify] https://crrev.com/6f903f3a228bc8e10cd7bda76342954feb3000c5/base/debug/proc_maps_linux.h [modify] https://crrev.com/6f903f3a228bc8e10cd7bda76342954feb3000c5/base/debug/stack_trace_posix.cc [modify] https://crrev.com/6f903f3a228bc8e10cd7bda76342954feb3000c5/base/third_party/symbolize/README.chromium [modify] https://crrev.com/6f903f3a228bc8e10cd7bda76342954feb3000c5/base/third_party/symbolize/symbolize.cc
,
Nov 4 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/65c533858f1d0c507e91b34f64fce93f5d1cf595 commit 65c533858f1d0c507e91b34f64fce93f5d1cf595 Author: Sky Malice <skym@chromium.org> Date: Sat Nov 04 00:07:42 2017 Revert "Compute base addresses from program headers while reading /proc/self/maps." This reverts commit 6f903f3a228bc8e10cd7bda76342954feb3000c5. Reason for revert: https://build.chromium.org/p/chromium.linux/builders/Android%20Clang%20Builder%20%28dbg%29/builds/104887 Original change's description: > Compute base addresses from program headers while reading /proc/self/maps. > > This cherry picks this glog change: > https://github.com/google/glog/pull/261 > with a reimplementation of the program header reading logic for the > sandboxed symbolizer. > > This should cause unsymbolized stack traces to contain the correct > addresses for binaries linked with lld. > > Bug: 772559 > Change-Id: Ief9cbb463b3b4c32149da893c89c2eefd76b05d4 > Reviewed-on: https://chromium-review.googlesource.com/752753 > Commit-Queue: Peter Collingbourne <pcc@chromium.org> > Reviewed-by: Mark Mentovai <mark@chromium.org> > Cr-Commit-Position: refs/heads/master@{#513956} TBR=ukai@chromium.org,pcc@chromium.org,mark@chromium.org Change-Id: Ie02672bcfb2d28729530d61975083ba2187efd4e No-Presubmit: true No-Tree-Checks: true No-Try: true Bug: 772559 Reviewed-on: https://chromium-review.googlesource.com/753864 Reviewed-by: Sky Malice <skym@chromium.org> Commit-Queue: Sky Malice <skym@chromium.org> Cr-Commit-Position: refs/heads/master@{#513978} [modify] https://crrev.com/65c533858f1d0c507e91b34f64fce93f5d1cf595/base/debug/proc_maps_linux.h [modify] https://crrev.com/65c533858f1d0c507e91b34f64fce93f5d1cf595/base/debug/stack_trace_posix.cc [modify] https://crrev.com/65c533858f1d0c507e91b34f64fce93f5d1cf595/base/third_party/symbolize/README.chromium [modify] https://crrev.com/65c533858f1d0c507e91b34f64fce93f5d1cf595/base/third_party/symbolize/symbolize.cc
,
Nov 4 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/1a3ed318cbbe2b53d9d6410a376e75439b9e6007 commit 1a3ed318cbbe2b53d9d6410a376e75439b9e6007 Author: Peter Collingbourne <pcc@chromium.org> Date: Sat Nov 04 02:37:47 2017 Reland "Compute base addresses from program headers while reading /proc/self/maps." This is a reland of 6f903f3a228bc8e10cd7bda76342954feb3000c5 Fixed Android build issue. Original change's description: > Compute base addresses from program headers while reading /proc/self/maps. > > This cherry picks this glog change: > https://github.com/google/glog/pull/261 > with a reimplementation of the program header reading logic for the > sandboxed symbolizer. > > This should cause unsymbolized stack traces to contain the correct > addresses for binaries linked with lld. > > Bug: 772559 > Change-Id: Ief9cbb463b3b4c32149da893c89c2eefd76b05d4 > Reviewed-on: https://chromium-review.googlesource.com/752753 > Commit-Queue: Peter Collingbourne <pcc@chromium.org> > Reviewed-by: Mark Mentovai <mark@chromium.org> > Cr-Commit-Position: refs/heads/master@{#513956} Bug: 772559 Change-Id: Id6a8c62c0b05a7d12f817e389c7b96223fec8073 Reviewed-on: https://chromium-review.googlesource.com/754232 Reviewed-by: Mark Mentovai <mark@chromium.org> Commit-Queue: Peter Collingbourne <pcc@chromium.org> Cr-Commit-Position: refs/heads/master@{#514018} [modify] https://crrev.com/1a3ed318cbbe2b53d9d6410a376e75439b9e6007/base/android/library_loader/library_prefetcher_unittest.cc [modify] https://crrev.com/1a3ed318cbbe2b53d9d6410a376e75439b9e6007/base/debug/proc_maps_linux.h [modify] https://crrev.com/1a3ed318cbbe2b53d9d6410a376e75439b9e6007/base/debug/stack_trace_posix.cc [modify] https://crrev.com/1a3ed318cbbe2b53d9d6410a376e75439b9e6007/base/third_party/symbolize/README.chromium [modify] https://crrev.com/1a3ed318cbbe2b53d9d6410a376e75439b9e6007/base/third_party/symbolize/symbolize.cc
,
Nov 10 2017
,
Jan 29 2018
,
Jan 29 2018
Worth a try I think. There was a separate issue with msan and lld ( issue 777980 ) but I suspect that it was fixed by r319860.
,
Jan 31 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/db86e994b531cfdf9dca8295d8f528fbe8dc460c commit db86e994b531cfdf9dca8295d8f528fbe8dc460c Author: Nico Weber <thakis@chromium.org> Date: Wed Jan 31 20:38:10 2018 Try using lld in msan builds again. The issue making this fail last time has since been fixed; let's try again. Bug: 772559 Change-Id: Ic8201c00533cfb47ff8509c3a8594b7703400bca Reviewed-on: https://chromium-review.googlesource.com/895805 Reviewed-by: Peter Collingbourne <pcc@chromium.org> Commit-Queue: Nico Weber <thakis@chromium.org> Cr-Commit-Position: refs/heads/master@{#533388} [modify] https://crrev.com/db86e994b531cfdf9dca8295d8f528fbe8dc460c/build/config/compiler/compiler.gni
,
Jan 31 2018
Maybe it'll stick; the msan trybot came back green at least. |
||||||
►
Sign in to add a comment |
||||||
Comment 1 by bugdroid1@chromium.org
, Oct 6 2017