Bug 1807872

Summary:	capnproto test suite segfaults on ARM when compiled with default compiler flags with gcc10
Product:	[Fedora] Fedora	Reporter:	Neal Gompa <ngompa13>
Component:	capnproto	Assignee:	Neal Gompa <ngompa13>
Status:	CLOSED ERRATA	QA Contact:	Fedora Extras Quality Assurance <extras-qa>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	rawhide	CC:	aoliva, avi.kivity, dmalcolm, fweimer, jakub, jwakely, law, mpolacek, msebor, ngompa13, nickc
Target Milestone:	---
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:	capnproto-0.7.0-6.fc32	Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2020-03-16 20:38:54 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:
Bug Blocks:	245418, 1788297

Description Neal Gompa 2020-02-27 12:11:34 UTC

Description of problem:
During the course of working on bug 1788297 and the affiliated upstream bug[1],
I discovered that capnproto's test suite only segfaults on ARM when I use the Fedora default compiler flags. If I compile with just "-g", everything passes.

Upstream has suggested this is a GCC bug, and I'm inclined to agree.

[1]: https://github.com/capnproto/capnproto/issues/937

Version-Release number of selected component (if applicable):
10.0.1-0.8.fc33

How reproducible:
Always

Steps to Reproduce:
1. fedpkg clone -a capnproto && cd capnproto
2. fedpkg scratch-build --arches armv7hl

Actual results:
Segfault while trying to run "table-test.c++:786: large tree table" test

Expected results:
All tests pass.

Additional info:
This is blocking resolving a FTBFS issue with capnproto.

Comment 1 Jakub Jelinek 2020-02-29 13:23:57 UTC

Tried building it on x86_64 with -O0 -g -fsanitize=undefined,address and it FAILs there too, e.g.
src/capnp/compiler/module-loader.c++:118:65: runtime error: signed integer overflow: 1535247712000000000 * 73 cannot be represented in type 'long int'
src/kj/array.h:466:14: runtime error: reference binding to misaligned address 0x7f9ca0e74dae for type 'const struct ArrayDisposer', which requires 8 byte alignment
src/kj/string.c++:190:25: runtime error: negation of -9223372036854775808 cannot be represented in type 'long int'; cast to an unsigned type to negate this value to itself
src/kj/array.h:466:14: runtime error: reference binding to misaligned address 0x000000000002 for type 'const struct ArrayDisposer', which requires 8 byte alignment
0x000000000002: note: pointer points here
<memory cannot be printed>
(just a couple of different messages from src/capnp/compiler/capnp-test.sh.log).
Furthermore, lots of warnings like:
./src/kj/common.h:1267:23: warning: initializing ‘kj::ArrayPtr<const kj::ArrayPtr<const unsigned char> >::ptr’ from ‘std::initializer_list<kj::ArrayPtr<const unsigned char> >::begin’ does not extend the lifetime of the underlying array [-Winit-list-lifetime]
So not really sure I want to spend too much time looking into it, because it is very likely just buggy package.

Comment 2 Jakub Jelinek 2020-03-02 16:05:57 UTC

As mentioned in https://github.com/capnproto/capnproto/issues/937 , this is a capnproto aliasing violation.

Comment 3 Fedora Update System 2020-03-12 13:23:40 UTC

FEDORA-2020-3e12574632 has been submitted as an update to Fedora 32. https://bodhi.fedoraproject.org/updates/FEDORA-2020-3e12574632

Comment 4 Fedora Update System 2020-03-12 18:50:14 UTC

capnproto-0.7.0-6.fc32 has been pushed to the Fedora 32 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2020-3e12574632

Comment 5 Fedora Update System 2020-03-16 20:38:54 UTC

capnproto-0.7.0-6.fc32 has been pushed to the Fedora 32 stable repository. If problems still persist, please make note of it in this bug report.