[Bug 1091538] New: golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

Bug ID: 1091538
           Summary: golang crashing on ARMv7
           Product: Fedora
           Version: rawhide
         Component: golang
          Assignee: [e-mail]
          Reporter: [e-mail]
        QA Contact: [e-mail]
                CC: [e-mail], [e-mail],
                    [e-mail], [e-mail],
                    [e-mail], [e-mail], [e-mail],
                    [e-mail]
            Blocks: 245418 (ARMTracker)

We're getting an crash with go when I try and enable go support for ARM in the
swig package.

http://koji.fedoraproject.org/koji/taskinfo?taskID=6780663

It's currently just a scratch build as I obviously can't push it if it fails.

checking Examples/go/callback
checking Examples/go/class
unexpected fault address 0x40440000
fatal error: fault
[signal 0xb code=0x1 addr=0x40440000 pc=0x40440000]
goroutine 1 [running]:
runtime.throw(0x15691f)
    /usr/lib/golang/src/pkg/runtime/panic.c:464 +0x5c fp=0xb6ba5f7c
runtime: unexpected return pc for runtime.sigpanic called from 0x40440000
runtime.sigpanic()
    /usr/lib/golang/src/pkg/runtime/os_linux.c:237 +0xa8 fp=0xb6ba5f88
goroutine 3 [syscall]:
runtime.goexit()
    /usr/lib/golang/src/pkg/runtime/proc.c:1394
checking Examples/go/constants
make[3]: *** [go_run] Error 2
make[2]: *** [check] Error 2
make[1]: *** [class.actionexample] Error 2
checking Examples/go/enum
checking Examples/go/extend
checking Examples/go/funcptr
checking Examples/go/multimap
checking Examples/go/pointer
checking Examples/go/reference
checking Examples/go/simple
checking Examples/go/template
checking Examples/go/variables

Referenced Bugs:

https://bugzilla.redhat.com/show_bug.cgi?id=245418
[Bug 245418] Tracker for ARM support

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

--- Comment #1 from Vincent Batts  ---
Wacky. I wonder what code it is compiling/running to get a panic on a syscall
...

Peter, code you paste the code from the swig class.actionexample ?

(In reply to Peter Robinson from comment #0)
> We're getting an crash with go when I try and enable go support for ARM in
> the swig package.
>
> http://koji.fedoraproject.org/koji/taskinfo?taskID=6780663
>
> It's currently just a scratch build as I obviously can't push it if it fails.
>
> checking Examples/go/callback
> checking Examples/go/class
> unexpected fault address 0x40440000
> fatal error: fault
> [signal 0xb code=0x1 addr=0x40440000 pc=0x40440000]
> goroutine 1 [running]:
> runtime.throw(0x15691f)
>    /usr/lib/golang/src/pkg/runtime/panic.c:464 +0x5c fp=0xb6ba5f7c
> runtime: unexpected return pc for runtime.sigpanic called from 0x40440000
> runtime.sigpanic()
>    /usr/lib/golang/src/pkg/runtime/os_linux.c:237 +0xa8 fp=0xb6ba5f88
> goroutine 3 [syscall]:
> runtime.goexit()
>    /usr/lib/golang/src/pkg/runtime/proc.c:1394
> checking Examples/go/constants
> make[3]: *** [go_run] Error 2
> make[2]: *** [check] Error 2
> make[1]: *** [class.actionexample] Error 2
> checking Examples/go/enum
> checking Examples/go/extend
> checking Examples/go/funcptr
> checking Examples/go/multimap
> checking Examples/go/pointer
> checking Examples/go/reference
> checking Examples/go/simple
> checking Examples/go/template
> checking Examples/go/variables

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

--- Comment #2 from Peter Robinson  ---
(In reply to Vincent Batts from comment #1)
> Wacky. I wonder what code it is compiling/running to get a panic on a
> syscall ...
>
> Peter, code you paste the code from the swig class.actionexample ?

Can't you retrieve it from the .src.rpm?

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

--- Comment #3 from Vincent Batts  ---
Peter Robinson, Is this still an issue on swig-3.0.2-1? looks like Swig is
building clean lately
http://koji.fedoraproject.org/koji/packageinfo?packageID=400

(In reply to Peter Robinson from comment #0)
> We're getting an crash with go when I try and enable go support for ARM in
> the swig package.
>
> http://koji.fedoraproject.org/koji/taskinfo?taskID=6780663
>
> It's currently just a scratch build as I obviously can't push it if it fails.
>
> checking Examples/go/callback
> checking Examples/go/class
> unexpected fault address 0x40440000
> fatal error: fault
> [signal 0xb code=0x1 addr=0x40440000 pc=0x40440000]
> goroutine 1 [running]:
> runtime.throw(0x15691f)
>    /usr/lib/golang/src/pkg/runtime/panic.c:464 +0x5c fp=0xb6ba5f7c
> runtime: unexpected return pc for runtime.sigpanic called from 0x40440000
> runtime.sigpanic()
>    /usr/lib/golang/src/pkg/runtime/os_linux.c:237 +0xa8 fp=0xb6ba5f88
> goroutine 3 [syscall]:
> runtime.goexit()
>    /usr/lib/golang/src/pkg/runtime/proc.c:1394
> checking Examples/go/constants
> make[3]: *** [go_run] Error 2
> make[2]: *** [check] Error 2
> make[1]: *** [class.actionexample] Error 2
> checking Examples/go/enum
> checking Examples/go/extend
> checking Examples/go/funcptr
> checking Examples/go/multimap
> checking Examples/go/pointer
> checking Examples/go/reference
> checking Examples/go/simple
> checking Examples/go/template
> checking Examples/go/variables

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

--- Comment #4 from Peter Robinson  ---
(In reply to Vincent Batts from comment #3)
> Peter Robinson, Is this still an issue on swig-3.0.2-1? looks like Swig is
> building clean lately
> http://koji.fedoraproject.org/koji/packageinfo?packageID=400

Only because go was disabled. I'll retest and update

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

--- Comment #5 from Peter Robinson  ---
It still fails.

http://koji.fedoraproject.org/koji/taskinfo?taskID=7058479

You can test yourself with the following change to the swig spec file:

diff --git a/swig.spec b/swig.spec
index 7c1b180..5bc6105 100644
--- a/swig.spec
+++ b/swig.spec
@@ -6,7 +6,7 @@
 %{!?lualang:%global lualang 1}
 %{!?rubylang:%global rubylang 1}

-%ifarch aarch64 %{arm} ppc64le ppc %{power64} s390 s390x
+%ifarch aarch64 ppc %{power64} s390 s390x
 %{!?golang:%global golang 0}
 %{!?Rlang:%global Rlang 0}
 %{!?javalang:%global javalang 0}

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

Adam Goode  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|[e-mail]          |[e-mail]

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

--- Comment #7 from Vincent Batts  ---
(In reply to Peter Robinson from comment #5)
> It still fails.
>
> http://koji.fedoraproject.org/koji/taskinfo?taskID=7058479
>
> You can test yourself with the following change to the swig spec file:
>
> diff --git a/swig.spec b/swig.spec
> index 7c1b180..5bc6105 100644
> --- a/swig.spec
> +++ b/swig.spec
> @@ -6,7 +6,7 @@
>  %{!?lualang:%global lualang 1}
>  %{!?rubylang:%global rubylang 1}
>
> -%ifarch aarch64 %{arm} ppc64le ppc %{power64} s390 s390x
> +%ifarch aarch64 ppc %{power64} s390 s390x
>  %{!?golang:%global golang 0}
>  %{!?Rlang:%global Rlang 0}
>  %{!?javalang:%global javalang 0}

Peter, there has been a lot of upstream improvements on arm. Can you retest?

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

Peter Robinson  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Blocks|                            |922257
                   |                            |(ARM64,F-ExcludeArch-aarch6
                   |                            |4), 1071880 (PPCTracker),
                   |                            |467765 (ZedoraTracker)

--- Comment #8 from Peter Robinson  ---
> Peter, there has been a lot of upstream improvements on arm. Can you retest?

It's going to take me a while to get to this, you could just drop the %{arm}
conditional and do a scratch build to see if it works.

In actual fact looking at the conditional it could likely be dropped entirely
as we have golang on aarch64/Power64 since then and gcc-go on s390(x) too so in
theory they should be supportable on all of the arches now.

Adding the other arches as someone might have some cycles to review them.

Referenced Bugs:

https://bugzilla.redhat.com/show_bug.cgi?id=467765
[Bug 467765] Fedora for System z (s390): Bug Tracker
https://bugzilla.redhat.com/show_bug.cgi?id=922257
[Bug 922257] Tracker for ARM64 support
https://bugzilla.redhat.com/show_bug.cgi?id=1071880
[Bug 1071880] (PPCTracker) Fedora for PowerPC architectures
(ppc64,ppc64le): Bug Tracker

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

--- Comment #9 from Peter Robinson  ---
> Peter, there has been a lot of upstream improvements on arm. Can you retest?

So kicked off some scratch builds with this diff enabling it on
ppc64/ppc64le/s390x/aarch64/armv7.

-%ifarch aarch64 %{arm} %{mips} ppc64le ppc %{power64} s390 s390x
+%ifarch %{mips} ppc s390
 %{!?golang:%global golang 0}

aarch64: looks to be some java issues
http://arm.koji.fedoraproject.org/koji/taskinfo?taskID=3537366

power64: golang issues
http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=3366693

s390x: looks to needs some conditionals to use gcc-go
http://s390.koji.fedoraproject.org/koji/taskinfo?taskID=2221906

ARMv7: still building
http://koji.fedoraproject.org/koji/taskinfo?taskID=13973607

I suspect it's probably worthwhile to untangle the golang/java/R enable
options, I've added it to my list to investigate but I don't have much time and
it's really something the maintainer should be doing.

[Bug 1091538] golang crashing on ARMv7

[Red Hat Bugzilla]

https://bugzilla.redhat.com/show_bug.cgi?id=1091538

--- Comment #10 from Peter Robinson  ---
ARMv7 appears to have completed:

http://koji.fedoraproject.org/koji/taskinfo?taskID=13973607