[Bug c++/30430] New: SIGABRT on run (Illumos on SPARC)

[Bug c++/30430] New: SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

            Bug ID: 30430
           Summary: SIGABRT on run (Illumos on SPARC)
           Product: gdb
           Version: 13.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
          Assignee: unassigned at sourceware dot org
          Reporter: mikeowens at gmail dot com
  Target Milestone: ---

This is specifically a SPARC issue. I am running the latest OpenIndiana on a
T4-1. I can provide SSH access to the machine to assist anyone willing to help
with this bug.

I experience the following problem with every version of gdb I can get: 10.1,
12.1-202, 13.1.

No matter how simple the program, the moment I start it to debug I get the same
result. Here is a simple case. I use the following program:

    int main() { return 0; }

Compile as follows:

    g++ -ggdb test.cpp

The gcc version is 11.3.0 (OpenIndiana 11.3.0-oi-4)

Then I run gdb as follows:

    # $ gdb a.out
    Exception caught while booting Guile.
    Error in function "load-thunk-from-memory":
    ELF file does not have native word size
    /usr/local/bin/gdb: warning: Could not complete Guile gdb module
initialization from:
    /usr/local/share/gdb/guile/gdb/boot.scm.
    Limited Guile support is available.
    Suggest passing --data-directory=/path/to/gdb/data-directory.
    GNU gdb (GDB) 13.1
    Copyright (C) 2023 Free Software Foundation, Inc.
    License GPLv3+: GNU GPL version 3 or later
<http://gnu.org/licenses/gpl.html>
    This is free software: you are free to change and redistribute it.
    There is NO WARRANTY, to the extent permitted by law.
    Type "show copying" and "show warranty" for details.
    This GDB was configured as "sparc-sun-solaris2.11".
    Type "show configuration" for configuration details.
    For bug reporting instructions, please see:
    <https://www.gnu.org/software/gdb/bugs/>.
    Find the GDB manual and other documentation resources online at:
        <http://www.gnu.org/software/gdb/documentation/>.

    For help, type "help".
    Type "apropos word" to search for commands related to "word"...
    Reading symbols from /var/data/tmp/a.out...
    (gdb) run
    Starting program: /tmp/a.out
    [Thread debugging using libthread_db enabled]


    Fatal signal: Abort
    ----- Backtrace -----
    Abort (core dumped)

If I run mbd on the core file I get no useful information:

    # mdb core
    Loading modules: [ libc.so.1 ld.so.1 ]
    > $C
    mdb: failed to get current register set: invalid thread identifier

When I run truss I see the following:

    /1:     open("/lib/64/libc.so.1", O_RDONLY)             = 18
    /1:     fcntl(18, F_GETFD, 0x00000000)                  = 0
    /1:     fcntl(18, F_SETFD, 0x00000001)                  = 0
    /1:     lseek(18, 1467912, SEEK_SET)                    = 1467912
    /1:     fstat(18, 0xFFFFFFFF7FFF9760)                   = 0
    /1:     fstat(18, 0xFFFFFFFF7FFF9580)                   = 0
    /1:     ioctl(18, TCGETA, 0xFFFFFFFF7FFF969E)           Err#25 ENOTTY
    /1:     read(18, "7F D O F02020102\b\b\0\0".., 131072)  = 131072
    /1:     open("/var/data/tmp/a.out", O_RDONLY)           = 20
    /1:     fcntl(20, F_GETFD, 0x00000000)                  = 0
    /1:     fcntl(20, F_SETFD, 0x00000001)                  = 0
    /1:     lseek(20, 4328, SEEK_SET)                       = 4328
    /1:     fstat(20, 0xFFFFFFFF7FFF5F70)                   = 0
    /1:     fstat(20, 0xFFFFFFFF7FFF5D90)                   = 0
    /1:     ioctl(20, TCGETA, 0xFFFFFFFF7FFF5EAE)           Err#25 ENOTTY
    /1:     read(20, "\0\0\0\0\0\0\001\0\0\0\0".., 10752)   = 6024
    /1:     lseek(17, 0, SEEK_SET)                          = 0
    /1:     read(17, "\t  \003\0\0\001\006\003".., 1392)    = 1392
    /1:     lseek(19, 4294967360, SEEK_SET)                 = 4294967360
    /1:     read(19, "\0\0\006\0\0\005\0\0\0\0".., 56)      = 56
    /1:     lseek(17, 0, SEEK_SET)                          = 0
    /1:     read(17, "\t  \003\0\0\001\006\003".., 1392)    = 1392
    /1:     lseek(19, 4294967416, SEEK_SET)                 = 4294967416
    /1:     read(19, "\0\0\003\0\0\004\0\0\0\0".., 56)      = 56
    /1:     lseek(17, 0, SEEK_SET)                          = 0
    /1:     read(17, "\t  \003\0\0\001\006\003".., 1392)    = 1392
    /1:     lseek(19, 4294967472, SEEK_SET)                 = 4294967472
    /1:     read(19, "\0\0\001\0\0\005\0\0\0\0".., 56)      = 56
    /1:     lseek(17, 0, SEEK_SET)                          = 0

    ...

    /1:     read(20, "\0\0\0\0\0\0\001\0\0\0\0".., 10752)   = 6024
    /1:     lseek(17, 0, SEEK_SET)                          = 0
    /1:     read(17, "\t  \003\0\0\001\006\003".., 1392)    = 1392
    /1:     lseek(19, 4296020768, SEEK_SET)                 = 4296020768
    /1:     read(19, "FFFFFFFF `84  \0", 8)                 = 8
    /1:     sigaction(SIGABRT, 0x00000000, 0xFFFFFFFF7FFF5820) = 0
    /1:     lwp_kill(1, SIGABRT)                            = 0
    /1:         Received signal #6, SIGABRT [caught]
    /1:           siginfo: SIGABRT pid=20069 uid=100 code=-1
    /1:     lwp_sigmask(SIG_SETMASK, 0x00000000, 0x00000000, 0x00000000,
0x00000000) = 0xFFBFFEFF [0xFFFFFFFF]

I get this exact same error on every version of gdb I test with.

I understand that SPARC is a dead platform, but thankfully it seems that gdb
still has theoretical support for it. The code built cleanly from source. You
can still get really good deals on SPARC hardware and I would love to be able
to
use this system for a long time to come. But without a working debugger, this
system is utterly useless.

Like I said, I can provide direct access to the machine for anyone willing to
work on the issue. It's a great machine with 250GB RAM and 64 cores. It's a lot
of fun to work on it. And I have pkgsrc set up on it so I can build any
additional tools that might be helpful.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug c++/30430] SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

--- Comment #1 from Mike Owens <mikeowens at gmail dot com> ---
If I run this as root with truss, I get a slightly different (shorter) result.
In both cases it seems the signal is being issued on read():


    /1:     open("/lib/sparcv9/libc.so.1-gdb.gdb", O_RDONLY|O_CLOEXEC) Err#2
ENOENT
    /1:     open("/opt/local/lib/debug/lib/sparcv9/libc.so.1-gdb.gdb",
O_RDONLY|O_CLOEXEC) Err#2 ENOENT
    /1:    
open("/opt/local/share/gdb/auto-load/lib/sparcv9/libc.so.1-gdb.gdb",
O_RDONLY|O_CLOEXEC) Err#2 ENOENT
    /1:     resolvepath("/lib/sparcv9/libc.so.1", "/lib/sparcv9/libc.so.1",
1024) = 22
    /1:     open("/lib/sparcv9/libc.so.1-gdb.py", O_RDONLY|O_CLOEXEC) Err#2
ENOENT
    /1:     open("/opt/local/lib/debug/lib/sparcv9/libc.so.1-gdb.py",
O_RDONLY|O_CLOEXEC) Err#2 ENOENT
    /1:     open("/opt/local/share/gdb/auto-load/lib/sparcv9/libc.so.1-gdb.py",
O_RDONLY|O_CLOEXEC) Err#2 ENOENT
    /1:     lseek(14, 0xFFFFFFFFFFFE0228, SEEK_CUR)         = 306960
    /1:     close(14)                                       = 0
    /1:     open("/lib/sparcv9/libc.so.1", O_RDONLY)        = 14
    /1:     fcntl(14, F_GETFD, 0x00000000)                  = 0
    /1:     fcntl(14, F_SETFD, 0x00000001)                  = 0
    /1:     lseek(14, 1467912, SEEK_SET)                    = 1467912
    /1:     fstat(14, 0xFFFFFFFF7FFEFFC0)                   = 0
    /1:     fstat(14, 0xFFFFFFFF7FFEFDE0)                   = 0
    /1:     ioctl(14, TCGETA, 0xFFFFFFFF7FFEFEFE)           Err#25 ENOTTY
    /1:     read(14, "7F D O F02020102\b\b\0\0".., 131072)  = 131072
    /1:     sigaction(SIGABRT, 0x00000000, 0xFFFFFFFF7FFEC290) = 0
    /1:     lseek(14, 0xFFFFFFFFFFFE1DC0, SEEK_CUR)         = 1475528
    /1:     sigaction(SIGABRT, 0xFFFFFFFF7FFEC0B0, 0xFFFFFFFF7FFEC1A0) = 0
    /1:     lwp_sigmask(SIG_SETMASK, 0x00000000, 0x00000000, 0x00000000,
0x00000000) = 0xFFBFFEFF [0xFFFFFFFF]
    /1:     lwp_kill(1, SIGABRT)                            = 0
    /1:         Received signal #6, SIGABRT [default]
    /1:           siginfo: SIGABRT pid=20523 uid=0 code=-1

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug c++/30430] SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

--- Comment #2 from Mike Owens <mikeowens at gmail dot com> ---
I found that the ABRT is triggered in procfs.c within procfs_xfer_memory(): the
lseek() followed by read():

  if (lseek (pi->as_fd, (off_t) memaddr, SEEK_SET) != (off_t) memaddr)
    return TARGET_XFER_E_IO;

  if (writebuf != NULL)
    {
      PROCFS_NOTE ("write memory:\n");
      nbytes = write (pi->as_fd, writebuf, len);
    }
  else
    {
      PROCFS_NOTE ("read  memory:\n");
      nbytes = read (pi->as_fd, readbuf, len);
    }

I noticed there was a procfs_trace variable so I enabled it and ran again. This
is the latter part of the debug output sent to "procfs_trace" file before the
ABRT:

    Thread 1: PR_STOPPED PR_ISTOP PR_RLC PR_MSACCT PR_PCOMPAT 
    PR_FAULTED FLTBPT 
    read  memory:

This is pretty much repeated through the file.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug c++/30430] SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

--- Comment #3 from Mike Owens <mikeowens at gmail dot com> ---
The file that the read() is generating the SIGABRT on is the proc address space
file (e.g. open("/proc/2740/as", O_RDWR))

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug c++/30430] SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

Mike Owens <mikeowens at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikeowens at gmail dot com
               Host|                            |SunOS t4-1 5.11
                   |                            |illumos-70143b9fb6 sun4v
                   |                            |sparc sun4v illumos

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug tdep/30430] SIGABRT on run (Illumos on SPARC)

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

Tom Tromey <tromey at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c++                         |tdep
                 CC|                            |tromey at sourceware dot org

--- Comment #4 from Tom Tromey <tromey at sourceware dot org> ---
(In reply to Mike Owens from comment #0)
> This is specifically a SPARC issue. I am running the latest OpenIndiana on a
> T4-1.

I suspect it's more about the OS than the architecture.
I know other sparc ports work ok.

Unfortunately I don't think anybody is actively maintaining this port.
So, it's possible that either it needs some changes for more recent versions
of the OS, or that the code in gdb bit-rotted a bit as other gdb changes
were made.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug tdep/30430] SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

--- Comment #5 from Mike Owens <mikeowens at gmail dot com> ---
Yeah I was afraid of that. I tried getting help on IRC and it's dead there so I
really don't expect this to be a priority to anyone. I will dig around when I
can to see if I can figure out what's going on but I knew when filing this PR
that it was a shot in the dark at best. Thanks for your response.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug tdep/30430] SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

Mike Owens <mikeowens at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED

--- Comment #6 from Mike Owens <mikeowens at gmail dot com> ---
This issue is no longer relevant.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug tdep/30430] SIGABRT on run (Illumos on SPARC)

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

--- Comment #7 from Tom Tromey <tromey at sourceware dot org> ---
(In reply to Mike Owens from comment #6)
> This issue is no longer relevant.

What changed?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug tdep/30430] SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

--- Comment #8 from Mike Owens <mikeowens at gmail dot com> ---
There's no point in trying to fix this. From what I can tell, the problem is
deep within the bowels of procfs. Furthermore mdb cannot print an actual stack
trace of the core dump, but rather prints some sort of cryptic message. So
there is no other debugger with which to debug this.

And, practically speaking, fixing this is not going to benefit anyone except
maybe me. I am just using it because I like SPARC and Solaris history -- both
are so beautifully engineered. But, honestly, nobody is using this hardware or
OS to do any kind of development. 

So I am just am facing the facts: this is a waste of time and it's not fair for
me to file this is a bug for other people to spend their time fixing.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug tdep/30430] SIGABRT on run (Illumos on SPARC)

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

--- Comment #9 from Tom Tromey <tromey at sourceware dot org> ---
Ok. Thanks for your response.
It's fine if you want to leave it open.
Maybe gdb is doing something wrong here (I don't know),
or maybe it's an OpenIndiana bug.
Anyway it's relatively normal for less-maintained targets to
have bugs open in gdb.  It doesn't really take up anybody's time
and sometimes (rarely I admit) these kinds of bugs do get
taken up by someone.
It's also fine if you'd rather just leave it closed as well.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug tdep/30430] SIGABRT on run (Illumos on SPARC)

[mikeowens at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30430

Mike Owens <mikeowens at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|FIXED                       |---
             Status|RESOLVED                    |UNCONFIRMED

--- Comment #10 from Mike Owens <mikeowens at gmail dot com> ---
Ok then, I will leave it open for anyone looking for punishment.

-- 
You are receiving this mail because:
You are on the CC list for the bug.