[Bug compile/29541] New: gdb.compile/compile-cplus-anonymous.exp shows failures with glibc debuginfo installed

[Bug compile/29541] New: gdb.compile/compile-cplus-anonymous.exp shows failures with glibc debuginfo installed

[kevinb at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29541

            Bug ID: 29541
           Summary: gdb.compile/compile-cplus-anonymous.exp shows failures
                    with glibc debuginfo installed
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: normal
          Priority: P2
         Component: compile
          Assignee: unassigned at sourceware dot org
          Reporter: kevinb at redhat dot com
  Target Milestone: ---

I'm seeing failures in the gdb test gdb.compile/compile-cplus-anonymous.exp
when run, with glibc debuginfo installed, on Fedora releases ranging from
Fedora 29 thru 37 (and rawhide, currently numbered 38 too).  I also see these
failures on Ubuntu 22.04 LTS once a missing symlink is installed for libcc1.so.

To reproduce this problem on Fedora, do (as root or using sudo):

[root@f36-1 gdb]# debuginfo-install -y glibc

Then, using a recent (or even old) build of upstream gdb, while in the gdb
directory (where the 'gdb' executable is found), do:

make check TESTS=gdb.compile/compile-cplus-anonymous.exp

I see the following failures / summary:

FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code anon_e (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code anon_e
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print anon_e (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code anon_u.aa (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code anon_u.aa
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print anon_u.aa
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code a.u.b (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code a.u.b
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print a.u.b (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code a.s.len (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code a.s.len
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print a.s.len (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code a.e (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code a.e
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print a.e (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code (*a.s.ptr != 'h')
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code (*a.s.ptr
!= 'h') (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print (*a.s.ptr != 'h')
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code A::BB (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code A::BB
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print A::BB (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code ABC (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code ABC
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print ABC (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code DEF (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code DEF
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print DEF (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code GHI (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code GHI
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print GHI (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code JKL (timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code JKL
(timeout)
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile print JKL (timeout)

                === gdb Summary ===

# of expected passes            1
# of unexpected failures        33
# of known failures             12

On Fedora, glibc debuginfo can be removed as follows:

[root@f36-1 gdb]# dnf remove -y glibc-debuginfo

Running gdb.compile/compile-cplus-anonymous.exp with glibc debuginfo NOT
installed shows this summary:

                === gdb Summary ===

# of expected passes            34
# of known failures             12

Testing an upstream build of GDB on Ubuntu 22.04 LTS shows:

# of untested testcases         1

I found that libcc1.so was missing.  So, I did:

root@u2204-1:~# cd /usr/lib/x86_64-linux-gnu/
root@u2204-1:/usr/lib/x86_64-linux-gnu# ln -sn libcc1.so.0 libcc1.so

With that symlink in place, I see the following failures / summary on Ubuntu
22.04 LTS:

FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code anon_e
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code anon_e
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code anon_u.aa
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code anon_u.aa
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code a.u.b
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code a.u.b
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code a.s.len
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code a.s.len
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code a.e
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code a.e
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code (*a.s.ptr != 'h')
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code (*a.s.ptr
!= 'h')
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code A::BB
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code A::BB
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code ABC
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code ABC
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code DEF
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code DEF
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code GHI
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code GHI
FAIL: gdb.compile/compile-cplus-anonymous.exp: compile code JKL
FAIL: gdb.compile/compile-cplus-anonymous.exp: result of compile code JKL

                === gdb Summary ===

# of expected passes            12
# of unexpected failures        22
# of known failures             12

Back on Fedora, if I run the initial part of the test by hand, without glibc
debuginfo installed AND without debuginfod, I see:

[kev@f36-1 gdb]$ ./gdb -q
testsuite/outputs/gdb.compile/compile-cplus-anonymous/compile-cplus-anonymous
Reading symbols from
testsuite/outputs/gdb.compile/compile-cplus-anonymous/compile-cplus-anonymous...
(gdb) set debuginfod enabled off
(gdb) set height 0
(gdb) set width 0
(gdb) break -qualified main
Breakpoint 1 at 0x401145: file
/ironwood1/sourceware-git/f36-master/bld/../../worktree-master/gdb/testsuite/gdb.compile/compile-cplus-anonymous.cc,
line 71.
(gdb) run
Starting program:
/mesquite2/sourceware-git/f36-master/bld/gdb/testsuite/outputs/gdb.compile/compile-cplus-anonymous/compile-cplus-anonymous 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".

Breakpoint 1, main () at
/ironwood1/sourceware-git/f36-master/bld/../../worktree-master/gdb/testsuite/gdb.compile/compile-cplus-anonymous.cc:71
71        A a;
(gdb) compile code -- ;
(gdb) break 75
Breakpoint 2 at 0x401165: file
/ironwood1/sourceware-git/f36-master/bld/../../worktree-master/gdb/testsuite/gdb.compile/compile-cplus-anonymous.cc,
line 75.
(gdb) continue
Continuing.

Breakpoint 2, main () at
/ironwood1/sourceware-git/f36-master/bld/../../worktree-master/gdb/testsuite/gdb.compile/compile-cplus-anonymous.cc:75
75          + static_cast<int> (anon_e) + anon_u.bb + anon_s.len; // break here
(gdb) compile code var = anon_e
(gdb)

I.e. 'compile code var = anon_e' works as expected.

Enabling debuginfod (but still without glibc debuginfo installed system-wide)
shows the following behavior:

(gdb) set debuginfod enabled on
... [same commands and similar output shown above]
(gdb) compile code var = anon_e
*** WARNING *** there are active plugins, do not report this as a bug unless
you can reproduce it without enabling any plugins.
Event                            | Plugins
PLUGIN_PRE_GENERICIZE            | libcp1plugin
PLUGIN_GGC_MARKING               | libcp1plugin
PLUGIN_PRAGMAS                   | libcp1plugin
gdb command line:1:1: internal compiler error: in plugin_start_enum_type, at
libcc1/libcp1plugin.cc:1782
Please submit a full bug report, with preprocessed source.
See <http://bugzilla.redhat.com/bugzilla> for instructions.

GDB is hung at this point, but Ctrl-C returns the user to the GDB prompt. This
explains the (many) timeouts that I see on Fedora.

I see this same behavior when I disable debuginfod, but have glibc debuginfo
installed via the 'debuginfo-install glibc' command shown earlier. In the
interest of brevity (!), I won't paste that nearly identical output.  (The part
that's different is the use of 'set debuginfod enabled off')

On Ubuntu 22.04 LTS, when running the test by hand, without debuginfod, but
with the libcc1.so symlink fixed as mentioned earlier, I see:

(gdb) set debuginfod enabled off
... [snipped session which is largely the same as that shown for Fedora]
(gdb) compile code var = anon_e
*** WARNING *** there are active plugins, do not report this as a bug unless
you can reproduce it without enabling any plugins.
Event                            | Plugins
PLUGIN_PRE_GENERICIZE            | libcp1plugin
PLUGIN_GGC_MARKING               | libcp1plugin
PLUGIN_PRAGMAS                   | libcp1plugin
gdb command line:1:1: internal compiler error: in plugin_start_enum_type, at
libcc1/libcp1plugin.cc:1937
0x7fd261edbd8f __libc_start_call_main
        ../sysdeps/nptl/libc_start_call_main.h:58
0x7fd261edbe3f __libc_start_main_impl
        ../csu/libc-start.c:392
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <file:///usr/share/doc/gcc-11/README.Bugs> for instructions.
Compilation failed.
(gdb) 

The main difference when running this on Ubuntu is that, after printing out the
messages related to the internal compiler error, GDB does not hang, but instead
prints "Compilation failed.", returning to the GDB prompt.  This explains the
difference between the number of failures shown for Ubuntu versus Fedora.  On
Fedora, the remaining tests failed due to the timeout. On Ubuntu, some gdb
tests succeeded; those which attempt to 'compile' something else fail.

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

[Bug compile/29541] gdb.compile/compile-cplus-anonymous.exp shows failures with glibc debuginfo installed

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29541

Tom Tromey <tromey at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tromey at sourceware dot org

--- Comment #1 from Tom Tromey <tromey at sourceware dot org> ---
I've run into this recently.

Looking into it, I see that the gdb plugin can hang while
waiting for results from gcc.  This happens due in a select()
call... this is unfortunate, to say the least.  I'm also
disappointed in my past self because there doesn't seem to
be any sort of sensible error-reporting system in the gdb
plugin.  That is, if a call fails, gdb will never really know
about it, but instead will happily continue trying to exercise
the API using nonsense :(

Meanwhile I do at least have a workaround for the test suite
timeouts.  These have been a real pain.

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

[Bug compile/29541] gdb.compile/compile-cplus-anonymous.exp shows failures with glibc debuginfo installed

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29541

--- Comment #2 from Tom Tromey <tromey at sourceware dot org> ---
https://sourceware.org/pipermail/gdb-patches/2022-November/194367.html

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

[Bug compile/29541] gdb.compile/compile-cplus-anonymous.exp shows failures with glibc debuginfo installed

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29541

--- Comment #3 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Tom Tromey <tromey@sourceware.org>:

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=9a7fde04ca4232d4b2efed61486fff65e83ab3a2

commit 9a7fde04ca4232d4b2efed61486fff65e83ab3a2
Author: Tom Tromey <tromey@adacore.com>
Date:   Wed Nov 30 11:27:10 2022 -0700

    Avoid timeouts in gdb.compile

    PR compile/29541 points out that some of the C++ tests in gdb.compile
    will time out when the glibc debuginfo is installed.  This was
    interfering with my hacking on gdb by making test runs extremely long,
    so I looked into it.

    Internally the bug seems to be that gdb tries to convert multiple
    symbols named "var" via the compiler interface; one such symbol (I
    didn't track it down too far) causes the C++ compiler plugin to crash.

    Unfortunately, the crash is reported as a timeout, as the gdb side of
    the plugin simply hangs.  This seems like a bug in the plugin RPC
    mechanism and, worse, apparently when I wrote this stuff I didn't
    really consider error reporting very much at all, so gdb can't really
    detect failures in the first place.

    Anyway... this patch works around the timeout by compiling a simple
    test that should provoke this bug, and then using "untested" if it
    notices a GCC crash.

    Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29541

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