Comment # 1 on bug 1192982 from Uli Laube

I ran into this error message a couple of times now and did some digging. Here
is what I found so far.

I am on openSUSE Tumbleweed 20241129 with ltrace-0.7.91+git20230705.8eabf68
installed. I am dabbling in some programming tasks of my own and tried to used
ltrace to find a bug in my own software. While doing so I got to see

malloc(ltrace: type.c:100: type_struct_add: Assertion `field_info->type !=
ARGTYPE_VOID' failed.
[1]    17111 IOT instruction (core dumped)  ltrace ./test

more than I liked. I started to experiment a little, first with C:

#include <stdlib.h>
#include <stdio.h>

int main() {
  void* ptr = malloc(100);
  printf("Hello World!\n");
  free(ptr);
  return 0;
}

$ gcc -o test -O0 test.c
$ ltrace ./test

I get the expected output:

malloc(100)            = 0x1c83e2a0
puts("Hello World!")   = 13
free(0x1c83e2a0)       = <void>
+++ exited (status 0) +++

Doing the same with debug information turned on

$ gcc -o test -O0 -g test.c
$ ltrace ./test

Same result, it works as expected. Switching to C++:

#include <cstdlib>
#include <print>

int main() {
  void* ptr = malloc(100);
  std::println("Hello World!");
  free(ptr);
  return 0;
}

$ g++ -o test -std=c++23 -O0 test.cpp
$ ltrace ./test

produces the expected result:

malloc(100)                                     = 0x2bc872b0
memchr("Hello World!", '{',12)                  = nil
memchr("Hello World!", '}', 12)                 = nil
memcpy(0x7ffcbea03640, "Hello World!", 12)      = 0x7ffcbea03640
memcpy(0x7ffcbea03750, "Hello World!", 12)      = 0x7ffcbea03750
memcpy(0x7ffcbea03840, "Hello World!\0", 13)    = 0x7ffcbea03840
memchr("{}\n", '{', 3)                          = 0x41a15a
memchr("{}\n", '}', 3)                          = 0x41a15b
memcpy(0x7ffcbea035c0, "Hello World!", 12)      = 0x7ffcbea035c0
memchr("\n", '{', 1)                            = nil
memchr("\n", '}', 1)                            = nil
memcpy(0x7ffcbea035cc, "\n", 1)                 = 0x7ffcbea035cc
fwrite("Hello World!\n", 1, 13, 0x7fc1a61fd5c0) = 13
free(0x2bc872b0)                                = <void>
+++ exited (status 0) +++

But turning on debug information:

$ g++ -o test -std=c++23 -O0 -g test.cpp
$ ltrace ./test

reproduces the error:

malloc(ltrace: type.c:100: type_struct_add: Assertion `field_info->type !=
ARGTYPE_VOID' failed.
[1]    8307 IOT instruction (core dumped)  ltrace ./test

That is what I was experiencing while searing for bugs in my own programs. They
are written in C++ and I was on a build with debug settings. Then ltrace fails
in the same way. Release builds of my programs where fine as show above with
the test programs.

I switched compilers from g++ 14.2.1 to clang++ 19.1.4 but the problem remains.

However, grabbing the upstream sources (https://gitlab.com/cespedes/ltrace.git)
and compiling ltrace 0.7.91 on my machine does NOT reproduce the problem. I get
the expected output:

malloc(100, 0x7ffd63d18868, 0x7ffd63d18878, 0x563399116bd8)       =
0x5633d167e2b0
memchr(0x56339910d4a8, 123, 12, 12)                               = 0
memchr(0x56339910d4a8, 125, 12, 12)                               = 0
memcpy(0x7ffd63d183f0, 0x56339910d4a8, 12, 0x7ffd63d17d78)        =
0x7ffd63d183f0
memcpy(0x7ffd63d18500, 0x7ffd63d183f0, 12, 15)                    =
0x7ffd63d18500
memcpy(0x7ffd63d186b8, 0x7ffd63d18500, 13, 0x7ffd63d18500)        =
0x7ffd63d186b8
memchr(0x56339910d4b5, 123, 3, 3)                                 =
0x56339910d4b5
memchr(0x56339910d4b5, 125, 3, 3)                                 =
0x56339910d4b6
memcpy(0x7ffd63d18340, 0x7ffd63d186b8, 12, 0x7ffd63d17a68)        =
0x7ffd63d18340
memchr(0x56339910d4b7, 123, 1, 1)                                 = 0
memchr(0x56339910d4b7, 125, 1, 1)                                 = 0
memcpy(0x7ffd63d1834c, 0x56339910d4b7, 1, 0x7ffd63d17c98)         =
0x7ffd63d1834c
fwrite(0x7ffd63d18340, 1, 13, 0x7f27317fd5c0)                     = 13
free(0x5633d167e2b0, 0x5633d167e320, 0, 0x7ffd63d186b8)           = 5
+++ exited (status 0) +++

I then grabbed the source rpm ltrace-0.7.91+git20230705.8eabf68-2.2.src.rpm
from download.opensuse.org-oss, rebuild it locally, installed it and this too
fails as before. Interestingly the testsuite fails in many places:

Running target unix
Using /usr/share/dejagnu/baseboards/unix.exp as board description file for
target.
Using /usr/share/dejagnu/config/unix.exp as generic interface file for target.
Using
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/config/unix.exp
as tool-and-target-specific interface file.
Running
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/XDG_CONFIG_DIRS.exp
...
Running
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/branch_func.exp
...
Running
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.exp
...
FAIL: dwarf->treetest({ 1, { 2, nil, nil }, { 3, nil, { 4, nil, nil } } }
<unfinished ...> in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: libdwarftest.so->treetest({ 2, nil, nil }) * = nil in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: libdwarftest.so->treetest({ 3, nil, { 4, nil, nil } } <unfinished ...> in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: libdwarftest.so->treetest({ 4, nil, nil }) * = nil in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: <... treetest resumed> ) * = { 5, nil, nil } in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: <... treetest resumed> ) * = { 2, { 3, nil, nil }, { 4, nil, { 5, nil,
nil } } } in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: libdwarftest.so->treetest({ 2, nil, nil }) * = nil in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: libdwarftest.so->treetest({ 3, nil, { 4, nil, nil } } <unfinished ...> in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: libdwarftest.so->treetest({ 4, nil, nil }) * = nil in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: <... treetest resumed> ) * = { 5, nil, nil } in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: test case execution failed
exec env LD_LIBRARY_PATH=$LD_LIBRARY_PATH_  {*}$LTRACE {*}$LTRACE_OPTIONS $BIN
{*}$LTRACE_ARGSltrace: type.c:100: type_struct_add: Assertion `field_info->type
!= ARGTYPE_VOID' failed.
    while executing
"exec env LD_LIBRARY_PATH=$LD_LIBRARY_PATH_  {*}$LTRACE {*}$LTRACE_OPTIONS $BIN
{*}$LTRACE_ARGS"
    invoked from within
"catch $command output"FAIL:  in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: libdwarftest.so->ftell(0x[0-9a-z]*) * = -1 in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: libdwarftest.so->strlen('t') * = 4 in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: dwarf->usleep(33) * = 0 in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: dwarf->nanosleep({ 0, 44 }, nil) * = 0 in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL:  in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: test case execution failed
exec env LD_LIBRARY_PATH=$LD_LIBRARY_PATH_  {*}$LTRACE {*}$LTRACE_OPTIONS $BIN
{*}$LTRACE_ARGSltrace: type.c:100: type_struct_add: Assertion `field_info->type
!= ARGTYPE_VOID' failed.
    while executing
"exec env LD_LIBRARY_PATH=$LD_LIBRARY_PATH_  {*}$LTRACE {*}$LTRACE_OPTIONS $BIN
{*}$LTRACE_ARGS"
    invoked from within
"catch $command output"FAIL:  in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: ftell@libc.so.6(0x[0-9a-z]* in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: strlen@libc.so.6('t') * = 4 in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: usleep@libc.so.6(33 <unfinished ...> in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: nanosleep@libc.so.6({ 0, 33000 }, nil) * = 0 in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: <... usleep resumed> ) * = 0 in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL: nanosleep@libc.so.6({ 0, 44 }, nil) * = 0 in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
FAIL:  in
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/dwarf.ltrace
for 0 times, should be 1
Running
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/filters.exp
...
FAIL: test case execution failed
exec env
LD_LIBRARY_PATH=/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/../ltrace
-o /tmp/lt-RYMO2ZC8Cd.ltrace -e*
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/filtltrace:
type.c:100: type_struct_add: Assertion `field_info->type != ARGTYPE_VOID'
failed.
    while executing
"exec env
LD_LIBRARY_PATH=/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main
/usr/src/packages/BUILD/ltrace-0.7.91+git2023..."
    ("eval" body line 1)
    invoked from within
"eval $command"FAIL: filt(\.pie)?->func1 appears in /tmp/lt-RYMO2ZC8Cd.ltrace 0
times, expected == 1
FAIL: libfilt1.so->func2 appears in /tmp/lt-RYMO2ZC8Cd.ltrace 0 times, expected
== 1
FAIL: libfilt2.so->puts appears in /tmp/lt-RYMO2ZC8Cd.ltrace 0 times, expected
== 1
FAIL: func2 resumed appears in /tmp/lt-RYMO2ZC8Cd.ltrace 0 times, expected == 1
FAIL: func1 resumed appears in /tmp/lt-RYMO2ZC8Cd.ltrace 0 times, expected == 1
FAIL: test case execution failed
exec env
LD_LIBRARY_PATH=/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/../ltrace
-o /tmp/lt-bU8KwolCaw.ltrace -e*
/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main/filt.pieltrace:
type.c:100: type_struct_add: Assertion `field_info->type != ARGTYPE_VOID'
failed.
    while executing
"exec env
LD_LIBRARY_PATH=/usr/src/packages/BUILD/ltrace-0.7.91+git20230705.8eabf68/testsuite/ltrace.main
/usr/src/packages/BUILD/ltrace-0.7.91+git2023..."
    ("eval" body line 1)
    invoked from within
"eval $command"FAIL: filt(\.pie)?->func1 appears in /tmp/lt-bU8KwolCaw.ltrace 0
times, expected == 1
FAIL: libfilt1.so->func2 appears in /tmp/lt-bU8KwolCaw.ltrace 0 times, expected
== 1
FAIL: libfilt2.so->puts appears in /tmp/lt-bU8KwolCaw.ltrace 0 times, expected
== 1
FAIL: func2 resumed appears in /tmp/lt-bU8KwolCaw.ltrace 0 times, expected == 1
FAIL: func1 resumed appears in /tmp/lt-bU8KwolCaw.ltrace 0 times, expected == 1

The testsuite fails with the exact assertion too. Running make check on the
upstream repo has the testsuite failing as well but without the assertion
showing up.

/usr/src/packages/SPECS/ltrace.spec has the patches that are applied on top of
the tarball. Commenting them out 

#Patch3:         ppc-ptrace.patch
#Patch5:         gcc9-printf-s-null-argument.patch
#Patch6:         lens-double-free.patch
#Patch7:         gcc9-Wlto-type-mismatch.patch
#Patch8:         s390x-ptrace.patch
#Patch9:         ppc64le-use-after-free.patch

rebuilding

$ sudo rpmbuild -bb /usr/src/packages/SPECS/ltrace.spec

reinstalling 

$ sudo zypper install
/usr/src/packages/x86_64/ltrace-0.7.91+git20230705.8eabf68-2.2.x86_64.rpm

does not make a difference, ltrace still fails. Unpacking the tarball 
ltrace-0.7.91+git20230705.8eabf68.tar.bz2 and comparing

$ diff -qr ~/repos/ltrace
/usr/src/packges/SOURCE/ltrace-0.7.91+git20230705.8eabf68

confusingly shows no difference in the source files besides what ./autogen.sh
created:

Only in /home/laube/repos/ltrace: aclocal.m4
Only in /home/laube/repos/ltrace: autom4te.cache
Only in /home/laube/repos/ltrace: BUILD
Only in /home/laube/repos/ltrace: config
Only in /home/laube/repos/ltrace: config.h.in
Only in /home/laube/repos/ltrace: configure
Common subdirectories: ltrace-0.7.91+git20230705.8eabf68/etc and
/home/laube/repos/ltrace/etc
Only in /home/laube/repos/ltrace: .git
Only in /home/laube/repos/ltrace: Makefile.in
Common subdirectories: ltrace-0.7.91+git20230705.8eabf68/sysdeps and
/home/laube/repos/ltrace/sysdeps
Common subdirectories: ltrace-0.7.91+git20230705.8eabf68/testsuite and
/home/laube/repos/ltrace/testsuite

I am out of ideas. 

I tried the test programs from above on a fresh AlmaLinux9.5 VM and the
assertion is not triggered at all.

Any suggestions what to try next?