Example C program crashes with "cannot open shared object file"

[kamahen]

The traceback doesn't make sense. When I ran under gdb, the crash seems to have happened somewhere inside PL_initialise(), where it calls prologTopLevel() (pl-init.c:1247).

[peter@penguin man (master=)]$ swipl-ld -goal true -o calc calc.c calc.pl
% Disabled autoloading (loaded 30 files)
% Disabled autoloading (loaded 0 files)
% halt
[peter@penguin man (master=)]$ ./calc pi/2
ERROR: /home/peter/.local/lib/swipl/library/uri.pl:59: Initialization goal raised exception:
ERROR: uri: cannot open shared object file: No such file or directory
ERROR: In:
ERROR:   [18] throw(error(shared_object(open,'uri: cannot open shared object file: No such file or directory'),context(...,_4354)))
ERROR:   [16] <meta call>
ERROR:   [15] with_mutex('$foreign',load_foreign_library(foreign(uri),uri,[])) <foreign>
ERROR:   [11] '$run_init_goal'('$syspreds':use_foreign_library_noi(...)) at /home/peter/src/swipl-devel/boot/init.pl:823
ERROR:   [10] catch(system:'$run_init_goal'(...),_4488,system:'$initialization_error'(_4510,...,...)) at /home/peter/src/swipl-devel/boot/init.pl:563
ERROR:    [9] catch_with_backtrace(system:'$run_init_goal'(...),_4546,system:'$initialization_error'(_4568,...,...)) at /home/peter/src/swipl-devel/boot/init.pl:643
ERROR: 
ERROR: Note: some frames are missing due to last-call optimization.
ERROR: Re-run your program in debug mode (:- debug.) to get more detail.
ERROR: /home/peter/.local/lib/swipl/library/time.pl:134: Initialization goal raised exception:
ERROR: time: cannot open shared object file: No such file or directory
ERROR: In:
ERROR:   [18] throw(error(shared_object(open,'time: cannot open shared object file: No such file or directory'),context(...,_218)))
ERROR:   [16] <meta call>
ERROR:   [15] with_mutex('$foreign',load_foreign_library(foreign(time),time,[])) <foreign>
ERROR:   [11] '$run_init_goal'('$syspreds':use_foreign_library_noi(...)) at /home/peter/src/swipl-devel/boot/init.pl:823
ERROR:   [10] catch(system:'$run_init_goal'(...),_352,system:'$initialization_error'(_374,...,...)) at /home/peter/src/swipl-devel/boot/init.pl:563
ERROR:    [9] catch_with_backtrace(system:'$run_init_goal'(...),_410,system:'$initialization_error'(_432,...,...)) at /home/peter/src/swipl-devel/boot/init.pl:643
ERROR: 
ERROR: Note: some frames are missing due to last-call optimization.
ERROR: Re-run your program in debug mode (:- debug.) to get more detail.
1.5707963267948966
[peter@penguin man (master=)]$ ldd ./calc
	linux-vdso.so.1 (0x00007ffd9c1d4000)
	libswipl.so.9 => /home/peter/.local/lib/swipl/lib/x86_64-linux/libswipl.so.9 (0x00007ec3c7af4000)
	libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007ec3c7907000)
	libtinfo.so.6 => /lib/x86_64-linux-gnu/libtinfo.so.6 (0x00007ec3c78d8000)
	libgmp.so.10 => /lib/x86_64-linux-gnu/libgmp.so.10 (0x00007ec3c7857000)
	libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007ec3c783a000)
	libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007ec3c7832000)
	libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007ec3c76ee000)
	libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007ec3c76cc000)
	/lib64/ld-linux-x86-64.so.2 (0x00007ec3c7c9d000)

[kamahen]

Probably unrelated, but it seems that the wrong compiler is used by swipl-ld:

$ swipl-ld --version
cc (Debian 10.2.1-6) 10.2.1 20210110
Copyright (C) 2020 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

but I built with:

$ gcc --version
gcc (GCC) 13.2.0
Copyright (C) 2023 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

[JanWielemaker]

Probably requires --no-pce to be passed to creating the saved state. Probably that should be default. Not sure what selects the default compiler these days. May be hardcoded to use cc. It is unrelated.

[kamahen]

It seems that gcc is used for the link step by /usr/bin/cc is hard-coded for the step that generates the .o file.

I tried running the individual steps manually, changing the /usr/bin/cc to gcc and adding --pce=no to the swipl/qsave line. It didn't make a difference (and ldd showed the same libraries):

++ swipl --dump-runtime-variables
+ eval 'PLBASE="/home/peter/.local/lib/swipl";' 'PLARCH="x86_64-linux";' 'PLBITS="64";' 'PLVERSION="90114";' 'PLSOEXT="so";' 'PLSOPATH="LD_LIBRARY_PATH";' 'PLLIBDIR="/home/peter/.local/lib/swipl/lib/x86_64-linux";' 'PLLIB="-lswipl";' 'PLLIBSWIPL="/home/peter/.local/lib/swipl/lib/x86_64-linux/libswipl.so.9";' 'PLSHARED="yes";' 'PLTHREADS="yes";'
++ PLBASE=/home/peter/.local/lib/swipl
++ PLARCH=x86_64-linux
++ PLBITS=64
++ PLVERSION=90114
++ PLSOEXT=so
++ PLSOPATH=LD_LIBRARY_PATH
++ PLLIBDIR=/home/peter/.local/lib/swipl/lib/x86_64-linux
++ PLLIB=-lswipl
++ PLLIBSWIPL=/home/peter/.local/lib/swipl/lib/x86_64-linux/libswipl.so.9
++ PLSHARED=yes
++ PLTHREADS=yes
+ gcc -c -g -O0 -D_REENTRANT -D__SWI_PROLOG__ -D__SWI_EMBEDDED__ -I/home/peter/.local/lib/swipl/include -o calc.o calc.c
+ gcc -o calc -g -Wl,-rpath=/home/peter/.local/lib/swipl/lib/x86_64-linux calc.o -L/home/peter/.local/lib/swipl/lib/x86_64-linux -lswipl
+ swipl --pce=no -f none -F none -g true -t 'consult(['\''calc.pl'\'']),qsave_program('\''pltmp-8637'\'',[goal=true,toplevel=prolog,init_file=none])'
% Disabled autoloading (loaded 30 files)
% Disabled autoloading (loaded 0 files)
% halt
+ cat pltmp-8637
+ chmod 755 calc
+ rm pltmp-8637

The expected external references were there:

$ nm calc
0000000000404068 B __bss_start
0000000000404068 b completed.0
0000000000404058 D __data_start
0000000000404058 W data_start
00000000004010f0 t deregister_tm_clones
00000000004010e0 T _dl_relocate_static_pie
0000000000401160 t __do_global_dtors_aux
0000000000403dd8 d __do_global_dtors_aux_fini_array_entry
0000000000404060 D __dso_handle
0000000000403de0 d _DYNAMIC
0000000000404068 D _edata
0000000000404070 B _end
0000000000401344 T _fini
0000000000401190 t frame_dummy
0000000000403dd0 d __frame_dummy_init_array_entry
000000000040214c r __FRAME_END__
0000000000404000 d _GLOBAL_OFFSET_TABLE_
                 w __gmon_start__
0000000000402010 r __GNU_EH_FRAME_HDR
0000000000401000 T _init
0000000000403dd8 d __init_array_end
0000000000403dd0 d __init_array_start
0000000000402000 R _IO_stdin_used
                 w _ITM_deregisterTMCloneTable
                 w _ITM_registerTMCloneTable
0000000000401340 T __libc_csu_fini
00000000004012e0 T __libc_csu_init
                 U __libc_start_main@GLIBC_2.2.5
0000000000401196 T main
                 U PL_call_predicate
                 U PL_halt
                 U PL_initialise
                 U PL_new_term_refs
                 U PL_predicate
                 U PL_put_atom_chars
0000000000401120 t register_tm_clones
00000000004010b0 T _start
                 U strcpy@GLIBC_2.2.5
                 U strlen@GLIBC_2.2.5
0000000000404068 D __TMC_END__

[kamahen]

When I copied calc.{c,cpp} to another directory, everything worked. So, there appears to be something in the man directory that's being picked up, causing the problem. One hint is that when I stripped the code down to the minimum (just calling PL_initialise()), and removed the calc.pl from the swipl-ld line, I still got a Could not find system resources error.

[kamahen]

Update - it appears to not be anything in the man directory but something to do with where the directory is.
When I copied (recursively) the contents of the man directory to /tmp, everything worked fine.

[JanWielemaker]

May well be that the algorithm to find the home dir is disturbed at some place by the stuff that is in the build environment. See docs for the various steps tried to find the resource file.

[kamahen]

That seems to be the problem -- when I set the environment variable SWI_HOME_DIR or SWIPL, it works properly.

The problem seems to be that at pl_init.c:314, the code determines that directory /home/peter/src/swipl-devel exists, and it then uses /home/peter/src/swipl-devel/swipl.home (because PLHOMEFILE is defined), which contains "." . Presumably this is desired behaviour for some testing?

[kamahen]

After all, the traceback doesn't indicate a crash -- calc.c runs (and calls calc/1 (in calc.pl, which it is able to find even though it can't findthe "home" directory) ... the output is the one line without ERROR: (which is to stdout; the error lines are to stderr).

Is the traceback from a different thread than what runs calc.c's main()?

I read the documentation on resources (which doesn't mention the swipl.home file -- I got that from tracing the code) but I'm a bit confused about how it works or what uses it. The traceback is missing a bit of information. Also, does swipl-ld --no-pce do anything? - swipl-ld seems to just ignore options it doesn't understand, and the -v output showed no difference with no-pce (presumably, it's pce that wants library(time) and library(uri)).

[kamahen]

I tried running swipl-ld with `-pl-options '--no-pce' but misbehaved (I'll try to figure this out later).

Instead, I ran the commands by hand, adding --no-pce to the qsave step:

swipl --no-pce -f none -F none -g true -t "consult(['calc.pl']),qsave_program('pltmp-2246',[goal=true,toplevel=prolog,init_file=none])"

This also got the traceback (to stderr) and the expected output (to stdout).
So, pce doesn't seem to be the cause.

I wonder if this should be debugged using --threads=false ... if so, how would I specify this to the qsave step?

[JanWielemaker]

--no-pce doesn't seem to be necessary as it uses -F none, which also disables xpce. It works for me when not creating the sate in the build tree but copying cal.c and calc.pl to some other directory. I wonder why ...

[JanWielemaker]

I see. In both cases the state includes library(ansi_term) which depends on uri_file_name/2 and thus includes library(uri) which depends on the uri.so foreign object. When started outside the Prolog source tree, it still gets the actual installation location and can load the foreign object. When started inside the man dir, using swipl.home in the parent directory of the executable, it thinks the home is the source tree. But now, the foreign object is not in the expected place and thus we cannot find the uri object.

Now it is a little unclear what to do. One is to use the foreign(save) option of qsave_program to include uri.so in the saved state. Another is to avoid the dependency on uri_file_name/2 or, probably even better on library(ansi_term). Better control over the dependency is probably the best way to go in the long term. How exactly is non-trivial. Part of the unwanted dependencies consider the support for debugging and interactive behavior.

[JanWielemaker]

Pushed a commit to remove the old swipl.home and swipl.rc from the top of the source. These were old remains from the pre-cmake build setup. This makes calc in man work, but does not resolve the issue otherwise.

[kamahen]

Is there a way to turn on trace at the beginning of the executable?

Anyway, I noticed that toplevel.pl has predicate setup_colors/0, which uses library(ansi_term). There's also a call to library(time) in the traceback (which also seems to come from library(ansi_term) - see below). setup_colors/0 is called from initalise_prolog/0, which is called from $initialise/0.

I tried this, but it didn't fix the problem (I reinstated swipl.home, swipl.rc so that the bug would be trigered):

~/src/swipl-devel/build/src/swipl-ld  -pl-options='--no-threads' -v -o calc -class runtime -cc gcc -ld gcc -goal true -toplevel halt calc.c calc.pl

which expanded to:

swipl --no-threads -f none -F none -g true -t "consult(['calc.pl']),qsave_program('pltmp-3995',[goal=true,toplevel=halt,init_file=none,class=runtime])"

hmmm ... shouldn't it have called ~/src/swipl-devel/build/src/swipl instead of swipl? Anyway, I solved that by adding -pl /home/peter/src/swipl-devel/build/src/swipl and got this (no surprise; the binary is in the build tree):

	/home/peter/src/swipl-devel/build/src/swipl --no-threads -f none -F none -g true -t "consult(['calc.pl']),qsave_program('pltmp-4083',[goal=true,toplevel=halt,init_file=none,class=runtime])"
ERROR: /home/peter/src/swipl-devel/build/home/library/uri.pl:59:
ERROR:    /home/peter/src/swipl-devel/build/home/library/uri.pl:59: Initialization goal raised exception:
ERROR:    open_shared_object/3: uri: cannot open shared object file: No such file or directory
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_iri/2 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_encoded/3 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_authority_components/2 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_query_components/2 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_normalized_iri/3 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_resolve/3 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_normalized/3 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_normalized_iri/2 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:iri_normalized/2 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_normalized/2 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:iri_normalized/3 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_components/2 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:45:
ERROR:    Exported procedure uri:uri_is_global/1 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/time.pl:134:
ERROR:    /home/peter/src/swipl-devel/build/home/library/time.pl:134: Initialization goal raised exception:
ERROR:    open_shared_object/3: time: cannot open shared object file: No such file or directory
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:47:
ERROR:    Exported procedure time:alarm_at/4 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:47:
ERROR:    Exported procedure time:uninstall_alarm/1 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:47:
ERROR:    Exported procedure time:install_alarm/2 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:47:
ERROR:    Exported procedure time:remove_alarm/1 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:47:
ERROR:    Exported procedure time:install_alarm/1 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:47:
ERROR:    Exported procedure time:alarm/3 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:47:
ERROR:    Exported procedure time:alarm_at/3 is not defined
ERROR: /home/peter/src/swipl-devel/build/home/library/ansi_term.pl:47:
ERROR:    Exported procedure time:alarm/4 is not defined
% Disabled autoloading (loaded 30 files)
% Disabled autoloading (loaded 0 files)
% halt

[kamahen]

I tried commenting out setup_colors/0 (plus a few other "setup" predicates") in toplevel.pl and that didn't change anything.