XS test hangs

[hvds]

Hi, I was trying to start investigating Perl/perl5#19439, but after installing the first 30-odd dependencies, I got as far as Devel-Debug-DBGp-0.22 and found that it seems to hang forever at the first of the XS tests:

...
t/616_xdebug_temporary_breakpoint.t .. ok   
All tests successful.
Files=48, Tests=351, 12 wallclock secs ( 0.21 usr  0.01 sys +  6.63 cusr  0.39 csys =  7.24 CPU)
Result: PASS
PERL_DL_NONLAZY=1 "/opt/with_cfavn/bin/perl" -I. "-Mt::lib::Harness" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(q[xs], 0, 'blib/lib', 'blib/arch')" t/*.t

Running tests for xs version

t/001_load.t ......................... 1/2 ^C

I tried this with a stock 5.34.0 perl and had the same result.

If I comment out the run_debugger() call in t/001_load.t, the following test file gives an indication of what may be a bigger problem:

t/003_xs_sub.t ....................... perl: inline.h:194: Perl_POPMARK: Assertion `(PL_markstack_ptr > PL_markstack) || !"MARK underflow"' failed.
t/003_xs_sub.t ....................... No subtests run 

I notice a few problems show up in Makefile.PL, I don't know if any of these are related; I assume some or all of the uninits are triggered by the XS => { 'perl5db.xs' => undef } clause, not sure if that's supposed to be a valid use:

% ${PERL} Makefile.PL
Checking if your kit is complete...
Warning: the following files are missing in your kit:
        README
Please inform the author.
Use of uninitialized value $_ in pattern match (m//) at /opt/v5.34.0-d/lib/perl5/5.34.0/ExtUtils/MM_Unix.pm line 1428.
Use of uninitialized value $_ in pattern match (m//) at /opt/v5.34.0-d/lib/perl5/5.34.0/ExtUtils/MM_Unix.pm line 1429.
Use of uninitialized value in sort at /opt/v5.34.0-d/lib/perl5/5.34.0/ExtUtils/MM_Any.pm line 713.
Use of uninitialized value in list assignment at /opt/v5.34.0-d/lib/perl5/5.34.0/ExtUtils/MM_Any.pm line 768.
Use of uninitialized value in sort at /opt/v5.34.0-d/lib/perl5/5.34.0/ExtUtils/MM_Unix.pm line 3143.
Use of uninitialized value in join or string at /opt/v5.34.0-d/lib/perl5/5.34.0/ExtUtils/MM_Unix.pm line 3143.
Generating a Unix-style Makefile
Writing Makefile for Devel::Debug::DBGp
Writing MYMETA.yml and MYMETA.json
% 

During build there are also a couple of macro redefined warnings for WIDEST_UTYPE and MUTABLE_PTR, but those look less likely to be an issue.

Please let me know if there's anything I can do to help diagnose this.

perl -V

Summary of my perl5 (revision 5 version 34 subversion 0) configuration:
  Commit id: 79a7b254d85a10b65126ad99bf10e70480569d68
  Platform:
    osname=linux
    osvers=5.4.0-87-generic
    archname=x86_64-linux
    uname='linux zen2 5.4.0-87-generic #98~18.04.1-ubuntu smp wed sep 22 10:45:04 utc 2021 x86_64 x86_64 x86_64 gnulinux '
    config_args='-des -Dcc=gcc -Dprefix=/opt/v5.34.0-d -Doptimize=-g -O0 -DDEBUGGING -Dusedevel -Uversiononly'
    hint=recommended
    useposix=true
    d_sigaction=define
    useithreads=undef
    usemultiplicity=undef
    use64bitint=define
    use64bitall=define
    uselongdouble=undef
    usemymalloc=n
    default_inc_excludes_dot=define
  Compiler:
    cc='gcc'
    ccflags ='-fwrapv -DDEBUGGING -fno-strict-aliasing -pipe -fstack-protector-strong -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2'
    optimize='-g -O0'
    cppflags='-fwrapv -DDEBUGGING -fno-strict-aliasing -pipe -fstack-protector-strong -I/usr/local/include'
    ccversion=''
    gccversion='7.5.0'
    gccosandvers=''
    intsize=4
    longsize=8
    ptrsize=8
    doublesize=8
    byteorder=12345678
    doublekind=3
    d_longlong=define
    longlongsize=8
    d_longdbl=define
    longdblsize=16
    longdblkind=3
    ivtype='long'
    ivsize=8
    nvtype='double'
    nvsize=8
    Off_t='off_t'
    lseeksize=8
    alignbytes=8
    prototype=define
  Linker and Libraries:
    ld='gcc'
    ldflags =' -fstack-protector-strong -L/usr/local/lib'
    libpth=/usr/local/lib /usr/lib/x86_64-linux-gnu /usr/lib /lib/x86_64-linux-gnu /lib
    libs=-lpthread -lnsl -ldl -lm -lcrypt -lutil -lc
    perllibs=-lpthread -lnsl -ldl -lm -lcrypt -lutil -lc
    libc=libc-2.27.so
    so=so
    useshrplib=false
    libperl=libperl.a
    gnulibc_version='2.27'
  Dynamic Linking:
    dlsrc=dl_dlopen.xs
    dlext=so
    d_dlsymun=undef
    ccdlflags='-Wl,-E'
    cccdlflags='-fPIC'
    lddlflags='-shared -g -O0 -L/usr/local/lib -fstack-protector-strong'


Characteristics of this binary (from libperl): 
  Compile-time options:
    DEBUGGING
    HAS_TIMES
    PERLIO_LAYERS
    PERL_COPY_ON_WRITE
    PERL_DONT_CREATE_GVSV
    PERL_MALLOC_WRAP
    PERL_OP_PARENT
    PERL_PRESERVE_IVUV
    PERL_USE_DEVEL
    USE_64_BIT_ALL
    USE_64_BIT_INT
    USE_LARGE_FILES
    USE_LOCALE
    USE_LOCALE_COLLATE
    USE_LOCALE_CTYPE
    USE_LOCALE_NUMERIC
    USE_LOCALE_TIME
    USE_PERLIO
    USE_PERL_ATOF
  Built under linux
  Compiled at Mar 25 2022 15:06:33
  %ENV:
    PERL="/opt/v5.34.0-d/bin/perl"
  @INC:
    /opt/v5.34.0-d/lib/perl5/site_perl/5.34.0/x86_64-linux
    /opt/v5.34.0-d/lib/perl5/site_perl/5.34.0
    /opt/v5.34.0-d/lib/perl5/5.34.0/x86_64-linux
    /opt/v5.34.0-d/lib/perl5/5.34.0

[hvds]

A bit more info: running a partially unrolled version of 001load.t with strace -ff -s1024 -etrace=write shows that the child process writes the same perl: inline.h:194: Perl_POPMARK: Assertion (PL_markstack_ptr > PL_markstack) || !"MARK underflow"' failed.message and then terminates with ABORT. (In the full strace, this happens just after dynaloadingperl5db.so`.)

It would certainly be an improvement if a timer could be wrapped around the $LISTEN->accept() call at some level, but clearly that isn't the underlying problem here.

[hvds]

Hmm, running 003_xs_sub.t just up to the XSLoader::load('dbgp-helper::perl5db') with $PERL -Dtsv the debug output finishes with:

STACK 0: MAIN
  CX 0: BLOCK  =>
  CX 1: SUB    =>
  retop=nextstate
  CX 2: SUB    => *  PV("dbgp-helper::perl5db"\0)  \CV(bootstrap)
  retop=leavesub

(6416:/opt/v5.34.0-d/lib/perl5/5.34.0/x86_64-linux/DynaLoader.pm:210)   entersub
MARK top  564e105a9524 0
MARK pop  564e105a9520 0
MARK top  564e105a9520 0
MARK pop  564e105a951c 0
perl: inline.h:194: Perl_POPMARK: Assertion `(PL_markstack_ptr > PL_markstack) || !"MARK underflow"' failed.

Stepping through in gdb shows that the first top/pop pair is from the invocation of boot_dbgp_helper__perl5db, the second is from call_pv("DB::setup_lexicals", G_VOID | G_DISCARD | G_NODEBUG); in reinit_my_cxt. So there appears to be a PUSHMARK missing somewhere, and after grepping perl's pod files for that and then reading perldoc perlcall, I tried the patch below, and found that I can rebuild and all tests pass.

I do not have a clue why this would be needed for me when it doesn't appear to be needed for anyone else.

--- perl5db.xs.old      2022-03-26 15:19:10.788517606 +0000
+++ perl5db.xs  2022-03-26 15:19:22.651916636 +0000
@@ -52,9 +52,10 @@
 #endif
 
 static void reinit_my_cxt(pTHX_ pMY_CXT) {
+    dSP;
     MY_CXT.ldebug = get_sv("DB::ldebug", 0);
     MY_CXT.stack_depth = gv_fetchpv("DB::stack_depth", 0, 0);
-
+    PUSHMARK(SP);
     call_pv("DB::setup_lexicals", G_VOID | G_DISCARD | G_NODEBUG);
 }
 

[hvds]

I do not have a clue why this would be needed for me when it doesn't appear to be needed for anyone else.

@andk notes in Perl/perl5#19439 (comment) that -DDEBUGGING is probably the difference here.