Re: mod_perl Silent Failure, Very Mysterious

Hi Jie,
If I temporarily disable B::Hooks::OP::Check, as shown below, then it
tells me it was included from Devel::Declare, which was in turn
(indireclty) included from Method::Signatures::Simple.
I am not the author of either Catalyst or ShinyCMS, so unfortunately I
don't know any more in-depth details off the top of my head.
Does that answer your question?
Thanks,~ Will
wbraswell@cloud-comp0-00:~$ pm_location.sh
B::Hooks::OP::Check/home/wbraswell/perl5/lib/perl5/x86_64-linux-gnu-thread-multi/B/Hooks/OP/Check.pm
wbraswell@cloud-comp0-00:~$ mv
/home/wbraswell/perl5/lib/perl5/x86_64-linux-gnu-thread-multi/B/Hooks/OP/Check.pm
/home/wbraswell/perl5/lib/perl5/x86_64-linux-gnu-thread-multi/B/Hooks/OP/Check.pm.DISABLED
wbraswell@cloud-comp0-00:~/public_html/cloud-comp0-00.autoparallel.com-latest$
perl -e 'use ShinyCMS;'Can't locate B/Hooks/OP/Check.pm in @INC (you
may need to install the B::Hooks::OP::Check module) (@INC contains
....) at
/home/wbraswell/perl5/lib/perl5/x86_64-linux-gnu-thread-multi/Devel/Declare.pm
line 18.BEGIN failed--compilation aborted at
/home/wbraswell/perl5/lib/perl5/x86_64-linux-gnu-thread-multi/Devel/Declare.pm
line 18.Compilation failed in require at
/home/wbraswell/perl5/lib/perl5/x86_64-linux-gnu-thread-multi/Devel/Declare/Context/Simple.pm
line 5.BEGIN failed--compilation aborted at
/home/wbraswell/perl5/lib/perl5/x86_64-linux-gnu-thread-multi/Devel/Declare/Context/Simple.pm
line 5.Compilation failed in require at /usr/share/perl/5.22/base.pm
line 97. ...propagated at /usr/share/perl/5.22/base.pm line 106.BEGIN
failed--compilation aborted at
/home/wbraswell/perl5/lib/perl5/x86_64-linux-gnu-thread-multi/Devel/Declare/MethodInstaller/Simple.pm
line 3.Compilation failed in require at /usr/share/perl/5.22/base.pm
line 97. ...propagated at /usr/share/perl/5.22/base.pm line 106.BEGIN
failed--compilation aborted at
/home/wbraswell/perl5/lib/perl5/Method/Signatures/Simple.pm line
19.Compilation failed in require at blib/lib/ShinyCMS.pm line 21.BEGIN
failed--compilation aborted at blib/lib/ShinyCMS.pm line
21.Compilation failed in require at -e line 1.BEGIN
failed--compilation aborted at -e line 1.
On 11/22/2016 at 11:03 PM, "Jie Gao"  wrote:Hi William

Where excatly is B::Hooks::OP::Check required / its Check.xs called?
Regards,
Jie 

* William N. Braswell, Jr.  wrote:

> Date: Tue, 22 Nov 2016 06:04:36 -0600
> From: "William N. Braswell, Jr." 
> To: Jie Gao 
> CC: [email protected]
> Subject: RE: mod_perl Silent Failure, Very Mysterious
> 
> Hello Again Jie,
> I have run the modperl_testreport.pl script per your direction, and
> have attached the output file.
> Also, I have posted this as a mod_perl issue:
> 
> https://rt.cpan.org/Ticket/Display.html?id=118900
> I am attempting to run an already-bug-free Catalyst application
under
> mod_perl, which should work just fine but instead gives a segfault
in
> the Check.xs file.
> The name of the Catalyst application is ShinyCMS, it works just fine
> using the PSGI stand-alone testing web server as well as fastcgi,
and
> mod_perl is the only option which causes a segfault.  I must have
> mod_perl support for my specific use of ShinyCMS, so I really really
> need to fix this segfault and get Shiny working under mod_perl. 
Yes,
> Shiny is specifically designed to work with mod_perl, as are most
> Catalyst apps.  In fact, my simple ">" print statements in
ShinyCMS.pm
> show that Shiny is totally loading and returning control back to
> Apache (or mod_perl?) before the actual segfault occurs.
> My system info:
> $ apache2 -vServer version: Apache/2.4.18 (Ubuntu)Server built:  
> 2016-07-14T12:32:26
> $ perl -vThis is perl 5, version 22, subversion 1 (v5.22.1) built
for
> x86_64-linux-gnu-thread-multi(with 58 registered patches, see perl
-V
> for more detail)
> $ pm_version.pl mod_perl2$mod_perl2::VERSION = 2.000009
> $ pm_version.pl B::Hooks::OP::Check$B::Hooks::OP::Check::VERSION =
> 0.19
> $ cat /etc/issueUbuntu 16.04.1 LTS
> I have spent the last week trying all of the following and more, but
> the segfault persists:
> 1.  manually-compiled perl v5.22.1 & mod_perl2 v2.000009 w/out
> threads2.  mod_perl2 back to v2.0000053.  B::Hooks::OP::Check back
to
> v0.17 (maybe earlier)4.  move all possible 'use' statements to be
> called as early as possible in Shiny_dependencies.pm
> I also tried Apache v2.2 & mod_perl2 v2.000008 & Perl v5.18.2 in
> Ubuntu v14.04, it did not segfault but instead it served up totally
> blank pages with 0 bytes of data.  I have not dug deeper because I
use
> Ubuntu v16.04 in production and I figure it is harder to debug w/out
> an actual segfault to start from.
> I have set up SSH access on a cloud server so you can see and debug
> the segfault directly, all 3 of the usernames below are set as
sudoers
> so please wield your root powers with caution:
> HOSTNAME: cloud-comp0-00.autoparallel.comHOST IP: 
> 64.137.255.53USERNAME: shay OR fmoyer OR jgao OR
modperl_teamPASSWORD:
> check your CPAN e-mail address
> Pertinent paths:
>
/home/wbraswell/public_html/cloud-comp0-00.autoparallel.com-latest/etc/apache2/sites-available/cloud-comp0-00.autoparallel.com.conf
> Here is an example gdb showing Check.xs as the likely culprit,
> starting with `sudo` and then `source` to load Apache env vars:
> [[[ BEGIN GDB SESSION ]]]
> zefram@cloud-comp0-00:/$ sudo -i...root@cloud-comp0-00:~# source
> /etc/apache2/envvarsroot@cloud-comp0-00:~# gdb
> /usr/sbin/apache2...Reading symbols from /usr/sbin/apache2...(no
> debugging symbols found)...done.
> (gdb) run -k start -XStarting program: /usr/sbin/apache2 -k start
> -X[Thread debugging using libthread_db enabled]Using host
libthread_db
> library "/lib/x86_64-linux-gnu/libthread_db.so.1".>: top of
> ShinyCMS.pm>: in ShinyCMS.pm, about to use Moose>: in ShinyCMS.pm,
> about to use Catalyst::Runtime>: in ShinyCMS.pm, about to use
> Catalyst>: in ShinyCMS.pm, about to use CatalystX::RoleApplicator>:
in
> ShinyCMS.pm, about to use Method::Signatures::Simple>: in
ShinyCMS.pm,
> about to call config()>: in ShinyCMS.pm, have __PACKAGE__ =
ShinyCMS>:
> in ShinyCMS.pm, about to call setup()...>: in ShinyCMS.pm, returned
> from setup()>: in ShinyCMS.pm, about to return 1
> Program received signal SIGSEGV, Segmentation
fault.0x00007fffebabcb10
> in ?? ()(gdb) bt#0  0x00007fffebabcb10 in ?? ()#1 
0x00007ffff3e5302b
> in Perl_newUNOP (my_perl=my_perl@entry=0x5555577ddde0,
> type=type@entry=17, flags=, flags@entry=8192, first=0x5555562141b8)
at
> op.c:4811#2  0x00007ffff3e54a1d in Perl_newCVREF
> (my_perl=my_perl@entry=0x5555577ddde0, flags=flags@entry=8192, o=)
at
> op.c:9367#3  0x00007ffff3e8b686 in Perl_yylex
> (my_perl=my_perl@entry=0x5555577ddde0) at toke.c:6693#4 
> 0x00007ffff3e97228 in Perl_yyparse
> (my_perl=my_perl@entry=0x5555577ddde0, gramtype=gramtype@entry=258)
at
> perly.c:322...#44 0x000055555558709f in main ()
> (gdb) startThe program being debugged has been started already.Start
> it from the beginning? (y or n) yTemporary breakpoint 1 at
> 0x5555555867b0Starting program: /usr/sbin/apache2 -k start -X[Thread
> debugging using libthread_db enabled]Using host libthread_db library
> "/lib/x86_64-linux-gnu/libthread_db.so.1".
> Temporary breakpoint 1, 0x00005555555867b0 in main ()
> (gdb) break perl_parseFunction "perl_parse" not defined.Make
> breakpoint pending on future shared library load? (y or [n])
> yBreakpoint 2 (perl_parse) pending.
> (gdb) continueContinuing.
> Breakpoint 2, perl_parse (my_perl=0x55555581ae50,
> xsinit=0x7ffff420a280, argc=2, argv=0x7ffff7f88070, env=0x0) at
> perl.c:14731473    perl.c: No such file or directory.
> (gdb) watch -l PL_check[17]Hardware watchpoint 3: -location
> PL_check[17]
> (gdb) continueContinuing.
> Hardware watchpoint 3: -location PL_check[17]
> Old value = (OP *(*)(PerlInterpreter *, OP *)) 0x7ffff3e4d9e0 New
> value = (OP *(*)(PerlInterpreter *, OP *)) 0x7fffebabcb10
> 0x00007fffebabcd0e in hook_op_check (type=type@entry=OP_RV2CV,
> cb=cb@entry=0x7fffeb8b6d30 , user_data=user_data@entry=0x0) at
> Check.xs:6666                      PL_check[type] = check_cb;
> (gdb) info threads  Id   Target Id         Frame * 1    Thread
> 0x7ffff7fe2780 (LWP 6196) "/usr/sbin/apach" 0x00007fffebabcd0e in
> hook_op_check (type=type@entry=OP_RV2CV, cb=cb@entry=0x7fffeb8b6d30
,
> user_data=user_data@entry=0x0)    at Check.xs:66
> (gdb) info sourceCurrent source file is Check.xsCompilation
directory
> is
>
/home/wbraswell/.cpanm/work/1479735542.24787/B-Hooks-OP-Check-0.19Located
> in
>
/home/wbraswell/.cpanm/work/1479735542.24787/B-Hooks-OP-Check-0.19/Check.xsContains
> 106 lines.Source language is c.Producer is GNU C11 5.4.0 20160609
> -mtune=generic -march=x86-64 -g -O2 -fwrapv -fno-strict-aliasing
-fPIC
> -fstack-protector-strong.Compiled with DWARF 2 debugging format.Does
> not include preprocessor macro info.
> (gdb) continueContinuing.>: top of ShinyCMS.pm>: in ShinyCMS.pm,
about
> to use Moose>: in ShinyCMS.pm, about to use Catalyst::Runtime>: in
> ShinyCMS.pm, about to use Catalyst>: in ShinyCMS.pm, about to use
> CatalystX::RoleApplicator>: in ShinyCMS.pm, about to use
> Method::Signatures::Simple>: in ShinyCMS.pm, about to call
config()>:
> in ShinyCMS.pm, have __PACKAGE__ = ShinyCMS>: in ShinyCMS.pm, about
to
> call setup()...>: in ShinyCMS.pm, returned from setup()>: in
> ShinyCMS.pm, about to return 1
> ...
> Program received signal SIGSEGV, Segmentation
fault.0x00007fffebabcb10
> in ?? ()(gdb) bt#0  0x00007fffebabcb10 in ?? ()#1 
0x00007ffff3e5302b
> in Perl_newUNOP (my_perl=my_perl@entry=0x555559dd8ed0,
> type=type@entry=17, flags=, flags@entry=8192, first=0x55555824eb98)
at
> op.c:4811#2  0x00007ffff3e54a1d in Perl_newCVREF
> (my_perl=my_perl@entry=0x555559dd8ed0, flags=flags@entry=8192, o=)
at
> op.c:9367...#44 0x000055555558709f in main ()
> (gdb) print PL_check[17]$1 = (Perl_check_t) 0x7fffebabcb10
> (gdb) quit
> [[[ END GDB SESSION ]]]
> I have attached the unabbreviated gdb session and the system info in
> text files, for your reference.
> I am anxious (near desperate) to get this bug fixed so that I can
move
> forward with my work.  I will eagerly assist in any way I can,
please
> just let me know what I should do and I will do it!
> Thanks so much for your help.
> Perling,~ Will
> 
> On 11/16/2016 at 3:17 AM, "Jie Gao"  wrote:     For the completeness
> of issue reporting, please run the following script and post the
> output:
>  #!/usr/bin/perl
>  use strict;
>  use warnings FATAL => 'all';
>  use ModPerl::TestReport;
>  ModPerl::TestReport->new(@ARGV)->run;
>  Jie
> -------------------------
>  From: William N. Braswell, Jr. [[email protected]]
>  Sent: Tuesday, November 15, 2016 8:01 PM
>  To: Jie Gao
>  Cc: [email protected]
>  Subject: Re: mod_perl Silent Failure, Very Mysterious
>     Hello Jie & Friends, 
>   I got the core dump out of gdb and also some hopefully-helpful
> output. 
>   Here is the output from gdb (not the core dump which is binary of
> course): 
>   http://hastebin.com/okaxacuriq.scala
>  Then again maybe we have a problem because I'm missing debugging
> symbols inside /usr/sbin/apache2?
>  _Program received signal SIGSEGV, Segmentation fault.   
> 0x00007fffeed67a00 in ?? ()_
>  Note the _??_ part in the output above... 
>   Where do I go from here? 
>   Thanks, ~ Will 
>  On 11/15/2016 at 12:14 AM, "William N. Braswell, Jr."  wrote:   Hi
> Jie, 
>   I have captured the output of `strace -f /usr/sbin/apache2 -e
trace8
> -k start`: 
>   http://hastebin.com/unupocozop.pl 
>   Looks like it is dying with a segfault, but I'm not sure what is
> causing this??? 
>   Thanks so much for your help!!!  :-) 
>   ~ Will 
>   [[[ BEGIN PASTE SNIPPET ]]] 
>    [pid  6550] stat("blib/lib/DynaLoader.pmc", 0x7ffdda87f280) = -1
> ENOENT (No such file or directory) [pid  6550]
> stat("blib/lib/DynaLoader.pm", 0x7ffdda87f1b0) = -1 ENOENT (No such
> file or directory) [pid  6550] stat("/etc/perl/DynaLoader.pmc",
> 0x7ffdda87f280) = -1 ENOENT (No such file or directory) [pid  6550]
> stat("/etc/perl/DynaLoader.pm", 0x7ffdda87f1b0) = -1 ENOENT (No such
> file or directory) [pid  6550]
> stat("/usr/local/lib/x86_64-linux-gnu/perl/5.22.1/DynaLoader.pmc",
> 0x7ffdda87f280) = -1 ENOENT (No such file or directory) [pid  6550]
> stat("/usr/local/lib/x86_64-linux-gnu/perl/5.22.1/DynaLoader.pm",
> 0x7ffdda87f1b0) = -1 ENOENT (No such file or directory) [pid  6550]
> stat("/usr/local/share/perl/5.22.1/DynaLoader.pmc", 0x7ffdda87f280)
=
> -1 ENOENT (No such file or directory) [pid  6550]
> stat("/usr/local/share/perl/5.22.1/DynaLoader.pm", 0x7ffdda87f1b0) =
> -1 ENOENT (No such file or directory) [pid  6550]
> stat("/usr/lib/x86_64-linux-gnu/perl5/5.22/DynaLoader.pmc",
> 0x7ffdda87f280) = -1 ENOENT (No such file or directory) [pid  6550]
> stat("/usr/lib/x86_64-linux-gnu/perl5/5.22/DynaLoader.pm",
> 0x7ffdda87f1b0) = -1 ENOENT (No such file or directory) [pid  6550]
> stat("/usr/share/perl5/DynaLoader.pmc", 0x7ffdda87f280) = -1 ENOENT
> (No such file or directory) [pid  6550]
> stat("/usr/share/perl5/DynaLoader.pm", 0x7ffdda87f1b0) = -1 ENOENT
(No
> such file or directory) [pid  6550]
> stat("/usr/lib/x86_64-linux-gnu/perl/5.22/DynaLoader.pmc",
> 0x7ffdda87f280) = -1 ENOENT (No such file or directory) [pid  6550]
> stat("/usr/lib/x86_64-linux-gnu/perl/5.22/DynaLoader.pm",
> {st_mode=S_IFREG|0644, st_size=10748, ...}) = 0 [pid  6550]
> open("/usr/lib/x86_64-linux-gnu/perl/5.22/DynaLoader.pm", O_RDONLY)
=
> 5 [pid  6550] ioctl(5, TCGETS, 0x7ffdda87ef50) = -1 ENOTTY
> (Inappropriate ioctl for device) [pid  6550] lseek(5, 0, SEEK_CUR)  
 
>   = 0 [pid  6550] read(5, "# Generated from DynaLoader_pm.P"...,
8192)
> = 8192 [pid  6550] --- SIGSEGV {si_signo=SIGSEGV,
si_code=SEGV_MAPERR,
> si_addr=0x7f491064fa00} --- [pid  6550] chdir("/etc/apache2")      
=
> 0 [pid  6550] rt_sigaction(SIGSEGV, {SIG_DFL, [],
> SA_RESTORER|SA_INTERRUPT, 0x7f4918dea3d0}, {SIG_DFL, [],
> SA_RESTORER|SA_RESETHAND, 0x7f4918dea3d0}, 8) = 0 [pid  6550]
> kill(6550, SIGSEGV)         = 0 [pid  6550] rt_sigreturn({mask=[]}) 
 
>  = 139951873083648 [pid  6550] --- SIGSEGV {si_signo=SIGSEGV,
> si_code=SI_USER, si_pid=6550, si_uid=0} --- [pid  6549] +++ exited
> with 0 +++ +++ killed by SIGSEGV (core dumped) +++  
>   [[[ END PASTE SNIPPET ]]] 
>  On 11/14/2016 at 11:54 PM, "Jie Gao"  wrote:  Add the "-f" flag to
> strace and see what errors you get.
>  Regards,
>  Jie 
>  * William N. Braswell, Jr.  wrote:
>  > Date: Mon, 14 Nov 2016 23:27:53 -0600
>  > From: "William N. Braswell, Jr." 
>  > To: [email protected]
>  > Subject: mod_perl Silent Failure, Very Mysterious
>  > 
>  > Hello everyone, I'm trying to enable a web app (Catalyst
ShinyCMS)
> to
>  > run under mod_perl, it is definitely supposed to work but it
fails
>  > without any errors, and I've been trying for several days to get
it
>  > working.
>  > My system info:
>  > * Perl 5, version 22, subversion 1 (v5.22.1) built for
>  > x86_64-linux-gnu-thread-multi
>  > * libapache2-mod-perl2 v2.0.9-4ubuntu1
>  > * Apache/2.4.18 (Ubuntu)
>  > * $Apache::Test::VERSION = 1.39
>  > * $Catalyst::VERSION = 5.90112
>  > * $ShinyCMS::VERSION = 0.007
>  > I know the app already functions properly when running via the
>  > stand-alone test server script (PSGI) or fastcgi, and I am able
to
> run
>  > a trivial Catalyst test application successfully under mod_perl,
so
>  > this issue seems to have something to do with the interaction
> between
>  > the ShinyCMS app and mod_perl.
>  > I have added print debug statements in my app, so when I call my
> app
>  > from within the Apache site conf file then I can see the debug
>  > statements prove that it is successfully going all the way
through
> my
>  > app, and it even returns control to the embedded code w/in the
> Apache
>  > site conf file, but it still just silently fails anyway with the
>  > following non-descriptive error:
>  > "The apache2 instance did not start within 20 seconds. Please
read
> the
>  > log files to discover problems"
>  > Unfortunately the log files do not provide any additional
> information
>  > whatsoever!
>  > Here are my Apache config files and the output of trying to start
>  > Apache manually, you can see my ">" statements which prove that
my
>  > ShinyCMS app code is definitely executing all the way through and
>  > returning control to Apache:
>  > http://hastebin.com/rulixitole.php
>  > http://hastebin.com/wevicagano.rb
>  > This is a really mysterious issue, when I run `apache2ctl -e
debug
> -k
>  > start` as seen in the link above, it goes through the motions as
if
> it
>  > is starting correctly, and it does not give any error messages,
but
>  > when apache2ctl is done there simply are not any apache processes
>  > running...
>  > I only found 1 person on google with a possibly-similar issue,
but
> he
>  > told me that he never found a solution:
>  > https://github.com/timbunce/devel-nytprof/issues/89
>  > I also found 1 person on the mod_perl mailing list archives with
>  > another possibly-similar issue, but I have no way to follow up,
and
>  > I'm not sure of the best way to totally rebuild everything from
>  > source:
>  > 
>
https://mail-archives.apache.org/mod_mbox/perl-modperl/200207.mbox/%[email protected]%3E
>  > I have asked for help on IRC #httpd, user Unbeliever suggested I
>  > ascertain my real /usr/sbin/apache2 command line and run it
through
>  > strace, then post to this mailing list.
>  > I used -x in apache2ctl which showed me I was simply sourcing
>  > /tmp/apache2/envvars and then running `/usr/sbin/apache2 -e
trace8
> -k
>  > start` which are the exact same args as accepted by the
apache2ctl
>  > script.
>  > I have now run `strace /usr/sbin/apache2 -e trace8 -k start`
which
>  > generates 2.5MB of text output, I have pasted the last few pages
of
> it
>  > on hastebin, I can post the whole thing for download if needed:
>  > http://hastebin.com/idemogexiq.erl
>  > I am not sure how to interpret the results of the strace command
>  > output and debug this issue, help please?
>  > Any and all assistance is greatly appreciated! :-)
>  > Thanks,~ Will