perl-modperl mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William N. Braswell, Jr." <william.brasw...@autoparallel.com>
Subject RE: mod_perl Silent Failure, Very Mysterious
Date Tue, 22 Nov 2016 12:04:36 GMT
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. [william.braswell@autoparallel.com]
 Sent: Tuesday, November 15, 2016 8:01 PM
 To: Jie Gao
 Cc: modperl@perl.apache.org
 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: modperl@perl.apache.org
 > 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/%3C4A19F19A1BB4D51183220020ED069714012882B0@exch2.iso-ne.com%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       
Mime
View raw message