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.brasw...@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/%3c4a19f19a1bb4d51183220020ed06971401288...@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
modperl_testreport.out
Description: Binary data
segfault_gdb.out
Description: Binary data
segfault_versions.out
Description: Binary data
