Hi William Where excatly is B::Hooks::OP::Check required / its Check.xs called?
Regards, Jie * William N. Braswell, Jr. <william.brasw...@autoparallel.com> wrote: > Date: Tue, 22 Nov 2016 06:04:36 -0600 > From: "William N. Braswell, Jr." <william.brasw...@autoparallel.com> > To: Jie Gao <j....@sydney.edu.au> > CC: modperl@perl.apache.org > 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. [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