Weirdness with XML-Encoding-1.01

deprecated has asked for the wisdom of the Perl Monks concerning the following question:

Hi, Monks...

I'm installing the Everything Engine to test out some ideas I've got for the monastery. I have had to install a few modules for it, and one of them is XML::Encoding. When I attempted to install it with CPAN, it barfed. So I went to install it by hand, and some curious things happened.

I would run :

perl Makefile.PL
make
make test

# and what would happen is this:

08:28:20 [alex : macachu : ~/.cpan/build/XML-Encoding-1.01]
[40] $ make test
PERL_DL_NONLAZY=1 /bin/perl -Iblib/arch -Iblib/lib -I/usr/local/lib/pe
+rl5/5.6.0/ppc-linux-thread -I/usr/local/lib/perl5/5.6.0 test.pl
1..7
ok 1
ok 2
ok 3
ok 4
ok 5
ok 6
ok 7
make: *** [test_dynamic] Segmentation fault (core dumped)
[download]

I thought this was kind of odd given that it seemed like all the tests were fine. So I went to have a look at the test script and saw this:

my $err = $@;

unless ($err and $err =~ /^Len plus byte > 256/) {
  print "not ";
}
print "ok 7\n";
[download]

that is the very end of the script. so it would appear that `make` is segfaulting at exit of the script. I changed

"print "ok 7\n";

# to...

print "ok 7\n"; exit 0;
[download]

and it, of course, exited, and `make` was happy.

What gives? for the curious, this is GNU Make 3.77 on linux/PowerPC.

Confused,
brother dep.

--
transcending "coolness" is what makes us cool.

Comment on Weirdness with XML-Encoding-1.01 Select or Download Code

Replies are listed 'Best First'.
Re: Weirdness with XML-Encoding-1.01 by mirod (Canon) on Feb 24, 2001 at 20:24 UTC
Extremely odd. I have the same problem on a linux/PentiumII with GNU Make 3.79.1. and 5.6.0 It runs fine on Sun/Solaris using GNU Make 3.77 but with 5.005_03. Subsequent analysis reveals that perl test.pl segfaults too. The last test is clearly the one that triggers it (I have removed all other tests and I still get it but if I remove the test it disappears). I tried adding code to clear $@ (and even $!) both directly and through eval-ing something that does not die, to na avail. The only reason for that behaviour seems to be the last `eval` in the test, which is supposed to fail and set $@: if you comment it out and set the error to the right value then you avoid the segfault. So we have a block eval, with a die thrown in the middle of the eval. Everything works OK (including code after this piece of code) until the program exits without an `exit` statement, in which case it segfaults. I guess you really found a bug in 5.6.0, maybe checking under 5.6.1 would find out whether it's already been fixed or whether it deserves a bug report.	[reply]
Re: Re: Weirdness with XML-Encoding-1.01 by MeowChow (Vicar) on Feb 25, 2001 at 04:27 UTC
Perhaps this is a bug in the XS interface of XML::Parser::Expat, rather than a bug in Perl? MeowChow s aamecha.s a..a\u$&owag.print	[reply]
Re: Re: Re: Weirdness with XML-Encoding-1.01 by rpc (Monk) on Feb 25, 2001 at 15:22 UTC
You're right, it's definitely a problem with the XS code. `Program received signal SIGSEGV, Segmentation fault. doContent (parser=0x824c7a0, startTagLevel=0, enc=0x401c2bc0, s=0x824e3d1 "</prefix>\n</encmap>\n", end=0x824e3e5 "", nextPtr=0x +0) at xmlparse.c:1709 1709 xmlparse.c: No such file or directory. (gdb)` [download] The fact that it's only segfaulting on exit, and not immediately makes me think its munging the stack, but I don't have the xmlparse source in front of me, so it's hard to tell.	[reply] [d/l]