Robert Spier wrote:
AFAIK it's an RT problem. Even if you go to the home page
http://rt.perl.org/perlbug/ and enter a ticket in a window suggesting
that this is a guest access, it'll still ask you for a password. I've
CC'ed Robert, who's the right person to ask about this.
Less of an RT problem,
I have written a small patch to solve this problem. Please have a look
at it. If it looks good, I will write tests and submit the patch.
The scheme is :
1. Populate %ENV before all Perl../Perl, PerlRequire and PerlModule.
2. After every Perl../Perl, PerlRequire and PerlModule, sync
server tables
hi all...
it seems you can't call exit from a handler unless ModPerl::Util::exit has
been preloaded. the attached tarball reproduces, but it's impossible (well,
difficult) to prove from the mp2 test suite with all it's preloading...
--Geoff
exit-bug.tar.gz
Description: GNU Zip compressed data
Philippe M. Chiasson wrote:
Got to the bottom of this one, and it turns out it's a bug in Perl-5.6.1 w/o
threads
perl_destruct(perl);
perl_free(perl);
Doesn't cleanup sufficiently and a subsequent perl_parse() will segfault
Turns out this has been eventually fixed in Perl 5.8.2. There were plenty
Philippe M. Chiasson wrote:
So, unless anybody feels especially curious and finds the missing
entries, I
think it's reasonable to require perl-5.8.2+ for static prefork builds.
Thoughts?
Cut the tail. We aint messing with perl bugs.
--
The story goes like this, I run a smoke of some 70+ mp2 configurations
every other night, and very often I had the test suite hang on startup.
when I'd run things manually everything will work just fine.
most of the time the error log will stop at:
[Wed Dec 22 03:57:01 2004] [notice] Digest:
On Wed, 22 Dec 2004, Stas Bekman wrote:
Date: Wed, 22 Dec 2004 18:11:30 -0500
From: Stas Bekman [EMAIL PROTECTED]
To: mod_perl Dev dev@perl.apache.org
Subject: how to raise entropy
The story goes like this, I run a smoke of some 70+ mp2 configurations
every other night, and very often I
Geoffrey Young wrote:
hi all...
it seems you can't call exit from a handler unless ModPerl::Util::exit has
been preloaded. the attached tarball reproduces, but it's impossible (well,
difficult) to prove from the mp2 test suite with all it's preloading...
I have 3 solutions. Thanks to Philippe for
Jie Gao wrote:
On Wed, 22 Dec 2004, Stas Bekman wrote:
Date: Wed, 22 Dec 2004 18:11:30 -0500
From: Stas Bekman [EMAIL PROTECTED]
To: mod_perl Dev dev@perl.apache.org
Subject: how to raise entropy
The story goes like this, I run a smoke of some 70+ mp2 configurations
every other night, and very
On Wed, 2004-12-22 at 18:28 -0500, Stas Bekman wrote:
Use /dev/urandom, which will not block.
It's not us who uses it, but Apache, so it's not under our control.
I think that if you just cat some stuff to /dev/urandom it will raise
entropy.
- Perrin
Pratik wrote:
I have written a small patch to solve this problem. Please have a look
at it. If it looks good, I will write tests and submit the patch.
The scheme is :
1. Populate %ENV before all Perl../Perl, PerlRequire and PerlModule.
2. After every Perl../Perl, PerlRequire and PerlModule, sync
Stas Bekman wrote:
Geoffrey Young wrote:
hi all...
it seems you can't call exit from a handler unless ModPerl::Util::exit
has
been preloaded. the attached tarball reproduces, but it's impossible
(well,
difficult) to prove from the mp2 test suite with all it's preloading...
I have
Perrin Harkins wrote:
On Wed, 2004-12-22 at 18:28 -0500, Stas Bekman wrote:
Use /dev/urandom, which will not block.
It's not us who uses it, but Apache, so it's not under our control.
I think that if you just cat some stuff to /dev/urandom it will raise
entropy.
it's read only on my system:
l
Geoffrey Young wrote:
Stas Bekman wrote:
Geoffrey Young wrote:
hi all...
it seems you can't call exit from a handler unless ModPerl::Util::exit
has
been preloaded. the attached tarball reproduces, but it's impossible
(well,
difficult) to prove from the mp2 test suite with all it's preloading...
On Wed, 2004-12-22 at 18:57 -0500, Stas Bekman wrote:
it's read only on my system
Change it then. The random(4) manpage on my system shows it being
written to as root, and if this is a closed system you can probably
allow another user to do that.
- Perrin
you wouldn't want to just import the namespaces required during
modperl_perl_core_global_init?
Sorry, Geoff, I don't understand what do you ask. You talk about the 3rd
solution?
well, kinda. isn't the issue that ModPerl::Util isn't imported? I mean,
clearly ModPerl::Util::exit has been
Perrin Harkins wrote:
On Wed, 2004-12-22 at 18:57 -0500, Stas Bekman wrote:
it's read only on my system
Change it then. The random(4) manpage on my system shows it being
written to as root, and if this is a closed system you can probably
allow another user to do that.
I guess so. I just didn't
Geoffrey Young wrote:
you wouldn't want to just import the namespaces required during
modperl_perl_core_global_init?
Sorry, Geoff, I don't understand what do you ask. You talk about the 3rd
solution?
well, kinda. isn't the issue that ModPerl::Util isn't imported? I mean,
clearly
Stas Bekman wrote:
Jie Gao wrote:
Use /dev/urandom, which will not block.
It's not us who uses it, but Apache, so it's not under our control.
Notice that Digest: generating secret for digest authentication ... is
coming from an apache module.
If mod_ssl is enabled, too, you could set the
Stas Bekman wrote:
Use /dev/urandom, which will not block.
It's not us who uses it, but Apache, so it's not under our control.
Notice that Digest: generating secret for digest authentication ... is
coming from an apache module.
I believe you can compile APR to use urandom instead of random,
heh, how about patching Apache::Test to try to write to /dev/u?random if
the server takes a long time to start, assuming that this is where things
block?
--
__
Stas BekmanJAm_pH -- Just Another mod_perl Hacker
Garrett Rooney wrote:
Stas Bekman wrote:
Use /dev/urandom, which will not block.
It's not us who uses it, but Apache, so it's not under our control.
Notice that Digest: generating secret for digest authentication ... is
coming from an apache module.
I believe you can compile APR to use urandom
Stas Bekman wrote:
in one window start:
cat /dev/random
in the other:
perl -le 'system (cat /proc/sys/kernel/random/entropy_avail), sleep 1
while 1'
3712
0
0
688
and if in the third window I try to feed /dev/random while doing the above:
perl -le 'system (echo stas /dev/random) while 1'
it makes
Geoffrey Young wrote:
You don't need to import it. We did that in mp1, in mp2 we do:
*CORE::GLOBAL::exit = \ModPerl::Util::exit.
but in C.
yah, ok.
for the record, I have no idea why the other exit() call in Apache/Test.pm
ends up working, but I suppose it's not important :)
It works under mp2
Stas Bekman wrote:
Stas Bekman wrote:
in one window start:
cat /dev/random
in the other:
perl -le 'system (cat /proc/sys/kernel/random/entropy_avail), sleep 1
while 1'
3712
0
0
688
and if in the third window I try to feed /dev/random while doing the above:
perl -le 'system (echo stas
Just tested http://www.vanheusden.com/aed/ on my box, and it feeds random data
into the pool from your audio speakers quite decently.
Turns out that you can't just write() to /dev/random, you have to be a bit
fancier than that to feed entropy back in.
Would make a nice CPAN module at some point.
26 matches
Mail list logo
