Mailing List Archive: ModPerl: Dev

mod_perl 2.0.6

Index | Next | Previous | View Threaded

fred at redhotpenguin

Feb 16, 2011, 6:28 PM

Post #1 of 9 (831 views)
Permalink

mod_perl 2.0.6

I'm going to roll a 2.0.6 RC in the last week of April in two months.
That should put us on track for quarterly releases.

In the meantime, if you have tuits, there are 19 bugs in the RT queue:

https://rt.cpan.org/Public/Dist/Display.html?Name=mod_perl

Overall I think mp 2.0.5 is looking *solid*, but I'll be putting some
effort towards a few of those bugs, and also exploring the
$r->spawn_proc_prog method that seems to have some issues in certain
situations.

I don't think we need to decouple the Apache-* modules from the 2.0.6
- it was really easy to release with the last stable version once I
read the branching doc Gozer wrote. I am considering pulling
Apache2::Build out of the mp core distro and setting that up like
Apache-Test. There aren't any mod_perl dependencies in there, so it
could be a great standalone module for other Alien::* and other
packages that try to bootstrap mod_perl.

Another nit I have to pick (Torsten maybe this would be up your alley)
is that @miyagawa posted a benchmark a while ago showing that his
Starman pure perl server was slightly faster than mod_perl. We should
correct this bottleneck for 2.0.6 :) Devel::NYTProf seems like it can
be used to get the low hanging fruit there, but I get the feeling
we'll need some XS hacking. I'll try to get his benchmark code and
post it here.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

torsten.foertsch at gmx

Feb 17, 2011, 2:16 AM

Post #2 of 9 (798 views)
Permalink

Re: mod_perl 2.0.6 [In reply to]

On Thursday, February 17, 2011 03:28:47 Fred Moyer wrote:
> I'm going to roll a 2.0.6 RC in the last week of April in two months.
> That should put us on track for quarterly releases.

but, no pressure, ;-)

I like that.

> In the meantime, if you have tuits, there are 19 bugs in the RT queue:
>
> https://rt.cpan.org/Public/Dist/Display.html?Name=mod_perl

- https://rt.cpan.org/Ticket/Display.html?id=45463

Maybe this is already resolved. I can think of a way how this can be caused by
the "mod_perl closes standard file descriptors" bug that has been fixed in
2.0.5.

- https://rt.cpan.org/Ticket/Display.html?id=54153

According to the code, the script is run in void context the same as in any
other context. The only difference is that apache closes the reading ends of
the child's STDOUT and STDERR (and the writing end of the child's STDIN). So,
it may be that the requestor's script dies due to a SIGPIPE. He does not
provide details.

- https://rt.cpan.org/Ticket/Display.html?id=64999

could be closed, me thinks. At least in my linux box it compiles and passes
all tests with perl 5.13.9.

> Another nit I have to pick (Torsten maybe this would be up your alley)
> is that @miyagawa posted a benchmark a while ago showing that his
> Starman pure perl server was slightly faster than mod_perl. We should
> correct this bottleneck for 2.0.6 :) Devel::NYTProf seems like it can
> be used to get the low hanging fruit there, but I get the feeling
> we'll need some XS hacking. I'll try to get his benchmark code and
> post it here.

Do you have a link?

However, my highest priority is merging the threading branch. It works on
Linux with a threaded and non-threaded perls. Testing and perhaps fixing on
Windows needs to be done. I have managed to compile and test perl on a XP VM.
That's a 1st step. But I won't mind if someone else would jump in.

svn co https://svn.eu.apache.org/repos/asf/perl/modperl/branches/threading

Torsten Förtsch

--
Need professional modperl support? Hire me! (http://foertsch.name)

Like fantasy? http://kabatinte.net

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

perrin at elem

Feb 17, 2011, 7:15 AM

Post #3 of 9 (797 views)
Permalink

Re: mod_perl 2.0.6 [In reply to]

On Wed, Feb 16, 2011 at 9:28 PM, Fred Moyer <fred [at] redhotpenguin> wrote:
> Another nit I have to pick (Torsten maybe this would be up your alley)
> is that @miyagawa posted a benchmark a while ago showing that his
> Starman pure perl server was slightly faster than mod_perl.

No offense to Miyagawa, but this is probably due to skipping some
parts of HTTP that Apache HTTPD implements and/or being less thorough
about preventing remote attacks.

- Perrin

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

david at kineticode

Feb 17, 2011, 8:42 AM

Post #4 of 9 (805 views)
Permalink

Re: mod_perl 2.0.6 [In reply to]

On Feb 17, 2011, at 7:15 AM, Perrin Harkins wrote:

>> Another nit I have to pick (Torsten maybe this would be up your alley)
>> is that @miyagawa posted a benchmark a while ago showing that his
>> Starman pure perl server was slightly faster than mod_perl.
>
> No offense to Miyagawa, but this is probably due to skipping some
> parts of HTTP that Apache HTTPD implements and/or being less thorough
> about preventing remote attacks.

That's an assertion in need of evidence.

David

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

perrin at elem

Feb 17, 2011, 10:17 AM

Post #5 of 9 (802 views)
Permalink

Re: mod_perl 2.0.6 [In reply to]

On Thu, Feb 17, 2011 at 11:42 AM, David E. Wheeler <david [at] kineticode> wrote:
> That's an assertion in need of evidence.

It's only a guess, and I don't care enough to go through the code for
evidence. I think people generally underestimate the complexity of
HTTP and the difficulty of writing a bug-free network server based on
it. Apache HTTPD is the result of years worth of debugging, and I
wouldn't trade it for a recently developed server.

I also remember Matt Sergeant discovering that with AxKit if he
created a mod_perl alternative that had no hooks at all for the Apache
API stages (e.g. auth) that it was measurably faster than mod_perl. I
don't think we'd want to trade the hooks for more speed though.

All that being said, there's always room for improvement in
performance and I'm sure mod_perl could be faster with some
optimization work.

- Perrin

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

david at kineticode

Feb 17, 2011, 10:27 AM

Post #6 of 9 (797 views)
Permalink

Re: mod_perl 2.0.6 [In reply to]

On Feb 17, 2011, at 10:17 AM, Perrin Harkins wrote:

> It's only a guess, and I don't care enough to go through the code for
> evidence. I think people generally underestimate the complexity of
> HTTP and the difficulty of writing a bug-free network server based on
> it. Apache HTTPD is the result of years worth of debugging, and I
> wouldn't trade it for a recently developed server.

Sure, that's reasonable.

> I also remember Matt Sergeant discovering that with AxKit if he
> created a mod_perl alternative that had no hooks at all for the Apache
> API stages (e.g. auth) that it was measurably faster than mod_perl. I
> don't think we'd want to trade the hooks for more speed though.

Well yes. But Plack doesn't have those hooks because its architecture is completely different. You can get the same effect as the hooks by using nested middleware.

> All that being said, there's always room for improvement in
> performance and I'm sure mod_perl could be faster with some
> optimization work.

True for any software, of course. :-)

Best,

David

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

pgollucci at p6m7g8

Feb 19, 2011, 10:49 AM

Post #7 of 9 (773 views)
Permalink

Re: mod_perl 2.0.6 [In reply to]

A bit aggressive, but I like it. More releases will definitely lead to
more code-wise contributors.

On 2/16/2011 9:28 PM, Fred Moyer wrote:
> I'm going to roll a 2.0.6 RC in the last week of April in two months.
> That should put us on track for quarterly releases.
>
> In the meantime, if you have tuits, there are 19 bugs in the RT queue:
>
> https://rt.cpan.org/Public/Dist/Display.html?Name=mod_perl
>
> Overall I think mp 2.0.5 is looking *solid*, but I'll be putting some
> effort towards a few of those bugs, and also exploring the
> $r->spawn_proc_prog method that seems to have some issues in certain
> situations.
>
> I don't think we need to decouple the Apache-* modules from the 2.0.6
> - it was really easy to release with the last stable version once I
> read the branching doc Gozer wrote. I am considering pulling
> Apache2::Build out of the mp core distro and setting that up like
> Apache-Test. There aren't any mod_perl dependencies in there, so it
> could be a great standalone module for other Alien::* and other
> packages that try to bootstrap mod_perl.
>
> Another nit I have to pick (Torsten maybe this would be up your alley)
> is that @miyagawa posted a benchmark a while ago showing that his
> Starman pure perl server was slightly faster than mod_perl. We should
> correct this bottleneck for 2.0.6 :) Devel::NYTProf seems like it can
> be used to get the low hanging fruit there, but I get the feeling
> we'll need some XS hacking. I'll try to get his benchmark code and
> post it here.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe [at] perl
> For additional commands, e-mail: dev-help [at] perl
>

--
------------------------------------------------------------------------
1024D/DB9B8C1C B90B FBC3 A3A1 C71A 8E70 3F8C 75B8 8FFB DB9B 8C1C
Philip M. Gollucci (pgollucci [at] p6m7g8) c: 703.336.9354
VP Apache Infrastructure; Member, Apache Software Foundation
Committer, FreeBSD Foundation
Consultant, P6M7G8 Inc.
Sr. System Admin, Ridecharge Inc.

Work like you don't need the money,
love like you'll never get hurt,
and dance like nobody's watching.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

pgollucci at p6m7g8

Feb 19, 2011, 10:50 AM

Post #8 of 9 (776 views)
Permalink

Re: mod_perl 2.0.6 [In reply to]

On 2/17/2011 5:16 AM, Torsten Förtsch wrote:
> Do you have a link?
>
> However, my highest priority is merging the threading branch. It works on
> Linux with a threaded and non-threaded perls. Testing and perhaps fixing on
> Windows needs to be done. I have managed to compile and test perl on a XP VM.
> That's a 1st step. But I won't mind if someone else would jump in.
>
> svn co https://svn.eu.apache.org/repos/asf/perl/modperl/branches/threading
>
> Torsten Förtsch
>
I don't think esp Windows or $arch support should be a blocker. You'll
certaintly get more support if you put it in trunk. Its not like we
don't have stable releases and branches. Also, we can always revert it,
thats why we have VCs.

I say merge it, the longer before the next release, the better.

--
------------------------------------------------------------------------
1024D/DB9B8C1C B90B FBC3 A3A1 C71A 8E70 3F8C 75B8 8FFB DB9B 8C1C
Philip M. Gollucci (pgollucci [at] p6m7g8) c: 703.336.9354
VP Apache Infrastructure; Member, Apache Software Foundation
Committer, FreeBSD Foundation
Consultant, P6M7G8 Inc.
Sr. System Admin, Ridecharge Inc.

Work like you don't need the money,
love like you'll never get hurt,
and dance like nobody's watching.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

fred at redhotpenguin

Feb 19, 2011, 9:46 PM

Post #9 of 9 (777 views)
Permalink

Re: mod_perl 2.0.6 [In reply to]

2011/2/17 Torsten Förtsch <torsten.foertsch [at] gmx>:
> However, my highest priority is merging the threading branch. It works on
> Linux with a threaded and non-threaded perls. Testing and perhaps fixing on
> Windows needs to be done. I have managed to compile and test perl on a XP VM.
> That's a 1st step. But I won't mind if someone else would jump in.
>
> svn co https://svn.eu.apache.org/repos/asf/perl/modperl/branches/threading

Failed the upload on 2.2.15 prefork, perl 5.10.0 with threads. Passed
every other test. Nothing noticeable in the error log, will try to
dig in deeper.

t/modules/cgiupload.....................NOK 1/2# Failed test 1 in
t/modules/cgiupload.t at line 37
t/modules/cgiupload.....................NOK 2/2# Failed test 2 in
t/modules/cgiupload.t at line 37 fail #2
t/modules/cgiupload.....................FAILED tests 1-2
Failed 2/2 tests, 0.00% okay

This is perl, v5.10.0 built for darwin-thread-multi-2level

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe [at] perl
For additional commands, e-mail: dev-help [at] perl

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads