Apache::Reload - patch - fixes problems with using dynamic @INC
I started using a dynamic INC (set up in a TransHandler), and discovered
that Apache::Reload (v0.07) was not doing its job correctly in that case.
Note, changing INC in a transhandler won't have the desired Apache::Reload
effects unless the PerlInitHandler for Apache::Reload is placed in a
Location section of httpd.conf.
The following patch corrects this, though it adds a few extra stat() calls
if the module is found at the end of INC instead of near the beginning (as
mine are). You could make this behavior an optional one for strange folks
like me :). Oh, I just added that to the patch manually:
PerlSetVar ReloadDynamicInc 1
...will enable this INC-scanning behavior. I think I hacked the patch
correctly :)
Note, if a module is loaded with one INC, then on a subsequent request,
INC changes such that the module would be not-found, the module is not
removed from memory. If you do that and use it from a script (served when
INC doesn't have the module), you'll get erratic behavior - it'll work fine
sometimes and error out other times. So don't do that., as a wise man
said. Instead, make sure INC always has the right libs for your script -
test from a freshly-started server to be sure.
Separate from this strange Dynamic-@INC use case, there's another bug in the
release version of the module: if you moved a module from one INC dir to
another, the INC loop at the top of handler() would not find the file, as
it was looking for inc-dir/full-path-name instead of
inc-dir/relative-path-name. This patch should fix this too, whether
ReloadDynamicInc is used or not.
I haven't tested this with wildcard settings or ReloadAll.
Enjoy,
Randy
--- Reload.pm Sat Oct 12 16:22:02 2002
+++ Reload.pm.new Sat Oct 12 17:08:07 2002
-108,35 +108,40
while (my($key, $file) = each %Apache::Reload::INCS) {
local $^W;
warn Apache::Reload: Checking mtime of $key\n if $DEBUG;
my $mtime = (stat $file)[9];
-unless (defined($mtime) $mtime) {
-for (INC) {
-$mtime = (stat $_/$file)[9];
-last if defined($mtime) $mtime;
-}
-}
+
+my $found = $file;
+if( $r-dir_config(ReloadDynamicInc) || !$mtime ) {
+# seek out the file in INC.
+my $mt;
+for (INC) {
+$mt = (stat $_/$key)[9];
+$mtime = $mt, $found = $_/$key, last if defined($mt)
$mt;
+}
+}
+
warn(Apache::Reload: Can't locate $file\n),next
unless defined $mtime and $mtime;
unless (defined $Stat{$file}) {
$Stat{$file} = $^T;
}
-if ($mtime $Stat{$file}) {
+if( $found ne $file || $mtime $Stat{$file} ) {
delete $INC{$key};
if (my $symref = $UndefFields{$key}) {
#warn undeffing fields\n;
no strict 'refs';
undef %{$symref};
Apache::DProf not working
I'm having trouble making Apache::DProf work. I've installed the module with CPAN, and I've added PerlModule Apache::DProf at the top of my httpd.conf. I've verified with 'httpd -l' that PerlChildInitHandler is OK - therefore, pushing a child-init handler should be OK, right? I've started the server, ran the web page I want to profile (that uses HTML::Mason), and then stopped the server. ServerRoot/dprof is not created, nor when I create it does it create dprof/$$ files as documented (likewise ServerRoot/logs/dprof/, as described elsewhere in the perldoc - not self-consistent). Also searched the entire system for tmon.out files (not found). Any suggestions? Randy
RE: New mod_perl site and oddness with IE
IE 5.5 on win2k as well. -Original Message- From: Issac Goldstand [mailto:[EMAIL PROTECTED]] Not just you. I have the same problem under MSIE. Has anyone else had problems with this particular page under IE (6.0.2600 under XP) being extremly slow to update when paging up/down? http://perl.apache.org/docs/1.0/guide/performance.html#Improving_Perform
RE: TransHandler called a second time after I've returned DECLINED
Rick, I hope you don't mind I've followed this up to the list. Your reply
was *very* useful to me, and I hope others may benefit from it...since I
didn't find any archived information that was helpful and all.
-Original Message-
From: Rick Myers [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, July 03, 2002 7:00 PM
Sorry for the late reply, but I don't read the list daily.
(Erm, or even weekly it seems. :-)
On Jun 25, 2002 at 08:26:37 -0700, Randy Harmon wrote:
Heh, the weeks do tend to fly by. There goes another one now.
I'm using a TransHandler, and having a problem where it sometimes gets
called twice when I don't expect it
[...]
On an example request for /exists/non-existent-dir/file.html,
[on the second, unexpected, TransHandler call]
the uri() is screwed up as /file.html. If the request is
/exists/non/foo/file.html, then uri() is /foo/file.html on
the second (unexpected) call to my TransHandler.
[Rick:]
That's an example of Apache's simplistic approach to
matching URL's to physical disk space. I don't recall if it
was mod_perl or Apache that actually called the TransHandler
twice, but I suspect the latter. Here's how I worked around
it...
PerlPostReadRequestHandler AlterTrans
package AlterTrans
sub handler {
my ($r) = @_;
return DECLINED if $blah_blah_blah;
for (@{$r-get_handlers('PerlTransHandler')}) {
/::handler$/ or $_ .= '::handler';
my $rv = {\$_}($r);
$rv != DECLINED and return $rv;
}
$r-set_handlers('PerlTransHandler', undef);
OK;
}
Basically, what it does is calls your TransHandlers one at a
time from the PostReadRequest phase.
--rick
Hmm.. maybe this is of interest?
NAME
Apache::AlterTransChain - Alters the chaining behaviour of
[cut: rest of perldoc]
That is interesting. I remain with trouble, in that my TransHandler relies
on things that apparently aren't available at the PostReadRequest phase -
specifically dir_config settings don't seem to be available when I use this
approach.
I was successful in using $r-set_handlers('PerlTransHandler', undef) in my
TransHandler, which apparently avoids the implicit subrequest that occurs
after I decline my TransHandler for a non-existent file like described
above.
There's still an interesting issue, where a subrequest that I *intend* to do
(while serving a site-branded error message) still has the original
filename - so I had to call $r-filename('/path/to/error_message.html')
before my subrequest would do the right thing. But that problem is
independent of my above fix.
Thanks for the information, Rick. Very helpful!
Randy
TransHandler called a second time after I've returned DECLINED
I'm using a TransHandler, and having a problem where it sometimes gets
called twice when I don't expect it - in most cases it's called just once as
I expect.
When I specify a file in a directory that doesn't exist (I'm going to use
path_info in a Mason dhandler, either to deliver a custom 404 with
autohandler, or for various other nefarious purposes), my TransHandler is
being called twice. The first time, it gets the correct $r-uri(), and I
determine that I don't want to translate the request - returning DECLINED.
The second time, the uri() is jimmied.
On an example request for /exists/non-existent-dir/file.html, the uri() is
screwed up as /file.html. If the request is /exists/non/foo/file.html, then
uri() is /foo/file.html on the second (unexpected) call to my TransHandler.
I've trimmed my server's vhost configuration down to this bare minimum:
VirtualHost IP:PORT
ServerAdmin [EMAIL PROTECTED]
ServerName randy.www.customer.com
DocumentRoot/www/home/cust/dev/randy/www.customer.com/
PerlTransHandler My::Trans
Location /
SetHandler perl-script
PerlHandler My::MasonHandler
order allow,deny
allow from all
/Location
/VirtualHost
Note, if I remove the PerlHandler, the TransHandler isn't getting called at
all, even though I've specified it. The two packages are defined in a
single init script that's still doing fine for other vhosts.
I've troubleshot quite a bit, and the call-stack when I'm called twice is
exactly the same (/dev/null is the caller). The pid and perl interpreter is
the same for both, so I can work around it with a package-global variable
and $r-uri( $foo ) and a $r-register_cleanup( sub{$foo=0}). But not with
$r-pnotes() - it doesn't hold the value!
I don't see anything My::MasonHandler is doing (push_handlers, for instance)
that should cause a second transhandler to be called for - besides which,
that code isn't run until later anyway.
perl 5.005_03
Apache.pm v1.27
compiled into apache, not DSO
I know this Perl version isn't up to the moment, but I've had various
problems with the more up-to-date perls, and none with this version - unless
this turns out to be the exception :)
Has anybody experience with such a problem? Anybody using this perl and
mod_perl with a transhandler of such a nature, with success? Anyone have
suggestions I should try?
Thanks,
Randy
RE: Re[2]: mod_perl memory leaks on Windows
Apache 1.3.x has always been experimental and not-for-production-use on
Win32 :(
Hopefully these modules will support Apache 2.0 pretty soon, for your sake.
Randy
-Original Message-
From: Andrey Prokopenko [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, June 25, 2002 8:31 AM
To: [EMAIL PROTECTED]
Subject: Re[2]: mod_perl memory leaks on Windows
Hello Perrin,
Tuesday, June 25, 2002, 6:40:01 PM, you wrote:
PH Andrey Prokopenko wrote:
After a fresh restart, I started Apache/mod_perl. Then i issued a
little stress test using simple perl script with LWP::Simple.
I ran a performance test on /mod_perl/index.pl page for 10 minutes. The
source code of that page is given below :
-
#!/perl/bin/perl
use CGI qw(:all);
print header;
print Hello, World!;
-
After 10 mintues of execution, the memory consumption of Apahce.exe
had jumped from approx 5 MB to 120MB ( virtual memory ).
I ve noticed, that Apache child grown by 4-8kb per each request.
PH Have you tried it without using CGI.pm? Have you tried writing a
PH handler instead of using Apache::Registry?
yep, i wtite a totally simple cgi
--
#!/perl/bin/perl
use strict;
print Content-type: text/plain\n\n;
print Hello, World!;
--
and the same simpliest handler:
--
package MyTest;
use Apache;
use Apache::Constants;
sub handler{
my $r = shift;
print $r-send_http_header(text/html);
print h1HELLO WORLD!/h1;
return OK;
}
1;
--
and still no luck ;(((
Apache child still continued to grow in both cases. ;(((
I even turned off KeepAlive option in httpd.conf
I know that mod_perl on Windows is a development version but still
this memory leak thing is pretty obvious and should have been taken
care off.
PH There's a pretty good chance that it's not mod_perl causing
it. It may
PH be just that Win32 perl grows a little when you run that CGI::header
PH call over and over.
I tried a plain Perl cgi script, with no module used, and still the
same. ;((
Do you mean that this leak cannot be fixed from Apache/mod_perl side ?
PH The upcoming mod_perl 2 will have better support for Windows.
You can
PH find information on it here: http://perl.apache.org/release/docs/
Sorry, but for now we're stuck with Apache 1.3.x, because we use
module, which
cannot work with Apache 2.0. So i seek appropriate solution based on
Apache 1.3 version.
--
Best regards,
Andreymailto:[EMAIL PROTECTED]
Re: how do I store/retrieve a hash through Apache::Session?
Once you realize that %session entries have to be scalars (references) as
described by other responses, be sure when you change entries in the
sub-hash that you tell the top-level session that the data's been changed
(it won't realize it unless you change one of its scalars).
Check the docs or the archives for the "official way" to do this (a message
from Jeffrey Baker uses such a phrase), but any modification of a %session
scalar will do - $session{_changed}++; , f'rinstance.
Best,
Randy
On Fri, Nov 03, 2000 at 04:36:15PM +, Enrique I . Rodriguez wrote:
Maybe a silly question...
how do I store/retrieve a hash through Apache::Session?
This doesn't work...
# retrieve
%table=$session{table};
...
# store
$session{table}=%table;
What's my fault?
Re: VirtualSellers.com TAME patent
IANAL, but I understand that technologies that were in the public domain can not be patented. And if a patent is issued, it's not enforcable in that case. Again, IANAL, but I feel that FreeType falls in that same category (see http://www.freetype.org). Randy On Mon, Jan 10, 2000 at 11:43:47AM -0800, Cliff Rayman wrote: looks like they submitted a patent for an embedded mark-up language. the language can actually be embedded inside an html page (what a concept!). see virtualsellers.com and the TAME language section. patent filing info was on business wire. soapbox These patents scare me since the patent granters seem not to understand software very well - and it seems like it is diametrically opposed to what the open software movement stands for. /soapbox cliff rayman genwax.com
Re: mysql.pm on Apache/mod_perl/perl win98
[cc list trimmed; this is more a mod_perl issue than a mysql one.] Thanks a bunch for your efforts here. I understand that earlier such bundles were built with VC6, eh? Will you have the opportunity/ability/tuits to distribute one built with Cygwin, so that the resultant installation can freely install new/updated modules from CPAN, including those that need a C compiler? I wouldn't suggest to bundle Cygwin, but to work with its standard install location, at any rate. The reason I ask is that I'd like to have a distribution available that includes HTML::Mason, and which is open for folks to add the precise modules that they need. Combined with SSH and CVS, this becomes a pretty decent development environment even for a web developer with only a Windows box. Perhaps it's more reasonable to request that you add HTML::Mason into the existing bundle, but that will have its own share of difficulties, considering the other unspecified modules that people would still be without. A cygwin-based distribution perhaps could simply ship with Perl's base install - let the user install their packages through CPAN. I've looked at getting apache/perl/mod_perl to build wtih cygwin, but I lacked the knowledge to understand the build hints for Perl, speak nothing of the rest. By the way, HTML::Mason may still have a problem here or there on Windows, which we folks on the Mason list will be glad to help get ironed out, I believe. At least, I'd be willing to help. Randy On Mon, Jan 10, 2000 at 04:10:08PM -0600, Randy Kobes wrote: On Mon, 10 Jan 2000, Tim Tompkins wrote: There is a DBD::mysql available from ActiveState. You can install it with ppm. ppm install DBD-mysql Hi, This would normally work, but the perl binary being used most probably isn't ActiveState's, because as of yet mod_perl doesn't work with ActiveState's perl. So there would most likely be a binary incompatibitly using this perl binary with ActiveState's DBD::mysql. I'm currently testing out an apache/mod_perl/perl Win32 binary that will have DBD::mysql and DBD::msql support, as well as mod_ssl. As a new mod_perl version is in the works, I'll probably wait until that is released. best regards, Randy Kobes
Re: Caching with $r-no_cache(1)
On Sun, Jan 09, 2000 at 08:45:11PM +, G.W. Haywood wrote: On Fri, 7 Jan 2000, Randy Harmon wrote: Does anybody have experience detecting such a condition, perhaps through one of the client headers? I haven't had a chance to dump them - many hats. No idea - ditto. In any case, I could use some Javascript to package up the machine's current time and send it back to the server, for instance at the same point where I check whether a cookie was accepted. That'd indicate their "Javscript-OK"-ness too. I think I'm willing to assume that someone clueful enough to turn off Javascript is clueful enough to have the correct time. You might want to look at the excellent ntpd documentation which talks about things like network delays. I think your Javascript idea is as Fortunately, network lag only works in our favor when it comes to this technique. So it expires a few hundred milliseconds in the past instead of "now"... no biggie to me. good a solution as you're going to get until the Web Comes Of Age. Don't know what you're going to do when I visit with Lynx though... I'm going to hope that Lynx is smarter than Netscape on this point, and assume that you're clueful enough to have the correct time. Well, at least my clock _will_ be right, I run a level 3 timeserver. He... I'm with you. Oh, good, my assumption was right. :) Randy
Caching with $r-no_cache(1)
I notice that the Guide omits the mention of Netscape's ignorance of Expires: set to the same as Date: when it mentions $r-no_cache(1) performing that function. Currently, I'm experiencing the problem with Netscape 4.7, although I seem to recall the same problem in earlier releases, in the case where the target browser's clock is slow. Of course, the server-side workaround (since we can't just fix our visitor's clocks) is to set an Expires header that significantly in the past, for some definition of 'significant'. I'm going to assume that $r-no_cache(1) won't be kludged to fix Netscape's bug, although some would argue that it should be 'fixed' in mod_perl. As an update to /guide/correct_headers.html#2_1_3_Expires_and_Cache_Control, I'd suggest adding the following text at the end: [ ... works with the major browsers. ] However, Netscape clients with slow clocks may not honor the 'immediate' timeout, cacheing pages anyway. This can be corrected by explicitly setting an Expires header that's in the past. How far in the past depends on how lenient you wish to be with browsers with slow clocks. Something between 5 and 30 minutes seems reasonable to me, but discussion may demonstrate a different approach and/or timeframe. Thoughts? Randy
Additions to the Apache::Session docs
Unless someone wants to implement multi-level tied hashes (/arrays) to
correct the following problem, something like the following should
make it into the Apache::Session documentation. I hope it helps some
folks.
Randy
---
When you store a hashref in a tied Apache::Session hash, subsequent
accesses to that underlying sub-hash do not go through the
Apache::Session tied-hash interface, and Apache::Session will not
recognize that changes have been made (I suspect that similar direct
access to a listref will behave the same way).
The changes to Apache::Session to enable recognition of such accesses
are non-trivial, particularly when multi-level hashes are desired.
In order to force Apache::Session to recognize such a change, you may
wish to access a scalar entry, such as by:
$session{'fix_hashref_storage'}++;
Re: ApacheDBI vs DBI for TicketMaster
On Sun, Jan 02, 2000 at 01:48:58AM -0600, Ed Loehr wrote: My apache children are seg faulting due to some combination of DBI usage and the cookie-based authentication/authorization [...] child seg faults. If I comment out all DBI references in the Hm, are you connecting to your database prior to Apache's forking - i.e., in your startup.pl? That could cause trouble, especially on older versions of Apache and Apache::DBI. Any clues? Not much of one. Hope it helps. Randy
Re: mod_perl Programmers demand is going up...
On Fri, Dec 03, 1999 at 11:38:35PM -0700, Michael Dearman wrote: "G.W. Haywood" wrote: How about ``Eagle''? Exactly what I was just thinking. Apache Stronghold Apache Eagle Um, isn't that animal taken by O'Rielly? How about an oyster? Perhaps not as american-patriotic but just think how sexy it sounds. Randy Logo? Eagle floating over the feather? Or, the feather floating under the Eagle. An Eagle Feather.
Re: Redhat httpsd with mod_perl
On Wed, Oct 20, 1999 at 10:47:02PM -0700, Adi wrote: On Wed, 20 Oct 1999, Remi Fasol wrote: Does anyone know why the shared memory would decrease so dramatically? Perl code and data both live in the data segment. As it is used, any time it writes information into a new chunk of memory, the memory is copied-on-write, becoming not-shared. You may or may not benefit from exiting your Apache child when its shared memory size shrinks too much. I think a package of Stas Bekman's authorship may help you detect that occasion. It wraps the gtop library in perl, so you can monitor 'top' info from mod_perl. Your own tests will give you an indication of whether this saves memory and should also indicate whether there is any performance advantage one way or other. It'd be interesting to hear your story of the results. Randy
Re: Another IE5 complaint
On Sat, Oct 02, 1999 at 03:21:17PM -0400, Greg Stark wrote: "Joe Pearson" [EMAIL PROTECTED] writes: [] whenever a IE5 user visits a page in their "Favorites", IE5 also trys to GET favicon.ico from the same site. Therefor I have hundreds of "File does not exist:" errors in my log file. [...] Can the gimp save .ico files? How big and what colours etc does IE5 expect Unknown. these to be anyways? http://www.wdvl.com/Authoring/Design/Images/Favicon/index.html Randy
