fix inode/device usage for win32, take #1

--00c09ffb4d2ee7d4e80476257ee7 Content-Type: text/plain; charset=ISO-8859-1 hi, This is a first attempt (draft) to fix the pecl bug #16860 (http://pecl.php.net/bugs/bug.php?id=16860). Here are the changes I made so far: - add apc_ino_t and apc_dev_t - native ino_t/dev_t on non windows platform - unsigned int64 on windows - use string for the inode and device key (cache info and iterator) - device uses now the volume serial number (part of the file information) - inode uses now the file id (part of the file information) - add zend_file_handle support to apc_cache_make_

apc(4), php4 support?

Hi All! Do we really support php4 for APC? Reading the recent commits in APC4 I was surprised to read php4 fixes :) Cheers, -- Pierre http://blog.thepimp.net | http://www.libgd.org

Time for a Meeting?

Hi, I've been out of the whole conference thing for a year and I haven't talked to enough people about apc & what it needs, what works and what doesn't in a while. So how about a meeting on irc? I'm OK with a US afternoon slot, so how does 2 PM PST on thursday sound? The code in SVN pretty much works to most of the requirements I have in mind, but I'd like to steal as much of apc-4.0 features into the 3.1.x branch without breaking api BC. Rest of the agenda is wide open :) Cheers, Gopal -- If I learn from my mistakes, pretty soon I'll know everything.

remove apc.include_once_override ?

I like the original intent and have found this useful in the past but currently I feel like there's a lot of issues with the apc.include_once_override option, and it's not always obvious what these are. Considering that there's similar optimizations in later versions of PHP, and that I don't see any good way to fix the concurrency issues with this optimization, I vote that we just remove it or at least disable it and make it a non-obvious option? Any objections or interest in salvaging it? http://pecl.php.net/bugs/bug.php?id=15356 http://pecl.php.net/bugs/bug.php?id=16086 -shire

pcre dependencies on php-5.2.2+

It looks like pcre header files were not being installed for us prior to php-5.2.2 when compiled against the non-bundled library. http://pecl.php.net/bugs/bug.php?id=15968&thanks=1 Do we want to try and hack-up the code further to support php-5.2.2 or should we just make a hard dependency on this? Alternatively we could add an option to disable pcre support in APC so it would compile, or they can recomplie older versions with the bundled pcre libraries to get APC working. I'm not a big fan of trying to get the broken installs to work with APC, as it'll just make things more compli

atomaticity/thread safeness of apc shared memory functions + mutexes

Couldn't find it in the docs, so I hope you know it: Are the userland apc cache functions (apc_add, apc_fetch, apc_delete etc) *thread-safe/atomic*? Also: I need to ensure that conncurrent access to shared memory doesn't get me in race conditions. Wouldn't it be nice if apc could offer mutex functions and the like? There is a semaphore extension for php, but it is *nix-only according to the docs. The need for synchronization mechanisms arise almost allways when using shared memory.

Feature request: ability to filter out entire directories/folders

I would like to suggest a new feature: ability to exclude php files from entire folders. This is useful when we have limited RAM on the host machine and would then like to exclude the admin panels for the CMS or blogging software installed on the system. At the moment the filtering parameter only looks at the file name and ignores the folder or the full path - hence there is no real way to achieve this setup. I hope this is posted in the right place, otherwise let me know where and I appologize. Thank you! -Ron Peled

Politics

silliness fido offering flying lewiss twotothen weenie floating threetimeslonger

Politics

cancellation quarter tucked termed incorporated frobnitz partisans diablo nuke

Politics

mahnjed weapons diverging pufferfish wokked protections semantics circling supposed

APC Wiki

I've taken the liberty of creating an APC wiki like we discussed. I've added all the stuff I could think of that's on my plate. I've forgotten the date Gopal tossed out for the next release so we'll probably want to tag that on along with everyone else's todo's. http://wiki.php.net/pecl/apc Let me know if this should be in another location/URL. -shire

RFC: APCIterator class

Here is a rough patch of what I have currently in my working copy for an APCIterator class. Check out the tests for some examples but it looks something like: <?php $it = new APCIterator('user', $regex, $chunk_size); foreach($it as $key=>$value) { echo $key."\n"; } ?> Some code probably needs to be cleaned up and I need to add a delete () method along with the ability to only dump entry values rather than the full entry info (which it does by default currently). Locking names also conflict with sma.c which I'll need to resolve. The primary purpose here is to all

APC for PHP 5.3 on Windows

--------------060202040005060005040503 Content-Type: text/plain; charset=ISO-8859-2; format=flowed Content-Transfer-Encoding: 7bit Hi, the attached patch should repair the compilation of APC for PHP 5.3 on Windows. Regards Jaroslav Hanslik --------------060202040005060005040503 Content-Type: text/plain; name="apc.diff" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="apc.diff" --- apc.c 2008-05-13 17:51:46.000000000 +0200 +++ apc-patched.c 2008-06-09 20:17:30.375000000 +0200 @@ -32,7 +32,12 @@ /* $Id: apc.c,v 3.22 2008/05/13 15:51:46 gopalv

apc_delete_file() ?

I'd like to add a function that removes a specific compiled file from the cache. My initial thought is to have the function, given a filename: - use apc_cache_make_file_key() to determine the cache key for the filename - use apc_cache_find_slot() to determine the slot for the cache key - call remove_slot() on the slot to either free it immediately or put it on the deleted list if its refcount is > 0 - and optionally have the function call process_pending_removals() to flush the deleted list (if possible) Another possibility, more useful for bulk deletes (at least on fullpath entr

3.0.x vs. 3.1.x

It looks like we have a regression: http://pecl.php.net/bugs/bug.php?id=13505 The largest change from 3.0.16 was the expunge callback stuff. Gopal has new code for 3.1.0 that will make cache expunges much easier and this code will need to change completely. And since we don't actually use it in 3.0.x, I think we should just pull it from the 3.0 tree and see if that fixes the MediaWiki issue. We should also fix: http://pecl.php.net/bugs/bug.php?id=13504 Probably means we need a 3.0.18 sometime soon. -Rasmus

Support for multiple memory segments.

I'm working on multiple memory cache support which has been on my todo for a long time now (used for different cache behavior, testing, lock performance, smaller coredumps with meaningful information, and larger user caches). Do we still need to support the multiple segments to work around system memory limitations with traditional shared memory segments? I believe everyone can use mmap now to get around this, but perhaps there's some system out there that I'm not aware of that requires this?? If we don't need it I'll drop support in my patch as multiple-within- mu

APC b0rkage upcoming

Hi, I'm going to take a week or two slap up apc-3.1. My rewrite is in an un-mergeable mess, but I have a good idea on how to do it again quickly. The first bit of code is already pushed in http://cvs.php.net/viewvc.cgi/pecl/apc/apc_pool.c?view=log Feel free to review it/fix bugs if you find any. I'll be pushing it into sma land and on top of apc_compile.c by friday. Cheers, Gopal -- Yea from the table of my memory I'll wipe away all trivial fond records. -- Hamlet

[PATCH] APC signals + coredump shared memory unmapping

--B_3281224808_9444842 Content-type: text/plain; charset="US-ASCII" Content-transfer-encoding: 7bit Hi, I would like to propose the attached patch against APC CVS HEAD. It provides the configurable ability to trap coredump related signals and unmap the APC shared memory segment while still calling any previously installed signal handlers. This feature would be enabled via a new apc ini entry, apc.coredump_unmap, which is defaulted to disabled. Although this will technically work in production environments the functionality was engineered to solve a problem in shared developme

apc.stat_ctime

Is it possible to add apc.stat_ctime to the php.net documentation? I just kicked myself looking through the source code to make a similar patch and found it was already there but undocumented. -- Andrew Hutchings - LinuxJedi - http://www.linuxjedi.co.uk/ Windows is the path to the darkside...Windows leads to Blue Screen. Blue Screen leads to downtime. Downtime leads to suffering...I sense much Windows in you...

Spurious writes to /tmp

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello, I'm in the process of setting up a CentOS 5 installation, which ships with a tightened SELinux policy. APC is installed and working well, however I notice that on any access to a PHP script - or, confusingly, any CGI script also - these warnings are generated: avc: denied { read, write } for comm="test.sh" dev=dm-0 egid=48 euid=48 exe="/bin/bash" exit=0 fsgid=48 fsuid=48 gid=48 items=0 name=".apc.ZEAkwD" Elaborated by sealert: Source Context user_u:system_r:httpd_sys_script_t Target Context

APC showing wrong pages

------=_Part_12570_21496387.1174225315035 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Content-Disposition: inline Hello, I noticed APC returning the cached values of the wrong pages. I'd bring up x.php in my browser and see (apparently cached) content from y.php, then reload again and see the (correct) content from x.php. Is this a known problem? Is the current version of APC recommended for prod environments? Thanks, --Craig Version info: - APC 3.0.13 - Apache/2.0.55 (Ubuntu) - PHP/5.1.2 ------=_Part_12570_21496387.

3.0.13 Released

I just pushed out 3.0.13 so we can start breaking things again. There are a bunch of internal fixes from Gopal and Brian Shire in this release and a couple of user-visible features as well. The user-visible stuff being: Pthread mutex and spin locks Brian Shire did all the legwork on this. We stole (with permission) the spin locking code from PostgreSQL after I had a chat with Bruce Momjian about PostgreSQL internals. It turns out PostgreSQL has a very similar shared memory locking requirement and they had already done all the work of writing a cross-platform spin lock implemen