Re: [PHP-DEV] buildbot
Sebastian Bergmann wrote: I just read about buildbot [1] on the Mono mailinglist. Would this (or a similar solution) not be helpfull? -- [1] http://buildbot.sourceforge.net/ a) easy to set up? esp. if compared to mozilla tinderbox? ;) (http://www.mozilla.org/tinderbox.html) b) who is going to provide the builders? c) how does this approach scale with the multitude of different extension setups in addition with lots of external libraries relied upon that come in different versions? -- Hartmut Holzgraefe [EMAIL PROTECTED] -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
RE: [PHP-DEV] PDO (was: SQLite API deficiency)
Hi, I apologize for not going to the end of the thread to see if you posted a link online somewhere to this, but I would love a copy none the less. It wasn't posted yet, but you can find it here: http://www.guidance.nl/php/dbx/dbx.p Cheerio, Marc. -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
[PHP-DEV] overloading class properties
Hello! Trying to use ZE2's new features for my framework I fell in trouble with (probably) abnormal behaviour of overloaded class properties which has been declared as expected long time ago. First of all, I do admit this feature has been discussed before, but despite of all that I apologize for bringing this issue back and I'd like you to pay a bit of your attention to the subject. The problem is following: ?php class A { public $pub; private $pri; function __get($name) { echo __get($name)\n; return $this; } } $a = new A; echo $a-pub; // no error, but direct access. no __get() call is being made echo $a-pri; // results in Fatal error. no __get() call is being made ? In neither of two echo'es the call to __get() is being made. Short question: Why is that? Long question: Isn't that clear that when there is a __get() declared in a class, this 'magic' method should be called on ANY access to a class property? Those who want it to behave like it does at the moment, may always emulate desired functionality with Reflection API: function __get($name) { $class = new ReflectionClass(__CLASS__); if($class-getProperty($name)-isPrivate()) trigger_error( ... ); return $this-$name; } ...but, please, give some freedom to those who want it! Let __get()/__set() be called on EVERY attempt to access a property, because the current realization makes Overloading practically USELESS for any serious use, since you don't have a 100% control over accessing to class' properties. Thanks in advance for your time and effort to helping us out with this one. Maxx -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
[PHP-DEV] New PDFlib extension for PHP5
Hello, I'm maintaining the PDFlib wrapper code for the pdf extenions of PHP. I did a major rewrite of the wrapper for PDFlib 6 and PHP 5 and wanted to ask whether it is good idea to check in this code now (short before the release of PHP 5). My idea is to add this new Wrappercode as an extra module and keep the existing wrapper code for peoples using the old PDFlib 4 version. This is necessary because there are incompatible changes between PDFlib 4 and 5 which made a new wrapper module necessary. Here the details about the new wrapper code: This PDFlib Wrapper Code works with PDFlib 5 and higher. A new module was introduced, as PDFlib 5 supports a method to handle the special returnvalues needed for PHP in the PDFlib core now. This is necessary, as the new concept of optionlists in PDFlib allows to pass handles to PDFlib objects and filenames inside of optionlists. Before the wrapper added some offset to handles returned by PDFlib to fit into scheme used by php to return 0 in case of errors, whereas in PDFlib the errorcode is -1 and 0 is a valid handle. As the optionlist may contain filenames on various places the VIRTUAL_DIR support and the CHECK_OPEN_BASEDIR checks implemented in the wrapper will not work reliable too. So VIRTUAL_DIR support and the CHECK_OPEN_BASEDIR checking is disabled here too, as it would only work for some of the files used with PDFlib. The main changes: - only official PDFlib API's are included (exception pdf_open_memory_image) - supports PDFlib 5 and PDFlib 6 - builds with older PHP Version too (tested with PHP 4.1.0, 4.2.1, 4.3.0, and 5.0.0RC2.) - adds an Object Oriented API to PDFlib (for PHP 5 only) - uses PHP exceptions (for PHP 5 only) - disables VIRTUAL_DIR support (use SerachPath instead) - disables CHECK_OPEN_BASEDIR checks - changed old API's with varargs to only accept all args now I think it is worth to include this now, so that PHP5 comes with this new code, but I wanted to discuss this first. -- Best regards, Rainer Schaaf [EMAIL PROTECTED]http://www.pdflib.com Personalize PDF: PDFlib Personalization Server and Block plugin ___PDFlib - a library for generating PDF on the fly -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
RE: [PHP-DEV] New PDFlib extension for PHP5
Hello Rainer, A few of us have been planning to move the pdflib extension into PECL (http://pecl.php.net) for a little while now; it sounds like an excellent time to make this move, because it will allow you to make supplemental releases outside of the PHP 5 core release cycle. Here is a possible release plan: 1/ Move the current code into PECL 2/ Make a release (so that people may choose to use the code they are used to) 3/ Merge your new code and bump the extension version to the next major (eg: from 1.x to 2.0) 4/ Make a release 5/ Update and make additional releases whenever you like How does that sound to you? - disables VIRTUAL_DIR support (use SerachPath instead) - disables CHECK_OPEN_BASEDIR checks This doesn't sound too hot for shared hosters. --Wez. -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
[PHP-DEV] [PATCH] File Upload - Binary and Garbage Safety
Hello PHP people! My patch adds binary safety to the variables sent when uploading files. I see this very often requested. It also restores some functionality which you broke in PHP 4.2. When fieldnames are not trimmed PHP does not work. Hope I was help Sammy Nolz -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
[PHP-DEV] [PATCH] [REPOST] File Upload - Binary and Garbage Safety
Sorry but your list does not like hushmail attachments Sammy Nolz --- rfc1867.c.unpatched 2002-01-01 02:02:11.0 +0100 +++ rfc1867.c 2002-01-01 02:04:55.0 +0100 @@ -230,6 +230,13 @@ } } +static void safe_php_register_binary_variable(char *var, char *strval, int strlen, zval *track_vars_array, zend_bool override_protection TSRMLS_DC) +{ + if (override_protection || !is_protected_variable(var TSRMLS_CC)) { + php_register_variable_safe(var, strval, strlen, track_vars_array TSRMLS_CC); + } +} + static void register_http_post_files_variable(char *strvar, char *val, zval *http_post_files, zend_bool override_protection TSRMLS_DC) { @@ -761,7 +768,7 @@ SAPI_API SAPI_POST_HANDLER_FUNC(rfc1867_post_handler) { - char *boundary, *s=NULL, *boundary_end = NULL, *start_arr=NULL, *array_index=NULL; + char *boundary, *s=NULL, *boundary_end = NULL, *start_arr=NULL, *end_arr=NULL, *array_index=NULL; char *temp_filename=NULL, *lbuf=NULL, *abuf=NULL; int boundary_len=0, total_bytes=0, cancel_upload=0, is_arr_upload=0, array_len=0; int max_file_size=0, skip_upload=0, anonindex=0, is_anonymous; @@ -883,23 +890,24 @@ /* Normal form variable, safe to read all data into memory */ if (!filename param) { - char *value = multipart_buffer_read_body(mbuff TSRMLS_CC); + int length = 0; + char *value = multipart_buffer_read_body(mbuff, length TSRMLS_CC); unsigned int new_val_len; /* Dummy variable */ if (!value) { value = estrdup(); } - if (sapi_module.input_filter(PARSE_POST, param, value, strlen(value), new_val_len TSRMLS_CC)) { + if (sapi_module.input_filter(PARSE_POST, param, value, length, new_val_len TSRMLS_CC)) { #if HAVE_MBSTRING !defined(COMPILE_DL_MBSTRING) if (php_mb_encoding_translation(TSRMLS_C)) { php_mb_gpc_stack_variable(param, value, val_list, len_list, num_vars, num_vars_max TSRMLS_CC); } else { - safe_php_register_variable(param, value, array_ptr, 0 TSRMLS_CC); + safe_php_register_binary_variable(param, value, new_val_len, array_ptr, 0 TSRMLS_CC); } #else - safe_php_register_variable(param, value, array_ptr, 0 TSRMLS_CC); + safe_php_register_binary_variable(param, value, new_val_len, array_ptr, 0 TSRMLS_CC); #endif } if (!strcasecmp(param, MAX_FILE_SIZE)) { @@ -1000,9 +1008,14 @@ /* is_arr_upload is true when name of file upload field * ends in [.*] -* start_arr is set to point to 1st [ +* start_arr is set to point to 1st [ and +* end_arr is set to point to last ] */ - is_arr_upload = (start_arr = strchr(param,'[')) (param[strlen(param)- 1] == ']'); + is_arr_upload = (end_arr = strrchr(param, ']')) (start_arr = strchr(param, '[')) (end_arr start_arr); + /* cut away garbage after ] */ + if (is_arr_upload end_arr) { + end_arr[1] = '\0'; + } /* handle unterminated [ */ if (!is_arr_upload start_arr) { *start_arr = '_'; @@ -1014,6 +1027,8 @@ efree(array_index); } array_index = estrndup(start_arr+1, array_len-2); + start_arr = NULL; + end_arr = NULL; } /* Add $foo_name */ Concerned about your privacy? Follow this link to get FREE encrypted email: https://www.hushmail.com/?l=2 Free, ultra-private instant messaging with Hush Messenger https://www.hushmail.com/services.php?subloc=messengerl=434 Promote security and make money with the Hushmail Affiliate Program: https://www.hushmail.com/about.php?subloc=affiliatel=427 -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
[PHP-DEV] Re: Destructors
Jason Garber wrote: Based on the above example, the $oSession-__destruct() method relies on the $oDB object still being usable. How can this be structured to ensure that the DB object does not get released first? Adding something like $oSession-db = $oDB; should work in my opinion. Didn't test it and I have no real experience with destructors. - Chris -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
Re: [PHP-DEV] Re: Destructors
On Mon, 24 May 2004, Christian Schneider wrote: Jason Garber wrote: Based on the above example, the $oSession-__destruct() method relies on the $oDB object still being usable. How can this be structured to ensure that the DB object does not get released first? It can't. Adding something like $oSession-db = $oDB; should work in my opinion. Didn't test it and I have no real experience with destructors. That's not guaranteed to work. PHP 5 does not guarantee that object destructors will be called in any particular order or that the order will remain constant from one invocation to another. Before asking for changes to this behavior, please check the mailing list archives for a long discussion on the topic. :) -adam -- [EMAIL PROTECTED] author of o'reilly's php cookbook avoid the holiday rush, buy your copy today! -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
Re: [PHP-DEV] Re: Destructors
Adam Maccabee Trachtenberg wrote: That's not guaranteed to work. PHP 5 does not guarantee that object destructors will be called in any particular order or that the order will remain constant from one invocation to another. Ok, I was under the impression that PHP won't call any destructors of objects still in use but from your comment I guess it does two passes: First collect a list of objects to destruct (which will be everything at the end of the script) and then call all the destructors (in random order). Ignore my first comment, it was posted without any insight in the internals of the GC mechanism :-) - Chris -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
Re: [PHP-DEV] Re: Destructors
Christian Schneider wrote: Adam Maccabee Trachtenberg wrote: That's not guaranteed to work. PHP 5 does not guarantee that object destructors will be called in any particular order or that the order will remain constant from one invocation to another. Ok, I was under the impression that PHP won't call any destructors of objects still in use but from your comment I guess it does two passes: First collect a list of objects to destruct (which will be everything at the end of the script) and then call all the destructors (in random order). unless you manually unset the objects in the order in which you want them to be destructed. regards, Lukas Smith -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
Re: [PHP-DEV] Re: Destructors
On Mon, 24 May 2004, Lukas Smith wrote: unless you manually unset the objects in the order in which you want them to be destructed. Right. That'll also work. :) -adam -- [EMAIL PROTECTED] author of o'reilly's php cookbook avoid the holiday rush, buy your copy today! -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
[PHP-DEV] where to find all the classes loaded in PHP
Hi All, I am working on porting APC to PHP 5.0. I want to know where to get all the use defined class entries(zend_class_entry) to copy to the APC cache. Old APC code was copying this information from CG(class_table) I tried to do the same but could not get much of a info from here. For example I have a class by name 'Simple' which do not have any parent class. The apc code checks for parent of a given class while copying The code looks like this, 'elem' is a pointer to 'Simple' zend_class_entry if(elem-parent) { //code to copy elem-parent-name to a newly allocated memory. } In my class 'Simple' I expect this if to fail as I assume parent to be null for a class that don't have any parent. But parent is not null which makes the if to succeed and hence accessing the name function of parent to fail with Segmentation fault. Thanks in advance With regards Kamesh Jayachandran -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
Re: [PHP-DEV] Re: Destructors
On Mon, 2004-05-24 at 10:40, Adam Maccabee Trachtenberg wrote: Right. That'll also work. :) I suppose you could push object references on a stack and register a shutdown function which pops / unset()s if you really wanted a order to it.. (that's an untested theory though) John -- -=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=- John Coggeshall http://www.coggeshall.org/ The PHP Developer's Handbookhttp://www.php-handbook.com/ -=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=- -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
Re: [PHP-DEV] Re: Destructors
John Coggeshall wrote: On Mon, 2004-05-24 at 10:40, Adam Maccabee Trachtenberg wrote: Right. That'll also work. :) I suppose you could push object references on a stack and register a shutdown function which pops / unset()s if you really wanted a order to it.. (that's an untested theory though) Which is what PEAR does to emulate destructors. But its doesnt really give you much control over the order of the stack. Its just the order in which you have created your instances. Anyways this is more and more turning into php-general@ material :-) regards, Lukas Smith -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
[PHP-DEV] fw: binarycluod-dev / namespaces
hi all, I wanted to forward this message along, and also ask some questions: -Are there any plans to support packages / modules properly? (ala python, which looks very nice, or I suppose java, etc). -Are there plans to restore support for namespaces? The current PEAR method of naming classes is the only sensible one given the engine's capabilities, but we (and I would imagine they too) find it to be extremely ugly and inelegant - something forced on us by the missing features in PHP5. I would like to see full hierarchical namespace support. -Has anyone thought of including a unitTest framework in pecl, so unit tests could be tightly integrated with the engine? --- Of course all this complaining is because we see the engine has become quite seriously capable when compared to PHP4 - it's a HUGE leap - but we need juuust a little more to really compete with the biggies. I think PHP, with version 5 - has the potential to be a major language - but it needs just a bit more of a gentle push to get there. I think those features would do it. I love python's system of modules: incredibly easy to use and explain, it works well, and it provides a clear method for structuring code. Something similar in PHP is welcome, and I trust the PEAR people to help set guidelines. Of course we would like to contribute to that process as well, if it happens. The message: yeah, welcome to the gradual let-down that has been the history of PHP5. But seriously ... one thing after another was pulled from the language since it was hard to get it working properly. loss of namespaces, magic __toString() was made unmagical, recently optional typehinted parameters were removed ... namespaces would have been a *huge* benefit to PHP. pobably almost as huge as exceptions. IMO the pear prefix solution is pretty dumb; if prefixing makes sense (and it often does), then I think having a unique name (like BC_ for Binarycloud) is a lot more reasonable than PHP_Framework_Binarycloud_Init, which is what PEAR does (makes the category part of the class names). I sure hope that namespaces gets reconsidered for 5.1 (which is supposed to follow 5.0 fairly quickly, as there are many features now that are being postponed due to feature freeze). things like lack of namespace support makes me look increasingly at Mono/ASP.NET for building big OO web applications. ouch _alex -- alex black founder, ceo the turing studio, inc. project leader binarycloud mail: [EMAIL PROTECTED] url : http://www.turingstudio.com tel : 510.666.0074 fax : 510.666.0093 -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php