PHP 5.0.0 Test roll

21 years ago by Derrell.Lipman@UnwiredUniverse.com — view source

unread

Andi Gutmans andi@zend.com writes:

I did a test roll of PHP 5.0.0 (without version change). Please download and
make sure there are no serious show stoppers. You can grab it at
http://snaps.php.net/~andi/

5.0.0 Test Roll is working fine for me, including the sqlite 2.8.14 upgrade.

Thanks!

Derrell

21 years ago by Benj Carson — view source

unread

The file_get_contents() bug still seems to be present in the test roll, but
not in the latest snap:

Reproduce code (from the bug report for 29099):

#!/usr/bin/php
<?php
$ch = curl_init("http://www.php.net/";);
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
$str1 = curl_exec($ch);
curl_close($ch);

$str2 = file_get_contents("http://www.php.net/";);

echo $str2; // Note garbage at end of string

if ($str1 != $str2) {
echo "Both results differ.\n";
} else {
echo "Both results are the same.\n";
}

The latest snap (200407130430) seems to be just fine though.

Benj Carson

21 years ago by Derick Rethans — view source

unread

I did a test roll of PHP 5.0.0 (without version change). Please download
and make sure there are no serious show stoppers. You can grab it at
http://snaps.php.net/~andi/

I moved those to http://www.php.net/~andi/ and renamed them from RC4
(which it is not of course) to -pre-release-test in order to avoid
confusion later on.

regards,
Derick

21 years ago by Edin Kadribasic — view source

unread

I did a test roll of PHP 5.0.0 (without version change). Please download
and make sure there are no serious show stoppers. You can grab it at
http://snaps.php.net/~andi/
Edin, it might be a good idea for you to roll a test package too.
If all goes well, I'll do version changes and will roll the release within
about 24 hours.

Windows build from the CVS as of 6:00 GMT:

http://www.php.net/~edink/php-5.0.0-Win32-prerelease-test.zip

Edin

21 years ago by Robert Janeczek — view source

unread

Windows build from the CVS as of 6:00 GMT:

http://www.php.net/~edink/php-5.0.0-Win32-prerelease-test.zip

http://bugs.php.net/bug.php?id=29025 is still present

rash

21 years ago by Derick Rethans — view source

unread

Windows build from the CVS as of 6:00 GMT:

http://www.php.net/~edink/php-5.0.0-Win32-prerelease-test.zip

http://bugs.php.net/bug.php?id=29025 is still present

It's set to "Open", isn't it?

Derick

21 years ago by Chris — view source

unread

Edin Kadribasic wrote:

Please download
and make sure there are no serious show stoppers.

Windows build from the CVS as of 6:00 GMT:

http://www.php.net/~edink/php-5.0.0-Win32-prerelease-test.zip

Edin

Hi, I'm new to the list, so forgive me if I'm going about this the wrong
way.

I'm still experiencing the bug: http://bugs.php.net/bug.php?id=28829
(Crashes when BC Math functions get an argument of 0) with this
release. I have no problems duplicating it and should be around for a
while if someone needs more information, or is unable to duplicate it.

Chris

P.S. I'm not really a C++ guy, I have Visual Studio 6 installed on this
machine, so I have access to its tools, but I'm not very good at using
them. I mostly joined this list to lurk, but I'm willing to help where
I'm able.

21 years ago by Daniel Crookston — view source

unread

I'm glad you spoke up first Chris, because I have questions in a similar
vein. Namely, how can I help? I have a few years of C under my belt, but
I never got to the point where I'd be comfy hacking PHP. (I'd like to get
there eventually, I'm just not there yet.)

So, am I being lazy, have I missed some docs on the web regarding how to
be useful, or how to become familiar enough with the PHP code to start
lending a hand with the debugging? Or are we expected to learn it
ourselves, on our own, in order to be accepted into "the club" as it were?
(Not that there's anything wrong with that, it's just that a lot of groups
require self-validation like that.)

Thanks for any help,
Dan

--
The dark language of memory allocation is not to be spoken here.

P.S. I'm not really a C++ guy, I have Visual Studio 6 installed on this
machine, so I have access to its tools, but I'm not very good at using them.
I mostly joined this list to lurk, but I'm willing to help where I'm able.

21 years ago by Adam Maccabee Trachtenberg — view source

unread

So, am I being lazy, have I missed some docs on the web regarding how to
be useful, or how to become familiar enough with the PHP code to start
lending a hand with the debugging? Or are we expected to learn it
ourselves, on our own, in order to be accepted into "the club" as it were?
(Not that there's anything wrong with that, it's just that a lot of groups
require self-validation like that.)

The general method of helping out (at first) is to review open bug
reports and post patches to this list. Someone will then review and
merge into the code base.

A pattern of good patches leads to greater CVS karma, so you can apply
them yourself.

If you're not up for that right now, you could go through open reports
and help classify them, provide reproducing test cases (and/or
backtraces), mark them as bogus, etc.

There's no good online "here's how PHP's C internals work," but
George's "Advanced PHP Programming" book is a good place to start
reading.

-adam

--
adam@trachtenberg.com
author of o'reilly's "upgrading to php 5" and "php cookbook"
avoid the holiday rush, buy your copies today!

21 years ago by Robert Janeczek — view source

unread

There's no good online "here's how PHP's C internals work," but
George's "Advanced PHP Programming" book is a good place to start
reading.

nice. just asked someone to publish it in poland :)

rash

21 years ago by Robert Janeczek — view source

unread

I'm still experiencing the bug: http://bugs.php.net/bug.php?id=28829
(Crashes when BC Math functions get an argument of 0) with this
release. I have no problems duplicating it and should be around for a
while if someone needs more information, or is unable to duplicate
it.

yep, this bug is still present. i would like to say, that this bug is real
pain in the ass for windows users, because this extension cannot be turned
off and it causes some pear packages to be unuseble (i know of Benchmark,
probably everything that uses it is crushing too). please fix this guys :/

rash

21 years ago by Sara Golemon — view source

unread

"Robert Janeczek" rashid@ds.pg.gda.pl wrote in message
news:20040713224220.57700.qmail@pb1.pair.com...

I'm still experiencing the bug: http://bugs.php.net/bug.php?id=28829
(Crashes when BC Math functions get an argument of 0) with this
release. I have no problems duplicating it and should be around for a
while if someone needs more information, or is unable to duplicate
it.

yep, this bug is still present. i would like to say, that this bug is real
pain in the ass for windows users, because this extension cannot be turned
off and it causes some pear packages to be unuseble (i know of Benchmark,
probably everything that uses it is crushing too). please fix this guys :/

Just commited what should be a fix. Please try next snapshot.

-Sara

21 years ago by Chris — view source

unread

Sara Golemon wrote:

"Robert Janeczek" rashid@ds.pg.gda.pl wrote in message
news:20040713224220.57700.qmail@pb1.pair.com...

I'm still experiencing the bug: http://bugs.php.net/bug.php?id=28829
(Crashes when BC Math functions get an argument of 0) with this
release. I have no problems duplicating it and should be around for a
while if someone needs more information, or is unable to duplicate
it.

yep, this bug is still present. i would like to say, that this bug is real
pain in the ass for windows users, because this extension cannot be turned
off and it causes some pear packages to be unuseble (i know of Benchmark,
probably everything that uses it is crushing too). please fix this guys :/

Just commited what should be a fix. Please try next snapshot.

-Sara

All my testing code runs fine, so it appears to be working.

Thanks!

Chris

21 years ago by Robert Janeczek — view source

unread

Just commited what should be a fix. Please try next snapshot.

looks fine - thanks :)

rash

21 years ago by Sebastian Bergmann — view source

unread

Andi Gutmans wrote:

If all goes well, I'll do version changes and will roll the release
within about 24 hours.

Current HEAD still segfaults with "php -m" on Windows:

http://bugs.php.net/bug.php?id=29088

--
Sebastian Bergmann
http://sebastian-bergmann.de/ http://phpOpenTracker.de/

Das Buch zu PHP 5: http://professionelle-softwareentwicklung-mit-php5.de/

21 years ago by Stefan Esser — view source

unread

Hi,

Current HEAD still segfaults with "php -m" on Windows:

this is not reproduceable with latest snapshotbuild on Windows XP.

Stefan

21 years ago by Nuno Lopes — view source

unread

Current HEAD still segfaults with "php -m" on Windows:

this is not reproduceable with latest snapshotbuild on Windows XP.

Stefan

Using php5-win32-200407131030.zip, PHP still segfaults on my pc:

Program received signal SIGSEGV, Segmentation fault.
0x1003e2fd in zend_hash_destroy () from /cygdrive/c/php/php5ts.dll
(gdb) bt
#0 0x1003e2fd in zend_hash_destroy () from /cygdrive/c/php/php5ts.dll
#1 0x0012ff4c in ?? ()
#2 0x00362678 in ?? ()
#3 0x1007145a in php_stream_copy_to_stream () from
/cygdrive/c/php/php5ts.dll
#4 0xbaadf00d in ?? ()
#5 0x00362678 in ?? ()
#6 0x10338598 in php_info_print_table_row ()
#7 0x10061ead in php_log_err () from /cygdrive/c/php/php5ts.dll

Nuno

21 years ago by Andi Gutmans — view source

unread

Are you sure you're not loading old extensions?

At 12:39 PM 7/13/2004 +0100, Nuno Lopes wrote:

Current HEAD still segfaults with "php -m" on Windows:

this is not reproduceable with latest snapshotbuild on Windows XP.

Stefan

Using php5-win32-200407131030.zip, PHP still segfaults on my pc:

Program received signal SIGSEGV, Segmentation fault.
0x1003e2fd in zend_hash_destroy () from /cygdrive/c/php/php5ts.dll
(gdb) bt
#0 0x1003e2fd in zend_hash_destroy () from /cygdrive/c/php/php5ts.dll
#1 0x0012ff4c in ?? ()
#2 0x00362678 in ?? ()
#3 0x1007145a in php_stream_copy_to_stream () from
/cygdrive/c/php/php5ts.dll
#4 0xbaadf00d in ?? ()
#5 0x00362678 in ?? ()
#6 0x10338598 in php_info_print_table_row ()
#7 0x10061ead in php_log_err () from /cygdrive/c/php/php5ts.dll

Nuno

21 years ago by Nuno Lopes — view source

unread

Are you sure you're not loading old extensions?

Completly sure! I've also tried to disable all extensions and it still
segfaults.

At 12:39 PM 7/13/2004 +0100, Nuno Lopes wrote:

Current HEAD still segfaults with "php -m" on Windows:

this is not reproduceable with latest snapshotbuild on Windows XP.

Stefan

Using php5-win32-200407131030.zip, PHP still segfaults on my pc:

Program received signal SIGSEGV, Segmentation fault.
0x1003e2fd in zend_hash_destroy () from /cygdrive/c/php/php5ts.dll
(gdb) bt
#0 0x1003e2fd in zend_hash_destroy () from /cygdrive/c/php/php5ts.dll
#1 0x0012ff4c in ?? ()
#2 0x00362678 in ?? ()
#3 0x1007145a in php_stream_copy_to_stream () from
/cygdrive/c/php/php5ts.dll
#4 0xbaadf00d in ?? ()
#5 0x00362678 in ?? ()
#6 0x10338598 in php_info_print_table_row ()
#7 0x10061ead in php_log_err () from /cygdrive/c/php/php5ts.dll

Nuno

21 years ago by Andi Gutmans — view source

unread

Can you try and debug? I just can't get it to crash here :)

At 05:13 PM 7/13/2004 +0100, Nuno Lopes wrote:

Are you sure you're not loading old extensions?

Completly sure! I've also tried to disable all extensions and it still
segfaults.

At 12:39 PM 7/13/2004 +0100, Nuno Lopes wrote:

Current HEAD still segfaults with "php -m" on Windows:

this is not reproduceable with latest snapshotbuild on Windows XP.

Stefan

Using php5-win32-200407131030.zip, PHP still segfaults on my pc:

Program received signal SIGSEGV, Segmentation fault.
0x1003e2fd in zend_hash_destroy () from /cygdrive/c/php/php5ts.dll
(gdb) bt
#0 0x1003e2fd in zend_hash_destroy () from /cygdrive/c/php/php5ts.dll
#1 0x0012ff4c in ?? ()
#2 0x00362678 in ?? ()
#3 0x1007145a in php_stream_copy_to_stream () from
/cygdrive/c/php/php5ts.dll
#4 0xbaadf00d in ?? ()
#5 0x00362678 in ?? ()
#6 0x10338598 in php_info_print_table_row ()
#7 0x10061ead in php_log_err () from /cygdrive/c/php/php5ts.dll

Nuno

21 years ago by Antony Dovgal — view source

unread

Hi all!

That's what valgrind says:

Zend/tests/bug26166.phpt

Zend/zend_execute.c:1352
EX(Ts) = (temp_variable *) safe_emalloc(sizeof(temp_variable),
op_array->T, 0);

==8511== 40 bytes in 1 blocks are still reachable in loss record 2 of 7
==8511== at 0x3C01E375: malloc (vg_replace_malloc.c:105)
==8511== by 0x8222A55: _emalloc (zend_alloc.c:182)
==8511== by 0x8222C77: _safe_emalloc (zend_alloc.c:238)
==8511== by 0x8266E78: execute (zend_execute.c:1352)

Zend/tests/bug20240.phpt

==8478== 40 bytes in 1 blocks are still reachable in loss record 2 of 7
==8478== at 0x3C01E375: malloc (vg_replace_malloc.c:105)
==8478== by 0x8222A55: _emalloc (zend_alloc.c:182)
==8478== by 0x8231488: call_user_function (zend_execute_API.c:517)
==8478== by 0x816D03B: user_shutdown_function_call
(basic_functions.c:2088)

Zend/tests/bug22836.phpt:

Zend/zend_execute.c(609) : Freeing 0x3D172090 (4 bytes),
script=php-src/Zend/tests/bug22836.php
php-src/Zend/zend_variables.c(137) : Actual location (location was
relayed)
php-src/Zend/zend_execute.c(264) : Freeing 0x3D1726F4 (16 bytes),
script=php-src/Zend/tests/bug22836.php

There are some more issues, I'm still waiting for valgrind to
finish its work (it's sloooow)..

WBR,
Antony Dovgal aka tony2001
tony2001@phpclub.net || antony@dovgal.com

21 years ago by Antony Dovgal — view source

unread

Some more:

php-src/ext/soap/tests/server014.phpt:

==10122== Invalid read of size 4
==10122== at 0x8230D88: _zval_ptr_dtor (zend_execute_API.c:389)
==10122== by 0x823D45E: _zval_ptr_dtor_wrapper (zend_variables.c:193)
==10122== by 0x8247C41: zend_hash_destroy (zend_hash.c:519)
==10122== by 0x8255F67: zend_objects_free_object_storage
(zend_objects.c:88)==10122== Address 0x4FFFB588 is not stack'd,
malloc'd or free'd==10122==
==10122== Invalid read of size 4
==10122== at 0x8230D90: _zval_ptr_dtor (zend_execute_API.c:390)
==10122== by 0x823D45E: _zval_ptr_dtor_wrapper (zend_variables.c:193)
==10122== by 0x8247C41: zend_hash_destroy (zend_hash.c:519)
==10122== by 0x8255F67: zend_objects_free_object_storage
(zend_objects.c:88)==10122== Address 0x4FFFB588 is not stack'd,
malloc'd or free'd==10122==
==10122== Invalid read of size 4
==10122== at 0x8230DE9: _zval_ptr_dtor (zend_execute_API.c:393)
==10122== by 0x823D45E: _zval_ptr_dtor_wrapper (zend_variables.c:193)
==10122== by 0x8247C41: zend_hash_destroy (zend_hash.c:519)
==10122== by 0x8255F67: zend_objects_free_object_storage
(zend_objects.c:88)==10122== Address 0x4FFFB588 is not stack'd,
malloc'd or free'd==10122==
==10122== Invalid write of size 1
==10122== at 0x8230DF4: _zval_ptr_dtor (zend_execute_API.c:394)
==10122== by 0x823D45E: _zval_ptr_dtor_wrapper (zend_variables.c:193)
==10122== by 0x8247C41: zend_hash_destroy (zend_hash.c:519)
==10122== by 0x8255F67: zend_objects_free_object_storage
(zend_objects.c:88)==10122== Address 0x4FFFB58D is not stack'd,
malloc'd or free'd

WBR,
Antony Dovgal aka tony2001
tony2001@phpclub.net || antony@dovgal.com

21 years ago by Antony Dovgal — view source

unread

On Tue, 13 Jul 2004 15:34:28 +0400
Antony Dovgal tony2001@phpclub.net wrote:

Hi all!

That's what valgrind says:

Zend/tests/bug26166.phpt

Zend/zend_execute.c:1352
EX(Ts) = (temp_variable *) safe_emalloc(sizeof(temp_variable),
op_array->T, 0);

==8511== 40 bytes in 1 blocks are still reachable in loss record 2 of
7==8511== at 0x3C01E375: malloc (vg_replace_malloc.c:105)
==8511== by 0x8222A55: _emalloc (zend_alloc.c:182)
==8511== by 0x8222C77: _safe_emalloc (zend_alloc.c:238)
==8511== by 0x8266E78: execute (zend_execute.c:1352)

Zend/tests/bug20240.phpt

==8478== 40 bytes in 1 blocks are still reachable in loss record 2 of
7==8478== at 0x3C01E375: malloc (vg_replace_malloc.c:105)
==8478== by 0x8222A55: _emalloc (zend_alloc.c:182)
==8478== by 0x8231488: call_user_function (zend_execute_API.c:517)
==8478== by 0x816D03B: user_shutdown_function_call
(basic_functions.c:2088)

Dunno if it's the right thing to add if()'s around emalloc(), but with
attached patch these warnings disappear, reducing valgrind logs in a
half.

WBR,
Antony Dovgal aka tony2001
tony2001@phpclub.net || antony@dovgal.com

21 years ago by Andi Gutmans — view source

unread

Okay. Tag wise it's RC4 because I couldn't come up with something better
and didn't want to screw up the CVS tag names.

At 10:00 AM 7/13/2004 +0200, Derick Rethans wrote:

I did a test roll of PHP 5.0.0 (without version change). Please download
and make sure there are no serious show stoppers. You can grab it at
http://snaps.php.net/~andi/

I moved those to http://www.php.net/~andi/ and renamed them from RC4
(which it is not of course) to -pre-release-test in order to avoid
confusion later on.

regards,
Derick

21 years ago by Andi Gutmans — view source

unread

Can someone reproduce this? I couldn't. Not in debug nor in release.

At 12:59 PM 7/13/2004 +0200, Sebastian Bergmann wrote:

Andi Gutmans wrote:

If all goes well, I'll do version changes and will roll the release
within about 24 hours.

Current HEAD still segfaults with "php -m" on Windows:

http://bugs.php.net/bug.php?id=29088

--
Sebastian Bergmann
http://sebastian-bergmann.de/ http://phpOpenTracker.de/

Das Buch zu PHP 5: http://professionelle-softwareentwicklung-mit-php5.de/

21 years ago by romain bourdon — view source

unread

I tried pre-release-test package on "php -m" : Does not segfault at home
with standard conf.
So I tried activating all extensions. I had a lot of dependencies Warnings
(normal) but no segfault...

Can someone reproduce this? I couldn't. Not in debug nor in release.

At 12:59 PM 7/13/2004 +0200, Sebastian Bergmann wrote:

Andi Gutmans wrote:

If all goes well, I'll do version changes and will roll the release
within about 24 hours.

Current HEAD still segfaults with "php -m" on Windows:

http://bugs.php.net/bug.php?id=29088

--
Sebastian Bergmann
http://sebastian-bergmann.de/ http://phpOpenTracker.de/

Das Buch zu PHP 5: http://professionelle-softwareentwicklung-mit-php5.de/

21 years ago by Rob Richards — view source

unread

I can repeatedly reproduce this with cli (php -m) and apache2 (just starting
and stopping it) apache 2 crash probably same as bug 29025 - though with or
without xsl doesnt matter). cgi seems to be fine. Always same stack trace.

php5ts_debug.dll!_zend_is_inconsistent(_hashtable * ht=0xcdcdcdcd, char *
file=0x1053aaf0, int line=510) Line 53 + 0x3 C
php5ts_debug.dll!zend_hash_destroy(_hashtable * ht=0xcdcdcdcd) Line 510 +
0x19 C
php5ts_debug.dll!php_shutdown_stream_wrappers(int module_number=0, void * *

tsrm_ls=0x00954278) Line 1379 + 0x12 C
php5ts_debug.dll!php_module_shutdown(void * * * tsrm_ls=0x00954278) Line
1516 + 0xd C
php5ts_debug.dll!php_module_shutdown(void * * * tsrm_ls=0x00954278) Line
1516 + 0xd C
php-cli.exe!main() Line 1052 + 0xd C
php-cli.exe!mainCRTStartup() Line 338 + 0x11 C

Rob

----- Original Message -----
From: Andi Gutmans

Can someone reproduce this? I couldn't. Not in debug nor in release.

21 years ago by Derrell.Lipman@UnwiredUniverse.com — view source

unread

"Rob Richards" rrichards@ctindustries.net writes:

php5ts_debug.dll!_zend_is_inconsistent(_hashtable * ht=0xcdcdcdcd, char *
file=0x1053aaf0, int line=510) Line 53 + 0x3 C
php5ts_debug.dll!zend_hash_destroy(_hashtable * ht=0xcdcdcdcd) Line 510 +
0x19 C
php5ts_debug.dll!php_shutdown_stream_wrappers(int module_number=0, void * *

tsrm_ls=0x00954278) Line 1379 + 0x12 C
php5ts_debug.dll!php_module_shutdown(void * * * tsrm_ls=0x00954278) Line
1516 + 0xd C
php5ts_debug.dll!php_module_shutdown(void * * * tsrm_ls=0x00954278) Line
1516 + 0xd C
php-cli.exe!main() Line 1052 + 0xd C
php-cli.exe!mainCRTStartup() Line 338 + 0x11 C

I don't know the internals of PHP, but a hash table parameter of 0xcdcdcdcd
looks an awful lot to me like there's a debug malloc library of some sort that
fills freed memory with '\xcd', meaning that zend_hash_destroy() is being
called with previously freed memory.

Since php_shutdown_stream_wrappers() calls zend_hash_destroy() three times,
but the first and third call (directly or indirectly) pass a parameter which
is simply the address of a variable, my first guess would be that this call:

zend_hash_destroy(php_get_stream_filters_hash())

is the culprit, which implicates the variable 'stream_filters" in some way,
shape or form (via the function _php_get_stream_filters_hash() in
filters.c:41.

I hope maybe this helps someone who knows the code determine the problem. I
did a cursory search for something pertaining to 0xcd and could not find it,
but maybe someone more familiar with the code will recognize 0xcdcdcdcd...

Cheers,

Derrell

21 years ago by Nuno Lopes — view source

unread

Can someone reproduce this? I couldn't. Not in debug nor in release.

I can!
http://news.php.net/php.internals/11163

At 12:59 PM 7/13/2004 +0200, Sebastian Bergmann wrote:

Andi Gutmans wrote:

If all goes well, I'll do version changes and will roll the release
within about 24 hours.

Current HEAD still segfaults with "php -m" on Windows:

http://bugs.php.net/bug.php?id=29088