When second confirmation writing happen error

[dreamsxin]

The first and second write same 100000 items, if second write 1 item will success and then change to 100000 will be success.

mdbx_env_set_mapsize(env, 256 * 1024 * 1024);
mdbx_env_set_maxdbs(env, 256);
rc = mdbx_env_open(env, "./example-db", 0, 0664);

bits.h:1122: NODEPTR: Assertion `NUMKEYS(p) > (unsigned)(i)' failed.

(gdb) bt
#0  0x00007ffff6428428 in __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:54
#1  0x00007ffff642a02a in __GI_abort () at abort.c:89
#2  0x00007ffff6420bd7 in __assert_fail_base (fmt=<optimized out>, assertion=assertion@entry=0x7fffec062fbe "NUMKEYS(p) > (unsigned)(i)", file=file@entry=0x7fffec062cee "./storage/libmdbx/bits.h", 
    line=line@entry=1122, function=function@entry=0x7fffec064170 <__PRETTY_FUNCTION__.32848> "NODEPTR") at assert.c:92
#3  0x00007ffff6420c82 in __GI___assert_fail (assertion=assertion@entry=0x7fffec062fbe "NUMKEYS(p) > (unsigned)(i)", file=file@entry=0x7fffec062cee "./storage/libmdbx/bits.h", line=line@entry=1122, 
    function=function@entry=0x7fffec064170 <__PRETTY_FUNCTION__.32848> "NODEPTR") at assert.c:101
#4  0x00007fffec029e5b in NODEPTR (i=0, p=<optimized out>) at ./storage/libmdbx/bits.h:1122
#5  0x00007ffbits.h:1122: NODEPTR: Assertion `NUMKEYS(p) > (unsigned)(i)' failed.fec032f8f in NODEPTR (i=0, p=<optimized out>) at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:6533
#6  mdbx_cursor_first (mc=mc@entry=0x7fffffff9c20, key=key@entry=0x7fffffff9b50, data=data@entry=0x0) at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:7026
#7  0x00007fffec0341a5 in mdbx_cursor_get (mc=mc@entry=0x7fffffff9c20, key=key@entry=0x7fffffff9b50, data=data@entry=0x0, op=op@entry=MDBX_FIRST)
    at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:7228
#8  0x00007fffec031262 in mdbx_page_alloc (num=num@entry=2, mp=mp@entry=0x7fffffff9df0, flags=flags@entry=15, mc=<optimized out>, mc=<optimized out>)
    at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:2204
#9  0x00007fffec034ae0 in mdbx_page_new (mc=mc@entry=0x7fffffffa0a0, flags=flags@entry=4, num=2, mp=mp@entry=0x7fffffff9e60) at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:8022
#10 0x00007fffec03506d in mdbx_node_add (mc=mc@entry=0x7fffffffa0a0, indx=0, key=key@entry=0x7fffffffa080, data=<optimized out>, pgno=pgno@entry=0, flags=<optimized out>)
    at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:8169
#11 0x00007fffec0396c0 in mdbx_cursor_put (mc=mc@entry=0x7fffffffa0a0, key=key@entry=0x7fffffffa080, data=data@entry=0x7fffffffa090, flags=flags@entry=65536)
    at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:7768
#12 0x00007fffec03bc1d in mdbx_freelist_save (txn=0x555555f5b170) at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:3614
#13 mdbx_txn_commit (txn=0x555555f5b170) at /home/server/work/cphalcon7/ext/storage/libmdbx/mdbx.c:4194

[erthink]

Couldn't reproduce. Please provide a testcase.

[dreamsxin]

I use it directly in the PHP extension, and guess is due to different compilation parameters.

[erthink]

@dreamsxin, however I should warn you - this could be a bug.

Historically MDBX inherites enough code opacity and bugs from LMDB (
So, be free to fill an issue.

[dreamsxin]

If you want to compile PHP extensions, you can try.

Environment:

• OS: Ubuntu16.04

Step 1:

sudo apt-get install php7.0-dev
git clone -b libmdbx https://github.com/dreamsxin/cphalcon7.git
cd cphalcon7/ext
phpize
./configure --enable-storage-libmdbx=yes
make -j4 && sudo make install

Step 2:
Add extension=phalcon.so to file /etc/php/7.0/cli/php.ini at the end of the file.

Create file test.php code:

<?php

$lmdb = new Phalcon\Storage\Libmdbx('./testdb');
$total = 100000;

$lmdb->begin();
for ($i = 0; $i < $total; $i++) {
	$lmdb->put('key_'.$i, 'value'.$i);
}
$lmdb->commit();

$lmdb->begin();
for ($i = 0; $i < $total; $i++) {
	$lmdb->put('key_'.$i, 'value'.$i);
}
$lmdb->commit();

Step 3:

php test.php

[erthink]

Confirmed, this is the BUG.

[dreamsxin]

Thank you!