Bug #52756 InnoDB Plugin 1.0.7 crashes on particular statement.

Description:
This is split from bug #52626

InnoDB Plugin 1.0.7 crashes with different trace:

0x00000034292c6952 in __select_nocancel () from /lib64/libc.so.6
(gdb) c
Continuing.

Program received signal SIGABRT, Aborted.
[Switching to Thread 1210108224 (LWP 10632)]
0x0000003429230015 in raise () from /lib64/libc.so.6
(gdb) bt
#0  0x0000003429230015 in raise () from /lib64/libc.so.6
#1  0x0000003429231980 in abort () from /lib64/libc.so.6
#2  0x00002aaaaab8ad94 in srv_error_monitor_thread (arg=0x0) at srv/srv0srv.c:2287
#3  0x0000003429e061b5 in start_thread () from /lib64/libpthread.so.0
#4  0x00000034292cd39d in clone () from /lib64/libc.so.6
#5  0x0000000000000000 in ?? ()
(gdb) bt full
#0  0x0000003429230015 in raise () from /lib64/libc.so.6
No symbol table info available.
#1  0x0000003429231980 in abort () from /lib64/libc.so.6
No symbol table info available.
#2  0x00002aaaaab8ad94 in srv_error_monitor_thread (arg=0x0) at srv/srv0srv.c:2287
        fatal_cnt = 11
        old_lsn = 6911444109
        new_lsn = 6911444109
#3  0x0000003429e061b5 in start_thread () from /lib64/libpthread.so.0
No symbol table info available.
#4  0x00000034292cd39d in clone () from /lib64/libc.so.6
No symbol table info available.
#5  0x0000000000000000 in ?? ()
No symbol table info available.
(gdb) c
Continuing.

Program terminated with signal SIGABRT, Aborted.
The program no longer exists.
(gdb) 

How to repeat:
See bug #52626

Apparently, this is the 600-second InnoDB rw-lock or mutex wait watchdog killing the server.

try with a 512M innodb_buffer_pool_size so that the statement doesn't run out of free pages...

alas, 1.0.7 plugin hangs. on this testcase as described by sveta.
"LRU len: 9029, unzip_LRU len: 8368" keeps increasing until no pages are
free.

and, the hang is in code from btr_page_split_and_insert() and functions, plus btr_page_get_sure_split_rec() returns null, so this might be a duplicate of bug #52964 . I will test the fix for that using this testcase.

Can someone test whether Marko's fix for bug 52964 fixes this (I'd be happy to if I can get access to the test data)?  That bug also caused an infinite loop in btr_page_get_sure_split_rec().

OK, using the test data from the original report I reproduced the original bug in a clean 5.1.46 build.  Breaking into it in gdb shows an identical callstack to the infinite loop Marko fixed in bug 52756.  I build a second version of the community source with Marko's patch applied and the test completed successfully.  I believe his patch solves this particular issue.

I caution claiming the entire issue is resolved however.  I have seen other crashes in similar situations (uncompressable data in a compressed table) even with Marko's patch.  I would suggest using the test data from this bug as a starting point for trying to reproduce these other issues as I can't provide my repro case directly. (Please see my latest comments on support issue 46284).

Thanks, Ryan

duplicate of bug#52964 based on Ryan's comments.  We should open new issues/bugs for subsequent problems found after the patch is applied, or else the patches/bugs/docs get very confusing.