Fix raxRemove crash at memcpy() due to key size exceeds max Rax size

[VoletiRam]

Fix raxRemove crash at memcpy() (line 1181#) due to key size exceeds RAX_NODE_MAX_SIZE. Note that this could happen when key size was more than 512MB if we allow it by increasing the default proto-max-bulk-len. The crash could happen when we recompress the rax after removing a key due to expiry or DEL while memcpy() merge the key that exceed 512MB limit. While the counting phase has the size check, the actual compress logic is missing it which lead to this crash.

Crash explanation with example:


`

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 71.13%. Comparing base (2f2b8d1) to head (dff6d5e).
Report is 3 commits behind head on unstable.

Additional details and impacted files

@@             Coverage Diff              @@
##           unstable    #1722      +/-   ##
============================================
+ Coverage     70.97%   71.13%   +0.16%     
============================================
  Files           123      123              
  Lines         65536    65537       +1     
============================================
+ Hits          46511    46618     +107     
+ Misses        19025    18919     -106     

Files with missing lines	Coverage Δ
src/rax.c	82.78% <100.00%> (+0.02%)	⬆️

... and 11 files with indirect coverage changes

[zuiderkwast]

Great! Can you add a test case that would fail without this fix?

[madolson]

You also need to fix the DCO.

[VoletiRam]

Thank you for taking a look. @zuiderkwast @madolson

1. Fixed the missing DCO.
2. Added a regression unit test that reproduces the crash scenario:
   • The test is in test_rax.c.
   • It's disabled by default as it is a large memory test.

Please let me know if you need any adjustments to the change. Thank you.

[zuiderkwast]

Thanks for the test!

I have just a minor comment about the name and comment for the test case.

I have tried this test case without the fix and it indeed crashes. This is the stacktrace when running in GDB (with optimization so we can't see the args; sorry for that):

Program received signal SIGSEGV, Segmentation fault.
0x0000fffff7c94e80 in __memcpy_generic () from /lib64/libc.so.6
(gdb) bt
#0  0x0000fffff7c94e80 in __memcpy_generic () from /lib64/libc.so.6
#1  0x00000000004bc0fc in raxRemove (rax=0xfffff200c000, s=<optimized out>, len=<optimized out>, old=<optimized out>) at unit/../rax.c:1181
#2  0x00000000004c3540 in test_raxRemoveCrash (argc=<optimized out>, argv=<optimized out>, flags=<optimized out>) at unit/test_rax.c:1071
#3  0x00000000004762b4 in runTestSuite (test=test@entry=0x771a18 <unitTestSuite+176>, argc=argc@entry=4, argv=argv@entry=0xffffffffee08, flags=flags@entry=6) at unit/test_main.c:25
#4  0x000000000045a4cc in main (argc=4, argv=0xffffffffee08) at unit/test_main.c:61