main.mysqlpump_basic_lz4 dep8 test fails with a false positive

[Robie Basak]

--- /usr/lib/mysql-test/r/mysqlpump_basic_lz4.result 2019-06-25 13:23:30.000000000 +0300
+++ /tmp/tmp.Txe2A0qlVp/var/log/mysqlpump_basic_lz4.reject 2019-11-15 12:52:28.844432010 +0300
@@ -1,3 +1,4 @@
+*** LZ4 command line interface 64-bits v1.9.2, by Yann Collet ***
 CREATE DATABASE bug21644479_lz4;
 USE bug21644479_lz4;
 CREATE TABLE t1 (a INT);

mysqltest: Result content mismatch

This is against lz4 1.9.2-2 in focal-proposed.

I suspect that lz4 -V, as called by mysqlpump_basic_lz4.test, is now providing that output in stdout and not stderr, so it doesn't get redirected and ends up in the test output that now fails to compare exactly. I haven't verified this.

[Andreas Hasenack (ahasenack)]

In eoan it's in stderr, with 1.9.1-1:

andreas@nsnx:~$ lz4 -V
*** LZ4 command line interface 64-bits v1.9.1, by Yann Collet ***
andreas@nsnx:~$ lz4 -V 2>/dev/null
andreas@nsnx:~$ lz4 -V > /dev/null
*** LZ4 command line interface 64-bits v1.9.1, by Yann Collet ***
andreas@nsnx:~$

In focal-proposed it's in stdout, with 1.9.2-2:
ubuntu@f1:~$ lz4 -V
*** LZ4 command line interface 64-bits v1.9.2, by Yann Collet ***
ubuntu@f1:~$ lz4 -V > /dev/null
ubuntu@f1:~$ lz4 -V 2>/dev/null
*** LZ4 command line interface 64-bits v1.9.2, by Yann Collet ***
ubuntu@f1:~$

[Andreas Hasenack (ahasenack)]

mysql in focal already has a fix for this:
- --exec lz4 -V 2> $LZ4_EXEC_LOG
+ --exec lz4 -V > $LZ4_EXEC_LOG 2>&1

But that just makes sure we get the lz4 version in the log file, regardless of the lz4 version.

But this is odd, since the result file doesn't have that string, so it will never match the test result. If we add it to the results file, it will have to be updated every time lz4 is updated, which is suboptimal to say the least.

[Andreas Hasenack (ahasenack)]

If the test just wants to check if lz4 exists and runs, then its output should be redirected to /dev/null, maybe that's the best fix here?

[Andreas Hasenack (ahasenack)]

This is fixed actually, the failing test in 8.0.18 is another one. 8.0.17 is what failed with the lz4 error.