Indeed, 10.3-10.5 server starts with --expire_logs_days=p, but it is not changed in 10.3.24, but in 10.3.3.
KB says "Variables that take a numeric size can either be specified in full, or with a suffix for easier readability. [ https://mariadb.com/kb/en/server-system-variables/#setting-server-system-variables ] = p for petabytes
So server truncates --expire_logs_days=PMbjbjabfvj to 0 and starts,
but --expire_logs_days=HJHkbaxcp - returns "Unknown suffix 'H' used for variable 'expire_logs_days' (value 'HJHkbaxcp')"
-----------------
On 10.3.23 there was a typo 'expire_log_days' instead of 'expire_logs_days' -so server did not start, not because of 'p'

Isn't there mentioned about suffix and not prefix? Here we are giving P as prefix and still it didn't give any warning or error. Even if we give any other value like "HJH" something, server didn't start and no error mentioned in mariadb error log. I've tested it with 10.3.24. You are right that the same behaviour is in 10.3.24. (it was typo) but the question is if we put any other value like expire_logs_days=abc or expire_logs_days=def why it didn't give any warning or error? In this server didn't start but nothing in mariadb error log.

I get

 [ERROR] /10.3.24/bin/mysqld: Error while setting value 'jkhjk' to 'expire_logs_days'

immediately after attempt to start (not in a log)

It is documented that =<number><unit> works for command line options. That =<unit><garbage> works — this could be really unexpected. Better to fix it.

Still there's a question, =5garbage — should this work or not? This is =<number><unit><garbage>, not as clear-cut as two previous cases.

With 5garbage value, server is getting start and set value to 99.

nilnandan@yoga730:~/sandboxes/msb_10_3_24$ cat my.sandbox.cnf | grep expire

expire_logs_days=5garbage

nilnandan@yoga730:~/sandboxes/msb_10_3_24$ ./start

. sandbox server started

nilnandan@yoga730:~/sandboxes/msb_10_3_24$ mysql -umsandbox --socket=/tmp/mysql_sandbox10324.sock -p

Enter password: 

Welcome to the MariaDB monitor.  Commands end with ; or \g.

Your MariaDB connection id is 8

Server version: 10.3.24-8-MariaDB-enterprise MariaDB Enterprise Server

Copyright (c) 2000, 2018, Oracle, MariaDB Corporation Ab and others.

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

MariaDB [(none)]> show global variables like 'expire%';

+------------------+-------+

| Variable_name    | Value |

+------------------+-------+

| expire_logs_days | 99    |

+------------------+-------+

1 row in set (0.010 sec)

MariaDB [(none)]> 

I know that it works. The value is 5G, five gigabytes. The question was should it work? Do we want it to?
=garbage works and we definitely don't want it to. But what about =5garbage?

It's probably better if it wouldn't

• it's a minor bug, let's fix it in 10.6 only?
• the fix should be in both eval_num_suffix_ll() and eval_num_suffix_ull()
  like

  @@ -1019,7 +1019,7 @@ static longlong eval_num_suffix_ll(char *argument,

  *error= 0;

  errno= 0;

  num= strtoll(argument, &endchar, 10);

  - if (errno == ERANGE)

  + if (errno == ERANGE || argument == endchar)

  {

  my_getopt_error_reporter(ERROR_LEVEL,

  "Incorrect integer value: '%s'", argument);

• the rest can be done inside eval_num_suffix(), to have the check only in one place. like

  @@ -974,6 +974,11 @@

  static inline ulonglong eval_num_suffix(char *suffix, int *error)

  {

  + if (suffix[0] && suffix[1])

  + {

  + *error= 1;

  + return 0ULL;

  + }

  switch (*suffix) {

  case '\0':

  return 1ULL;