Change

If the environment variables ydb_routines and gtmroutines are undefined, the value of $ZROUTINES is set to a default. In this MR, the default has been updated to include all shared library object files in $ydb_dist/plugin/o In M mode or all shared object files in $ydb_dist/plugin/o/utf8 in UTF-8 mode.

Here we say utf8 but in line 115 and a few other lines we say utf8/. Can we be consistent?

The macro value strings across multiple lines used to be indented at the same column previously. Now they are arbitrary. Please fix that. It was a lot more readable before.

Yes that would be nice. But that would mean more enhancements to mupip dumpfhead -flush which I am not inclined to spend time on. That is because mupip_dump_fhead() calls wcs_flu() to do the flush. And whether there is nothing or something to flush is decided inside wcs_flu(). So mupip_dump_fhead() cannot issue an error in either case (the nothing to flush OR the something to flush cases) if we do not issue the DBRDONLY error before the wcs_flu() call. Hence my suggestion to issue the DBRDONLY error in mupip_dump_fhead() before the call to wcs_flu().

@ksbhaskar : The following is not accurate.

after clients have connected and disconnected to the sockets, while a subsequent first WRITE /WAIT(timeout) command would time out, a second WRITE /WAIT(timeout) command would hang ...

Instead, how about

after clients have connected to all listening sockets, a WRITE /WAIT(timeout) command on the server would time out correctly as many times as there are listening sockets but after that a subsequent WRITE /WAIT(timeout) command would hang ...

v53003/C9I09003044 fix

Since zshow I is included in zshow *, and the test stores zshow * output into a global variable, the new $ZYJOBPARENT adds some extra operations to the statistics. The test has been updated to expect the adjusted statistics that account for this.

1. Like I have mentioned before, in the MR description please include a mention of the thread which discussed this failure.
2. And do close that original thread (which is in some closed MR) with a mention of !2627 as having the fix.

Include [YDB#918] in the MR title as that is what introduced the $ZYJOBPARENT ISV.

zshow I -> zshow I in the MR description

Forgot to reply to your prior comment. You can say INST2's receiver instead of secondary's receiver. But note that the receiver server is going to be running only in one instance and that is INST2 (unlike the source server which is going to be running in INST1 and INST2). So there is not much need for this distinction.

@ksbhaskar : Yes it is an upstream issue. Label added as requested.

30 -> 300

Can we fold ydb1140a.m, ydb1140b.m and ydb1140c.m into labels inside ydb1140.m and maintain this in the repository?

30 seconds might not be enough on some slow systems. So just make it 300 to be safe.

MM does not support before image journaling. But this test does not run mupip journal -rollback -backward and so there is no need for before image journaling in this test.

Use $tst_jnl_str instead of enable,on,before in the mupip set -journal command that is done in this subtest. That should ensure nobefore is used in case of MM and avoid this error as well as remove the need for the gtm_test_setbeforeimage.csh invocation.

-r -> -run

Use full forms of qualifiers in test scripts.

The previous line (comment) still says disable this environment variable.

You had mentioned the following in your previous comment in this thread. I was referring to this.

or is that just the same as a dumpfhead with nothing to flush

Does that answer your question?

@ksbhaskar : Thanks. I did not know that.

Narayanan Iyer (4082b816) at 16 Mar 20:44

... and 1 more commit