MTR can very well do that. The test below ensures that the query fails with "Table 'test.passwords' doesn't exist".

--error ER_NO_SUCH_TABLE

select * from passwords;

Feature requests are filed with the type 'Task'.

Thank you. Is there a way I can make mysqltest do that?

I don't understand the question. Make mysqltest do what? It is an MTR test case, you put it in a file and run it as a usual MTR test.

I am asking if I can implement the aforementioned requirement with the syntax mysqltest accepts (which is, as far as I know, pure SQL).

It is a syntax which mysqltest accepts. There is a lot more in mysqltest than just pure SQL.

jacob.moorman,

MTR (the test framework), including mysqltest (the client program which actually runs the tests) are a part of the server and included in the distribution, but I don't think we have any meaningful documentation for mysqltest in the KB.

Ah, thank you. That is great to hear. Do you have any existing documentation about the syntax? I'm glad for any scrap of it I can get or find.

I couldn't find anything right now, maybe jacob.moorman will be able to point at something that I missed.

You can use the actual existing tests to get the idea, in particular I expect that mysqtest.test in the main suite should contain a lot of examples of the syntax (it's <sourcedir>/mysql-test/main/mysqltest.test or <sourcedir>/mysql-test/t/mysqltest.test, depending on which branch you use).
Naturally, other tests can also be looked int as examples, but most of the regression tests only use simple commands. mysqltest.test should have higher coverage for its syntax.

Thank you! That is already an enormous help.

This task is accepted to our backlog.

I have another question: Is there a way to measure the query execution time in microseconds or nanoseconds using mysqltest? I haven't found anything regarding that.

not that I know of. One can do, like

set @before=now(6);

then execute the query and then

select unix_timestamp(now(6))-unix_timestamp(@before);

but generally this is not something one should do in mysql-test, because tests are run on a variety of platforms and architectures, some are slow, some are fast, some are very loaded. Any timing-dependent tests are bound to fail eventually.

It's something one can do locally when experimenting or debugging or something, but it's never reliable enough to be pushed, so there's little interest in time measurements in mysqltest.

jacob.moorman from our discussion, I decided to reassign Kenneth's ticket to you?