The Samba-Bugzilla – Bug 6467
Feature request: smbtorture - timestamps
Last modified: 2009-07-07 03:13:24 UTC
smbtorture seems to be capable of running for long periods of time. This may be a bug, or it may just be normal. Either way:
PLEASE add a date/time stamp in the output, in between tests, at the beginning of tests, even in the middle of long tests. Whenever you see fit. But it would be a great help when looking at smbtorture output to understand when things happened. I am especially referring to running batteries of tests, such as "ALL".
Created attachment 4333 [details]
Here a small patch which adds timestamps on the beginning and the end of testsuites
The patch changes lib/subunit, which is an implementation of a standard unit testing framework. Are you sure that's a correct use of subunit?
What tests are you running that are so long that you need this kind of 'I'm still going' detail?
>>>What tests are you running that are so long that you need this kind of 'I'm
still going' detail?
That's a good question. I ran the full set of tests a couple a weeks ago, and when I came back to the office in the morning, it was still running! But as the case may be, it's probably fine to just have the timestamps before each individual test or subtest, and not in the middle of them.
As I mentioned in 6513, it would nice to see
- clear headers and whitespace before each test
- more verbose description of errors in a referenced external website
The output of sambatorture is cryptic and meaningless to the typical sysadmin. Don't you imagine? It looks like only used for internal development purposes. Maybe add a flag, like sambatorture -U (for user, or something) where there is only simple, useful and actionable information, and less internal diagnostics.
As you have noticed, smbtorture is purely designed as a testing tool for developers.
We are working to make the output more regular, and smbtorture has multiple output formats (to allow parsing by our test harnesses), but it isn't really intended as a tool for use by administrators.
That said, if done properly, within the framework, I don't object to the idea of better output formatting.
Since we have other test tools as well, I'd rather see the improved test formatting being done as a subunit filter. That way we don't have to patch the talloc, tdb and Python tests as well.
In my patch I changed the subunit filter. The "smbtorture.c" changes chould be reverted. Is this what you mean, Jelmer?
For now we agreed to close this bug as "WONTFIX" since we don't see the need to use smbtorture as general benchmark tool.