This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Simple tests to check all probe aliases


Frank Ch. Eigler wrote:
Mike Mason <mmlnx@us.ibm.com> writes:

[...]
  probe nfs.*.*, nfs.*.*.* {}
This simple script exposed 24 unresolvable probe aliases.

Indeed, this is a class of genuine bugs.


[...] I know there's been disagreement as to the value of using -u,
that it would report $var problems that we already know about and
can do little to fix.

Not just that -- a user cannot accidentally refer to $vars like she can to probes (there is no equivalent of "probe * {}" like "$*"), so this class is much less interesting.

Hmm... maybe I'm not being clear. Wildcarding target variables isn't
something I even considered. I just think we should do our best to make sure the arguments in the tapset probes are actually accessible.
The only way I know to do that is test them all and, when we find
problems, fix those we can.



I still think too much test data is better than too little.  We can
use the kfail mechnaism if it becomes too annoying.

It will become instantly annoying. If you like, check in two copies of the tests, once with -u (and you might as well instantly mark it kfail), and once without (which should pass).

-u exposes both fixable and unfixable target variable problems.
I don't think we should ignore problems we can fix because we might expose some we can't.


I'm not trying to be difficult. I just want to help make SystemTap
better for the users. I think it's reasonable for users to expect
the advertised probe arguments (those documented in the man pages) to be available for use.


Mike



- FChE


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]