Re: PATCH: Remove README.QMTEST

2011-03-06 Thread Mark Mitchell
On 3/6/2011 2:57 PM, Mike Stump wrote:

>> It was recently pointed out to me that we still have the
>> README.QMTEST file in gcc/testsuite. That file talks about how to
>> use QMTest instead of DejaGNU to test G++

> If you were to summarize the 3 most compelling features provided, I'd
> be interested.  I'm thinking it was parallel testing...

That's one.  A couple of others:

(a) QMTest is a structured system, i.e., you can enumerate all of the
tests and ask questions about them.  In contrast, with DejaGNU, you can
only figure out what tests you have by running them -- we depend on
executable .exp files to enumerate tests, not declarative information.
For example, try asking DejaGNU "how many tests do I have?", or "how
many tests do I have for Solaris targets that aren't applicable to Linux
targets?"

(b) QMTest has a facility for saving results in structured formats
(rather than text-based .log/.sum files) so that they can be processed
or queried later.  One immediate use of this is that you can compare
results today with results yesterday in ways more structured that "diff".

Thank you,

-- 
Mark Mitchell
CodeSourcery
m...@codesourcery.com
(650) 331-3385 x713


Re: PATCH: Remove README.QMTEST

2011-03-06 Thread Mike Stump
On Mar 6, 2011, at 1:39 PM, mitch...@codesourcery.com (Mark Mitchell) wrote:
> It was recently pointed out to me that we still have the README.QMTEST
> file in gcc/testsuite.  That file talks about how to use QMTest
> instead of DejaGNU to test G++

If you were to summarize the 3 most compelling features provided, I'd be 
interested.  I'm thinking it was parallel testing...


Re: PATCH: Remove README.QMTEST

2011-03-06 Thread FX
> It was recently pointed out to me that we still have the README.QMTEST
> file in gcc/testsuite.  That file talks about how to use QMTest
> instead of DejaGNU to test G++ -- but the information there is just
> not accurate anymore.  Various things have bit-rotted since the point
> at which that was written.  Given that nobody is actively working on
> making this work, it's best just to remove the file.

There are also qmtest targets in gcc/Makefile.in. Shouldn't these be removed 
too?

FX