Re: jacks oddities

[Ranjit Mathew <rmathew at gmail dot com>]

Andreas Tobler wrote:
> Hi all,
> 
> I recently wondered about having only 1400 passes in my jacks suite. 
> Bryce told me that there are quite a few more. And running jacks outside 
> the dejagnu environment really gave me about 4000 passes.

Right. As of today's mainline, we should be getting 4069 PASSes
and 785 XFAILs with Jacks (no FAILs and no XPASSes), at least on
i686-pc-linux-gnu.

I can't make out from your mail how exactly have you
checked out Jacks, but I for one keep the Jacks CVS tree
directly under "$GCC_CVS_DIR/libjava/testsuite/libjava.jacks"
and run it (after having bootstrapped GCC) from under the
"$GCC_BUILD_DIR/i686-pc-linux-gnu/libjava/testsuite" folder
using "make -k check" or using an explicit
"make -k check RUNTESTFLAGS=jacks.exp".

This has worked quite well for me so far. If your setup is
the same and it doesn't work for you quite as well, we
need to figure out what's going wrong.


> Anyway, during this investigation I wondered why we do not set 
> JAVAC_DEPRECATION_FLAG to -Wdeprecated in the gcj_setup?
> 
> Doing so would give us another 60 XPASSES.
> 
> Any comments?

Yes: I don't know whether to exclaim "D'oh!" or "Wow!". ;-)

I have verified that your patch indeed gets us 58 XPASSes
(58 testcases out of 68 in 4.7.10) and no FAILs.

Since Tom has basically approved it, please check in
your patch to jacks.exp as well as jacks.xfail.

Thank you!

Ranjit.

PS: BTW, much of Jacks can possibly be speeded up
by using "-fsyntax-only", but there are some tests
which execute the generated classes, so I couldn't
apply this globally.

-- 
Ranjit Mathew          Email: rmathew AT gmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/