autotools transition report: need a flag day

[Dan Kegel]

Phil Edwards wrote:
> On Tue, Aug 26, 2003 at 09:43:14AM -0700, Dan Kegel wrote:
> 
>>cd $TARGET/$LIBSTDCXX_DIR/testsuite
>>runtest --tool=$LIBSTDCXX_TEST --target=$TARGET -v || echo Some $LIBSTDCXX_TEST test failed...
> 
> Try adding "make site.exp" between those commands.

I already do that... my whole script is at
http://www.kegel.com/crosstool/current/crosstest.sh

>>(I'm invoking runtest directly
>>rather than 'make check'.  Guess I'm paying for that now, but
>>invoking it directly makes it a lot easier to figure out how
>>options get passed to runtest :-)
> 
> Simplifying only a tiny bit, the generated make lines are basically:
> 
>     AM_RUNTESTFLAGS =
>     RUNTESTDEFAULTFLAGS = --tool $$tool --srcdir $$srcdir
>     RUNTESTFLAGS =
>     DEJATOOL = libstdc++
> 
>     check-DEJAGNU: site.exp
>         srcdir=/path/to/source; \
>         l='$(DEJATOOL)'; for tool in $$l; do \
>           runtest $(AM_RUNTESTFLAGS) $(RUNTESTDEFAULTFLAGS) $(RUNTESTFLAGS); \
>         done; \
> 
> Once, the variables and loops come out, it's just
> 
>     runtest --tool libstdc++ --srcdir /path/to/source
> 
> with the target and everything else taken from the generated site.exp.
> 
> The way you pass additional flags is by setting RUNTESTFLAGS on the make
> command line:
> 
>     make check RUNTESTFLAGS="--verbose"

Yes, I'm aware of that.  All that machinery gives me the willies.  The
gcc makefiles are exceedingly complex, and when I was figuring out how
remote testing all worked, bypassing the makefile gunk really helped keep my head from exploding.
Still, now that I'm comfortable with my understanding of things,
I might try using "make check-gcc" and "make check-g++" again.

Nevertheless, it'd be nice if the naming for libstdc++ was consistent.
How hard would it be to drive a stake through the heart of the -v3 suffix?
- Dan

-- 
Dan Kegel
http://www.kegel.com
http://counter.li.org/cgi-bin/runscript/display-person.cgi?user=78045