dejagnu lied. was: 971023 on OpenServer test results; eh gets ugly.

[Robert Lipe]

> It's worth confessing that COFF and ELF on this platform exercise
> sjlj and dwarf2 eh respectively, so perhaps the eh problem is 
> independent of the method used.


It seemed really strange that the two forms of eh would fail 
identically.   So I ran some tests by hand.   For the one case
I tried (eh10.C) -mcoff (sjlj) worked but -melf, the default which
uses dwarf2,  failed.

I started slogging through the dejagnu logs and am pretty much
convinced that dejagnu tricked me into thinking it was running
both -melf and -mcoff.   Yes, it runs the tests twice but it never
actually passes those flags to the compiler.  Clearly I haven't met 
it on it's terms.

I was executing:
nice runtest --outdir ${COFFDIR} --tool gcc \
        --srcdir ${SRC}/gcc/testsuite \
        --target_board 'unix{-mcoff,-melf}' \
        GCC_UNDER_TEST=negcs \
        CXXFLAGS="-g -O"

What am I missing to get dejanu to run the tests twice and actually
*use* the -mcoff/-melf flags?

So, disregard the COFF/sjlj portions of my test results as they're
really ELF/dwarf results masquerading as COFF.


> 		=== g++ Summary for unix/-melf ===
> 
> # of expected passes		3339
> # of unexpected failures	41
> # of unexpected successes	1
> # of expected failures		82
> # of untested testcases		6
> 
> 		=== g++ Summary ===
> 
> # of expected passes		6678
> # of unexpected failures	82
> # of unexpected successes	2
> # of expected failures		164
> # of untested testcases		12
> negcs version egcs-2.90.14 971023 (gcc2-970802 experimental)

-- 
Robert Lipe       http://www.dgii.com/people/robertl       robertl@dgii.com