HPUX testsuite tweakage

[Zack Weinberg]

This patch corrects a number of spurious XPASSes and FAILs on
ia64-hpux and hppa*-hpux.

1) g++.dg/init/init-ref4.C and g++.old-deja/g++.pt/static3.C were
   using a hardwired list of platforms to XFAIL instead of
   dg-require-weak.  Recent versions of HPUX support weak symbols so
   these tests were getting XPASS.

2) g++.dg/parse/attr-ctor1.C is similar, except that the property it
   cares about is named sections.  (This one must have been getting a
   spurious failure on a whole bunch of a.out targets.)  For this I
   had to write a new dg-require function.  I'd appreciate a second
   pair of eyes on this code - my Tcl-fu is weak.

3) HP's linker prints a count of errors on unsuccesful exit, which was
   being picked up as an excess error, causing a spurious failure of
   g++.pt/const2.C.  Compounding the problem was that old-deja.exp
   overrode g++-dg-prune with its own definition, so adding an
   appropriate regexp to prune_gcc_output wasn't enough.  I nuked that
   and merged the handful of additional prunes into prune.exp.

4) GAS for IA64 issues warnings about code that may not execute
   correctly on A/B-step Itanium.  There is no way to turn this 
   off.  This causes spurious failures in the libstdc++ testsuite.  
   The most expeditious fix is to prune them.

Bootstrapped ia64-hpux, hppa2.0w-hpux, hppa64-hpux (3.4 branch); the
net benefit is to eliminate the following FAILs and XPASSes:

(ia64)
-XPASS: g++.dg/init/init-ref4.C execution test
-FAIL:  g++.old-deja/g++.pt/const2.C (test for excess errors)
-XPASS: g++.old-deja/g++.pt/static3.C execution test
-FAIL:  25_algorithms/min_max.cc (test for excess errors)
-FAIL:  26_numerics/cmath/overloads.cc (test for excess errors)
-FAIL:  26_numerics/complex/13450.cc (test for excess errors)
-FAIL:  26_numerics/complex/pow.cc (test for excess errors)

(hppa)
-XPASS: g++.dg/init/init-ref4.C execution test
-XPASS: g++.dg/parse/attr-ctor1.C  (test for bogus messages, line 9)
-XPASS: g++.old-deja/g++.pt/static3.C execution test

Mainline testing in progress.  OK 3.4?  N.B.: the tree this was tested
in has my previous two ia64 patches included.

zw

gcc/testsuite:

        * lib/gcc-dg.exp (dg-require-named-sections): New.
        * lib/target-supports.exp (check_named_sections_available): New.
        * lib/prune.exp (prune_gcc_output): Incorporate prunes from
        old-deja.exp's definition of g++-dg-prune.  Also prune
        "[0-9]+ errors\." message emitted by HP linker.
        * g++.old-deja/old-deja.exp: Delete local definition of g++-dg-prune.
        
        * g++.dg/init/init-ref4.C: Use dg-require-weak.
        * g++.old-deja/g++.pt/static3.C: Likewise.
        * g++.dg/parse/attr-ctor1.C: Use dg-require-named-sections.

libstdc++-v3:
        * testsuite/lib/prune.exp: Also prune "Assembler messages:"
        and ia64 GAS's "Additional NOP may be necessary" warning.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/x-patch
Size: 7155 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20040409/9e5ce1c0/attachment.bin>