This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Testsuite patch: Fix PPC debug-[12].c regressions


> I believe this patch will fix the regression tester's complains about
> debug-[12].c.

Your patch fixes some but not all fails of gcc.dg/debug/debug-[1,2].c.
There are still fails at -O.  I'm wondering if these two tests are
correctly implemented.

The tests started failing on all PA ports between Apr 16 15:25:55 UTC
and Apr 19 17:07:50.  The first AIX fails occur at Apr 17 02:19:34.
I believe the following patch introduced the failures:

2003-04-16  J"orn Rennecke <>

        Re-apply this patch:

	2002-05-16  Dale Johannesen  <>
	  * combine.c (cant_combine_insn_p):  Reenable combinations
	  involving hard regs unless CLASS_LIKELY_SPILLED_P.

This is the code from debug-1.c:

long foo(long p)
      long xyzzy = 0;
      if (p)
	xyzzy = 2;
      return xyzzy;

This is the only active insn after the combine pass on hppa64-hp-hpux11.11:

(insn 30 28 33 0 0000000000000000 (set (reg/i:DI 28 %r28 [ <result> ])
        (if_then_else:DI (ne (reg:DI 26 %r26 [ p ])
		(const_int 0 [0x0]))
	    (const_int 2 [0x2])
	    (const_int 0 [0x0]))) 44 {*} (nil)
    (expr_list:REG_DEAD (reg:DI 26 %r26 [ p ])

The local xyzzy is no longer present in the rtl.  As a result, xyzzy
isn't present in the assembler output and the test fails.  The assembler
code looks perfectly valid to me but possibly I'm missing some subtle
issues with respect to lexical blocks.

J. David Anglin                        
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]