Re: [Gm2] Building under Solaris 9 on Sparc -- update 3 (mostly good)

[Gaius Mulley]

john o goyo <address@hidden> writes:

> Greetings:
> 
> Here are my latest runs.
> 
> -------------------------------------------------------------------------
> 1. Solaris-i32lp64 PASSED (including gm2.paranoid).
> 
> The dejagnu test had the following failures:
> 
> ===[Truncated log file]===
> Native configuration is sparcv9-sun-solaris2
> 
>                  === gm2 tests ===
> 
> Running target unix
> Running
> /home/build/prj/bld/gcc-3.3.3-gm2/gcc/testsuite/gm2/cpp/pass/gm2.exp
> FAIL: gm2/cpp/pass/subaddr.mod,  -g
> FAIL: gm2/cpp/pass/subaddr.mod,  -O
> FAIL: gm2/cpp/pass/subaddr.mod,  -g -O

Hi John,

this can be safely ignored as it assumes that SIZE(CARDINAL) =
SIZE(ADDRESS). Incidentally what cpp architecture values are defined,
ie: here is the cpp #if defined I use for Pure64 on the Opteron:

MODULE subaddr;

FROM SYSTEM IMPORT ADDRESS;

VAR
   x, y: ADDRESS;
   i   : CARDINAL;
BEGIN
#if !defined(__x86_64)
   i := CARDINAL(x - y)
#endif
END subaddr.


we need to add "&& !defined(...)" to the #if


> Running
> /home/build/prj/bld/gcc-3.3.3-gm2/gcc/testsuite/gm2/ulmlib/std/pass/gm2.exp
> FAIL: ../gm2/ulm-lib-gm2/std/Functions.mod,  -g
> FAIL: ../gm2/ulm-lib-gm2/std/Functions.mod,  -O
> FAIL: ../gm2/ulm-lib-gm2/std/Functions.mod,  -g -O
> FAIL: ../gm2/ulm-lib-gm2/std/TimeIO.mod,  -g
> FAIL: ../gm2/ulm-lib-gm2/std/TimeIO.mod,  -O
> FAIL: ../gm2/ulm-lib-gm2/std/TimeIO.mod,  -g -O

these fail on all platforms..

so the 64 bit sparc port works as well as can be expected.


> -------------------------------------------------------------------------
> 2. Solaris-ilp32 build PASSED but gm2.paranoid FAILED:
> 
> ===[gmake gm2.paranoid output]===
> /xgm2 -Wnil -Wcase -Wbounds -Wreturn -g -gdwarf-2
> -funbounded-by-reference -Wverbose-unbounded -Wpim
> -Wpedantic-param-names -Wextended-opaque
> -c -B./stage2/gm2 -I../../gcc-3.3.3-gm2/gcc/gm2/gm2-compiler
> -I../../gcc-3.3.3-gm2/gcc/gm2/gm2-libs
> -I../../gcc-3.3.3-gm2/gcc/gm2/gm2-libiberty
> ../../gcc-3.3.3-gm2/gcc/gm2/gm2-compiler/FifoQueue.mod -o
> gm2/gm2-compiler-paranoid/FifoQueue.o
> 
> ../../gcc-3.3.3-gm2/gcc/gm2/gm2-compiler/FifoQueue.mod:43:coersion
> expecting a type
> 
> ../../gcc-3.3.3-gm2/gcc/gm2/gm2-compiler/FifoQueue.mod:43:only one
> parameter expected in a TYPE coersion
> 
> ../../gcc-3.3.3-gm2/gcc/gm2/gm2-compiler/FifoQueue.mod:45:too many
> errors in pass 3
> 
> gmake: *** [gm2/gm2-compiler-paranoid/FifoQueue.o] Error 1


hmm these errors are very strange, my immediate thoughts are that it
is picking up some old .o (or source) pre 2005-06-08 which had the
extended opaque problem.. But to check/or to try a different test
could you recompile after removing the -Wextended-opaque from the
GM2_FLAGS in gm2/Make-lang.in? If this works then there is either a
remaining bug in the extended-opaque code - or the build is picking up
an old binary/source file.

Tis very odd that this fails whereas the 64 bit version passes..

regards,
Gaius