Re: Testsuite on Solaris 11.3/sparc

[Gaius Mulley]

john o goyo <jog37@riddermarkfarm.ca> writes:

> Greetings, Gaius.
>
> I confess to not understanding how the testsuite is run.  Here is the
> latest summary:
>
>                 === gm2 Summary ===
>
> # of expected passes            11866
> # of unexpected failures        27
> # of unresolved test cases       7
>
>
> (1) As I reported earlier, the various tests based on the following
> modules all pass if linked with the appropriate libraries (libnsl,
> libsocket):
>
>     gm2/projects/iso/run/pass/halma/halma.mod
>     gm2/switches/whole-program/pass/run/hello.mod
>     gm2/switches/whole-program/pass/run/hello2.mod
>     gm2/switches/whole-program/pass/run/tiny.mod
>     gm2/switches/whole-program/pass/run/tiny2.mod
>
> I  added LDFLAGS="-lnsl -lsocket" to the configuration script but that
> did not propagate down to the testsuite.  I added LD_OPTIONS="-lnsl
> -lsocket" to the environment but that was ignored in the testsuite. 
>  However, the installed gm2 builds all of them and they run without
> problems.
>
> How does one convince dejagnu, which I do not know, to link in the
> required libraries?
>
> (2)  In the test gm2/pim/pass/TestLong4.mod, the number
> 9223372036854775808 (8000000000000000H) is assigned a LONGCARD. Whence
> this number, Gaius?  It is much larger than MAX(LONGINT) + 1 on Sparc V9.
>
> Sincerely,
> john

Hi John,

re: (1) here is a patch which might be along the right lines?

diff --git a/gcc/testsuite/lib/gm2.exp b/gcc/testsuite/lib/gm2.exp
index 25b5e684b3a..9ed8c6ebd53 100644
--- a/gcc/testsuite/lib/gm2.exp
+++ b/gcc/testsuite/lib/gm2.exp
@@ -218,6 +218,10 @@ proc gm2_target_compile_default { source dest type options 
} {
     if [info exists TOOL_OPTIONS] {
        lappend options "additional_flags=$TOOL_OPTIONS"
     }
+    if [istarget *-*-solaris2*] {
+       lappend options "ldflags=-lnsl -lsocket"
+    }
+
     lappend options "timeout=[timeout_value]"
     lappend options "compiler=$GCC_UNDER_TEST"
     # puts stderr "options = $options\n"


I'm not sure if the solaris2 string is correct though.  But it maybe
worth just forcing lappend options "ldflags=-lnsl -lsocket" as an
experiment and refining afterwards.

Hope this helps?

regards,
Gaius