Re: ddd-3.3.9-test2 build feedback HPUX once again...

[Arnaud Desitter]

----- Original Message ----- 
From: "Tuomo Tikkanen" <address@hidden>
Newsgroups: gmane.comp.debugging.ddd.general
Sent: Friday, May 21, 2004 10:13 PM
Subject: Re: ddd-3.3.9-test2 build feedback HPUX once again...


> 
> >>- because above I had to use g++ on this platform (have tried versions
> >>3.4.0 and 3.3.3). Unfortunately the compilation does not go through
> >>because in the final link phase of ddd binary following happens:
> >>-----8<-----8<-----8<-----8<-----8<-----
> >>/usr/ccs/bin/ld: Unsatisfied symbols:
> >>   typeinfo for _Assoc<int, VarArray<int> >(data)
> >>   typeinfo for DynArray<int>(data)
> >>   typeinfo for Assoc<string, string>(data)
> >>   typeinfo for DynArray<EdgeAnnotation*>(data)
> >>   typeinfo for VarArray<void*>(data)
> >>   typeinfo name for ThemePattern(data)
> >>   typeinfo for VarArray<EdgeAnnotation*>(data)
> >>   typeinfo for DynArray<GraphNode*>(data)
> >>   typeinfo name for UndoBufferEntry(data)
> >>   typeinfo for _Assoc<string, string>(data)
> >>   typeinfo for Assoc<int, VarArray<int> >(data)
> >>   typeinfo for VarArray<_WidgetRec*>(data)
> >>   typeinfo for VarArray<GraphNode*>(data)
> >>   typeinfo for DynArray<EntryType>(data)
> >>   typeinfo for VarArray<UndoBufferEntry>(data)
> >>   typeinfo for _Assoc<_WidgetRec*, string (*)(string const&)>(data)
> >>   typeinfo for DynArray<UndoBufferEntry>(data)
> >>   typeinfo for Assoc<_WidgetRec*, MinMax>(data)
> >>   typeinfo for VarArray<DispValue*>(data)
> >>   typeinfo for DynArray<VarArray<int> >(data)
> >>   typeinfo for _Assoc<string, VarArray<VarArray<int> > >(data)
> >>   typeinfo for DynArray<void*>(data)
> >>   typeinfo for DynArray<stat>(data)
> >>   typeinfo for DynArray<CodeCacheEntry>(data)
> >>   typeinfo for VarArray<CodeCacheEntry>(data)
> >>   typeinfo for Assoc<string, ThemePattern>(data)
> >>   typeinfo for Assoc<string, VarArray<VarArray<int> > >(data)
> >>   typeinfo for _Assoc<string, SourceOrigin>(data)
> >>   typeinfo name for DispGraph(data)
> >>   typeinfo for _Assoc<_WidgetRec*, MinMax>(data)
> >>   typeinfo for DynArray<DispValue*>(data)
> >>   typeinfo name for DispValue(data)
> >>   typeinfo for VarArray<long>(data)
> >>   typeinfo for _Assoc<_WidgetRec*, string>(data)
> >>   typeinfo for Assoc<_WidgetRec*, string>(data)
> >>   typeinfo for DynArray<string>(data)
> >>   typeinfo for DynArray<long>(data)
> >>   typeinfo for VarArray<EntryType>(data)
> >>   typeinfo for VarArray<VarArray<int> >(data)
> >>   typeinfo name for DispNode(data)
> >>   typeinfo for VarArray<stat>(data)
> >>   typeinfo for Assoc<string, SourceOrigin>(data)
> >>   typeinfo for VarArray<string>(data)
> >>   typeinfo for VarArray<int>(data)
> >>   typeinfo for Assoc<_WidgetRec*, string (*)(string const&)>(data)
> >>   typeinfo for DynArray<_WidgetRec*>(data)
> >>   typeinfo for _Assoc<string, ThemePattern>(data)
> >>collect2: ld returned 1 exit status
> >>make[2]: *** [ddd] Error 1
> >>-----8<-----8<-----8<-----8<-----8<-----
> >>
> >>I would guess that this is problem with g++ on HPUX and not by anyway
> >>related to DDD itself. I have not yet had time to inspect the problem
> >>more carefully though.
> >>    
> >>
> >
> >Just a guess. Looks in PROBLEMS, section Cygwin. Could try the trick
> >to remove some #pragma and see if it makes it work ?
> >In any case, it will be useful to fill a bug report for gcc with a
> >simplified
> >test case.
> >(I thought that http://gcc.gnu.org/bugzilla/show_bug.cgi?id=9941
> >had it fixed in 3.4 and 3.3.3).
> >
> >  
> >
> Thank you pointing this out.
> 
> I made a test using g++ version 3.4.0 and it seems to be so that the 
> compiler bug is still there at least on some level:
> - on Linux the compilation went OK with the same g++ version (Linux does 
> have weak symbols)
> - on Solaris I used only native compiler version 5.5 on Solaris8 and 
> version 5.3 on Solaris 2.6 (but Solaris does have weak symbols too, so I 
> believe the g++ would work just as fine)
> - on HPUX 11 I used native aCC compiler, but at first I was using g++ 
> 3.4.0 there too to compile 3.3.9-test2 and -- if I remember it right -- 
> the reason why I turned to aCC was the fact that compilation ended to 
> very same error that happened now on HPUX 10.20
> 
> Interesting point is that if I download the test case from [1] and 
> repeat the compilation as shown there it really shows that particular 
> bug has been corrected in 3.4.0, but exists still in 3.3.2 and 3.3.3:
> 
> -----8<-----8<-----8<-----8<-----8<-----8<-----
> <hepux1020:/share/wrk/tikkanen/gccbug> diff pragma_output_from_340 
> no_pragma_output_from_340
> <hepux1020:/share/wrk/tikkanen/gccbug> diff pragma_output_from_333 
> no_pragma_output_from_333
> 11c11
> <          U typeinfo name for LS
> ---
>  > 00000000 t typeinfo name for LS
> <hepux1020:/share/wrk/tikkanen/gccbug> diff pragma_output_from_332 
> no_pragma_output_from_332
> 11c11
> <          U typeinfo name for LS
> ---
>  > 00000000 t typeinfo name for LS
> -----8<-----8<-----8<-----8<-----8<-----8<-----
> 
> Anyhow I managed to make working DDD version on HPUX 10.20 and using g++ 
> 3.4.0 by commenting out "#pragma interface" line from ddd/Assoc.h file.
> 
> [1] http://gcc.gnu.org/bugzilla/show_bug.cgi?id=9941

Obviously, there is another issue not covered by  bug9941.
It would be nice to fill a new gcc bug report as it is the only way to get
bugs fixed. Isolating a small test case may be a long task though.

You could try the test case in
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13420
, which illustrates the link failure of ddd on cygwin wich ddd 3.3.1.
This bug report was closed as duplicate of 9941 but it may well
be that the test case still fails.

Regards,