Re: [Boost-build] Bug in building shared libs on HP/UX(wrong linker arguments)

From: Boris Gubenko (Boris.Gubenko_at_[hidden])
Date: 2007-05-16 08:50:15

> This case int the switch shlould be remove [...]

I'm not sure if  the code under 'case HP_UX'  ever gets executed on HP-UX. 
I.e., I'm not sure if "HP_UX" is the correct name for the platform.

>From one of your earlier postings, it appears that -Wl,--start-group is what 
is killing linking on HP-UX. Since there is no case for HP-UX in 'switch 
$(linker)', it looks like 'case gnu' is executed. In init, you need to add 
case for linker on HP-UX. Something like:

        if [ os.name ] = HPUX
        {
            linker = hpux ;
        }

And in init-link-flags, you need to add case for hpux. Again, look at how it 
is done for osf.

HTH,
  Boris

----- Original Message ----- 
From: <Oliver.Kowalke_at_[hidden]>
To: <Boris.Gubenko_at_[hidden]>; <boost-build_at_[hidden]>
Sent: Wednesday, May 16, 2007 8:25 AM
Subject: RE: [Boost-build] Bug in building shared libs on HP/UX(wrong linker 
arguments)

> And you can find the jam file for aCC on HP-UX in
> tools/build/v2/tools/acc.jam. You can look at HP-UX Linker
> options in acc.jam. In gcc.jam, you can look at
> conditionalization for osf
> (Tru64) as an example.

Thanks - I still try to figure out how bjam v2 works (I believe some
changes have to be made).

> Btw, there may be a bug in gcc.jam:
>
>   switch [ modules.peek : JAMUNAME ]
>   ...
>   case HP_UX :
>
> Why is HP_UX and not HP-UX which is what uname(1) returns?
> And why is it not HPUX which is what os.name returns?
> What should it be?

This case int the switch shlould be remove because gcc (>= 3.x) supports
threads on HP/UX.

Oliver