Re: gcc-ada-3.3 task problems under solaris 9

[David Holm <david@realityrift.com>]

It might be more convenient to run gnatmake with --RTS=pthread than to relink adainclude/adalib.

//David Holm

On Mon, 16 Jun 2003 10:29:20 +0200
Thierry Lelegard <thierry.lelegard@cptechno.com> wrote:

> This is a known issue of GNAT 3.15. Its runtime used an unsupported
> feature of the Solaris libthread which is no longer valid on Solaris 9.
> GNAT 5.00a has a fix and works on both Solaris 8 and 9.
> 
> If you are not an ACT customer and do not have access to 'a' versions,
> you can still use GNAT 3.15p on Solaris 9, provided that you use the
> pthreads instead of the native Solaris threads. (Slightly worse performance
> may be expected).
> 
> Move to gnat/lib/gcc-lib/.../2.8.1 and modify the symbolic links adainclude
> and adalib. Make them point into rts-pthread instead of rts-native.
> 
> I do not know anything about gcc-ada-3.3 but it has a bad reputation however.
> Use GNAT release instead.
> 
> -Thierry
> 
> nathand@thx.vic.optushome.com.au wrote:
> > 
> > Hi All,
> > 
> > Just wondering if anyone else has experienced problems with ada 3.3 tasking under solaris 9.
> > I have compiled gcc-ada-3.3 under solaris 9 64bit using gnat-3.15 binaries to assist the
> > compilation. A simple single threaded process works just fine, and exits cleanly.
> > A program containing a task coredumps with the following information from gdb
> > 
> > Program received signal SIGSEGV, Segmentation fault.
> > 0x2931c in system__task_primitives__operations__self () at s-tpopse.adb:174
> > 174        if To_Iptr (Get_G7 - 120).all /=
> > 
> > When I compile a simple test program with either the gcc-ada-3.3 compiler, or the 3.15
> > compiler I get the same result as shown above.