This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [build] Support libgo on Solaris 2
- From: Ian Lance Taylor <iant at google dot com>
- To: Rainer Orth <ro at CeBiTec dot Uni-Bielefeld dot DE>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 11 Jan 2011 22:41:27 -0800
- Subject: Re: [build] Support libgo on Solaris 2
- References: <m3y66rnuku.fsf@pepe.airs.com>
> From: Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
> Date: Mon, 10 Jan 2011 16:23:05 +0100
> * runtime/go-go.c (__go_go) [!LINKER_SUPPORTS_SPLIT_STACK &&
> __sun__ && __svr4__]: Reduce default thread stack size.
> diff -r 72fce0443d71 libgo/runtime/go-go.c
> --- a/libgo/runtime/go-go.c Sat Jan 08 12:30:04 2011 +0100
> +++ b/libgo/runtime/go-go.c Sun Jan 09 12:31:06 2011 +0100
> @@ -196,7 +196,21 @@
> #endif
> i = pthread_attr_setstacksize (&attr, PTHREAD_STACK_MIN);
> __go_assert (i == 0);
> +
> +#else /* !LINKER_SUPPORTS_SPLIT_STACK */
> +
> +#if defined __sun__ && defined __svr4__
> + /* Reduce the thread stack size from the Solaris 2 default of 1 MB
> + (32-bit) or 2 MB (64-bit) to allow go.test/test/chan/goroutines.go to
> + work. The value is just a guess. */
> +#ifndef PTHREAD_STACK_MIN
> +#define PTHREAD_STACK_MIN (128*1024)
> #endif
> + i = pthread_attr_setstacksize (&attr, PTHREAD_STACK_MIN);
> + __go_assert (i == 0);
> +#endif /* __sun__ && __svr4__ */
> +
> +#endif /* !LINKER_SUPPORTS_SPLIT_STACK */
I don't think this is the way to go. What this does is reduce the
default stack size of a thread in order to permit more threads to run.
But the effect will be that more threads will overrun their stacks and
crash. I think it's more likely that a thread will need a lot of stack
than it is that a Go program will need a lot of goroutines. The
goroutines.go test creates 10000 goroutines. It should work, of course,
but I don't think we should make it work by sacrificing the size of the
stack for ordinary goroutines. We should instead let goroutines.go fail
until split stack is implemented.
Ian