Re: [PATCH 1/1] guile: workaround ice ssa corruption while DEBUG_BUILD

[Martin Jansa <martin.jansa@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 5621 bytes --]

On Sun, Aug 10, 2014 at 09:47:12PM +0800, Hongxu Jia wrote:
> Add:
> MACHINE = "qemux86-64"
> DEBUG_BUILD = "1"
> to local.conf and there was a ssa corruption to build guile:
> ..
> Unable to coalesce ssa_names 48 and 3476 which are marked as MUST COALESCE.
> sp_48(ab) and  sp_3476(ab)
> guile-2.0.11/libguile/vm-engine.c: In function 'vm_debug_engine':
> guile-2.0.11/libguile/vm.c:673:19: internal compiler error: SSA corruption
>  #define VM_NAME   vm_debug_engine
>                    ^
> guile-2.0.11/libguile/vm-engine.c:39:1: note: in expansion of macro 'VM_NAME'
>  VM_NAME (SCM vm, SCM program, SCM *argv, int nargs)
>  ^
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <http://gcc.gnu.org/bugs.html> for instructions.
> ...
> 
> Tweak libguile/vm-i-system.c to add boundary value check to workaround it.
> 
> Also fix qa issue while DEBUG_BUILD:
> ...
> WARNING: QA Issue: guile rdepends on ncurses-libncurses but its not a build dependency? [build-deps]
> WARNING: QA Issue: guile rdepends on readline but its not a build dependency? [build-deps]
> ...
> 
> [YOCTO #6411]
> 
> Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
> ---
>  .../files/workaround-ice-ssa-corruption.patch      | 52 ++++++++++++++++++++++
>  meta/recipes-devtools/guile/guile_2.0.11.bb        |  8 +++-
>  2 files changed, 59 insertions(+), 1 deletion(-)
>  create mode 100644 meta/recipes-devtools/guile/files/workaround-ice-ssa-corruption.patch
> 
> diff --git a/meta/recipes-devtools/guile/files/workaround-ice-ssa-corruption.patch b/meta/recipes-devtools/guile/files/workaround-ice-ssa-corruption.patch
> new file mode 100644
> index 0000000..1363d63
> --- /dev/null
> +++ b/meta/recipes-devtools/guile/files/workaround-ice-ssa-corruption.patch
> @@ -0,0 +1,52 @@
> +libguile/vm-i-system.c: workaround ice ssa corruption while compiling with option -g -O
> +
> +While compiling with option -g -O, there was a ssa corruption:
> +..
> +Unable to coalesce ssa_names 48 and 3476 which are marked as MUST COALESCE.
> +sp_48(ab) and  sp_3476(ab)
> +guile-2.0.11/libguile/vm-engine.c: In function 'vm_debug_engine':
> +guile-2.0.11/libguile/vm.c:673:19: internal compiler error: SSA corruption
> + #define VM_NAME   vm_debug_engine
> +                   ^
> +guile-2.0.11/libguile/vm-engine.c:39:1: note: in expansion of macro 'VM_NAME'
> + VM_NAME (SCM vm, SCM program, SCM *argv, int nargs)
> + ^
> +Please submit a full bug report,
> +with preprocessed source if appropriate.
> +See <http://gcc.gnu.org/bugs.html> for instructions.
> +...
> +
> +Tweak libguile/vm-i-system.c to add boundary value check to workaround it.
> +
> +Upstream-Status: Pending
> +
> +Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
> +---
> + libguile/vm-i-system.c | 11 +++++++----
> + 1 file changed, 7 insertions(+), 4 deletions(-)
> +
> +diff --git a/libguile/vm-i-system.c b/libguile/vm-i-system.c
> +index 5057fb0..8d31214 100644
> +--- a/libguile/vm-i-system.c
> ++++ b/libguile/vm-i-system.c
> +@@ -625,10 +625,13 @@ VM_DEFINE_INSTRUCTION (47, bind_optionals_shuffle, "bind-optionals/shuffle", 6,
> +   /* now shuffle up, from walk to ntotal */
> +   {
> +     scm_t_ptrdiff nshuf = sp - walk + 1, i;
> +-    sp = (fp - 1) + ntotal + nshuf;
> +-    CHECK_OVERFLOW ();
> +-    for (i = 0; i < nshuf; i++)
> +-      sp[-i] = walk[nshuf-i-1];
> ++    if (nshuf > 0)
> ++    {
> ++      sp = (fp - 1) + ntotal + nshuf;
> ++      CHECK_OVERFLOW ();
> ++      for (i = 0; i < nshuf; i++)
> ++        sp[-i] = walk[nshuf-i-1];
> ++    }
> +   }
> +   /* and fill optionals & keyword args with SCM_UNDEFINED */
> +   while (walk <= (fp - 1) + ntotal)
> +-- 
> +1.9.1
> +
> diff --git a/meta/recipes-devtools/guile/guile_2.0.11.bb b/meta/recipes-devtools/guile/guile_2.0.11.bb
> index d19460a..e380bed 100644
> --- a/meta/recipes-devtools/guile/guile_2.0.11.bb
> +++ b/meta/recipes-devtools/guile/guile_2.0.11.bb
> @@ -21,6 +21,9 @@ SRC_URI = "${GNU_MIRROR}/guile/guile-${PV}.tar.xz \
>             file://arm_endianness.patch \
>             "
>  
> +SRC_URI_append_class-target = " \
> +           ${@base_conditional( "DEBUG_BUILD", "1", " file://workaround-ice-ssa-corruption.patch", "", d )} \
> +           "
>  #           file://debian/0001-Change-guile-to-guile-X.Y-for-info-pages.patch
>  #           file://opensuse/guile-turn-off-gc-test.patch 
>  
> @@ -35,7 +38,10 @@ DEPENDS = "libunistring bdwgc gmp libtool libffi"
>  # add guile-native only to the target recipe's DEPENDS
>  DEPENDS_append_class-target = " guile-native libatomics-ops"
>  
> -RDEPENDS_${PN}_append_libc-glibc_class-target = "glibc-gconv-iso8859-1"
> +RDEPENDS_${PN}_append_class-target = " \
> +             ${@base_conditional( "DEBUG_BUILD", "1", " readline ncurses", "", d )} \
> +             "

This seems to be needed even withotu DEBUG_BUILD, see
http://lists.openembedded.org/pipermail/openembedded-core/2014-August/095585.html
and IMHO it's build time dependncy not runtime.

> +RDEPENDS_${PN}_append_libc-glibc_class-target = " glibc-gconv-iso8859-1"
>  
>  EXTRA_OECONF += "${@['--without-libltdl-prefix --without-libgmp-prefix --without-libreadline-prefix', ''][bb.data.inherits_class('native',d)]}"
>  
> -- 
> 1.9.1
> 
> -- 
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 188 bytes --]