Building of libopcode.so and gold can go on in parallel currently
and I am running into the issue that libopcode.so is being written
while it is opened by the host assembler (which is directed to
the libopcode build library directory via the toplevel machinery
setting HOST_LIB_PATH) - resulting in the dynamic loader complaining
that libopcode is corrupt.
This can be avoided by making all builds depend on libopcode - which
the following patch does for gold. That fixes my spurious build
failures (on fast machines).
Ok? Ian, can you take care of patching src?
Thanks,
Richard.
2011-07-01 Richard Guenther <rguent...@suse.de>
* Makefile.def (all-gold): Depend on all-opcodes.
* Makefile.in: Rebuild.
Index: Makefile.def
===================================================================
--- Makefile.def (revision 175749)
+++ Makefile.def (working copy)
@@ -402,6 +402,7 @@ dependencies = { module=configure-gold;
dependencies = { module=all-gold; on=all-libiberty; };
dependencies = { module=all-gold; on=all-intl; };
dependencies = { module=all-gold; on=all-bfd; };
+dependencies = { module=all-gold; on=all-opcodes; };
dependencies = { module=all-gold; on=all-build-bison; };
dependencies = { module=check-gold; on=all-binutils; };
dependencies = { module=check-gold; on=all-gas; };
Index: Makefile.in
===================================================================
--- Makefile.in (revision 175749)
+++ Makefile.in (working copy)
@@ -42822,6 +42822,14 @@ all-stage3-gold: maybe-all-stage3-bfd
all-stage4-gold: maybe-all-stage4-bfd
all-stageprofile-gold: maybe-all-stageprofile-bfd
all-stagefeedback-gold: maybe-all-stagefeedback-bfd
+all-gold: maybe-all-opcodes
+
+all-stage1-gold: maybe-all-stage1-opcodes
+all-stage2-gold: maybe-all-stage2-opcodes
+all-stage3-gold: maybe-all-stage3-opcodes
+all-stage4-gold: maybe-all-stage4-opcodes
+all-stageprofile-gold: maybe-all-stageprofile-opcodes
+all-stagefeedback-gold: maybe-all-stagefeedback-opcodes
all-gold: maybe-all-build-bison
all-stage1-gold: maybe-all-build-bison
