The alternative is to change the root makefile to call "$(MAKE) -C tools/ all" instead, but this way feels neater to me. -- Andrew Cooper - Dom0 Kernel Engineer, Citrix XenServer T: +44 (0)1223 225 900, http://www.citrix.com _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
Andrew Cooper writes ("[Xen-devel] tools/makefile: Add build
target"):> The alternative is to change the root makefile to call "$(MAKE) -C
> tools/ all" instead, but this way feels neater to me.
Yes.
> tools/makefile: Add build target
>
> The root Makefile has a build target which calls "$(MAKE) build"
in each of xen/
> tools/ stubdom/ and docs/, which fails because of the tools/ Makefile.
That this didn''t work is clearly a bug.
But when I tried it my build did this:
CC i386-stubdom/eeprom93xx.o
/u/iwj/work/xen-unstable-tools.hg/stubdom/mini-os-x86_32-c/test.o: In function
`app_main'':
/u/iwj/work/xen-unstable-tools.hg/extras/mini-os/test.c:441: multiple definition
of `app_main''
/u/iwj/work/xen-unstable-tools.hg/stubdom/mini-os-x86_32-c/main.o:/u/iwj/work/xen-unstable-tools.hg/extras/mini-os/main.c:187:
first defined here
CC i386-stubdom/eepro100.o
make[2]: ***
[/u/iwj/work/xen-unstable-tools.hg/stubdom/mini-os-x86_32-c/mini-os] Error 1
make[2]: Leaving directory
`/u/iwj/work/xen-unstable-tools.hg/extras/mini-os''
make[1]: *** [c-stubdom] Error 2
make[1]: *** Waiting for unfinished jobs....
That''s the result of:
(make -j4 build && echo ok.) 2>&1 | tee ../log
Ian.
On 07/08/12 19:27, Ian Jackson wrote:> Andrew Cooper writes ("[Xen-devel] tools/makefile: Add build target"): >> The alternative is to change the root makefile to call "$(MAKE) -C >> tools/ all" instead, but this way feels neater to me. > Yes. > >> tools/makefile: Add build target >> >> The root Makefile has a build target which calls "$(MAKE) build" in each of xen/ >> tools/ stubdom/ and docs/, which fails because of the tools/ Makefile. > That this didn''t work is clearly a bug. > > But when I tried it my build did this: > > CC i386-stubdom/eeprom93xx.o > /u/iwj/work/xen-unstable-tools.hg/stubdom/mini-os-x86_32-c/test.o: In function `app_main'': > /u/iwj/work/xen-unstable-tools.hg/extras/mini-os/test.c:441: multiple definition of `app_main'' > /u/iwj/work/xen-unstable-tools.hg/stubdom/mini-os-x86_32-c/main.o:/u/iwj/work/xen-unstable-tools.hg/extras/mini-os/main.c:187: first defined here > CC i386-stubdom/eepro100.o > make[2]: *** [/u/iwj/work/xen-unstable-tools.hg/stubdom/mini-os-x86_32-c/mini-os] Error 1 > make[2]: Leaving directory `/u/iwj/work/xen-unstable-tools.hg/extras/mini-os'' > make[1]: *** [c-stubdom] Error 2 > make[1]: *** Waiting for unfinished jobs.... > > That''s the result of: > > (make -j4 build && echo ok.) 2>&1 | tee ../log > > Ian.The issue there is that app_main() is defined as __attribute__((weak)), and has two definitions in the code. I have just tested and confirmed that minios (and therefore stubdom) is not -j safe. As a result, I would argue that this build failure is not nesseserally a barrier of entry to the patch itself; There are more issues which need fixing as well. stubdom itself has further build issues regarding relative paths to configure scripts, which I was going to around to fixing after some of my more important tasks. -- Andrew Cooper - Dom0 Kernel Engineer, Citrix XenServer T: +44 (0)1223 225 900, http://www.citrix.com
Andrew Cooper writes ("Re: [Xen-devel] tools/makefile: Add build
target"):> The issue there is that app_main() is defined as __attribute__((weak)),
> and has two definitions in the code.
>
> I have just tested and confirmed that minios (and therefore stubdom) is
> not -j safe. As a result, I would argue that this build failure is not
> nesseserally a barrier of entry to the patch itself; There are more
> issues which need fixing as well.
After struggling with testing a bit, -j appears to be a red herring.
The real bug is that if you say "make -C stubdom build" you end up
trying to run the "c-stubdom" target. However that target (some kind
of test application?) is what breaks. It''s not built or used by
"make -C stubdom install" which is what you get if you say
"make" at
the toplevel.
See patch below.
With that and your patch "make build" (and "make build -j4")
works for
me. I don''t understand how it can have worked for you with just your
original patch; without mine, "make build" (without -j) fails in the
same way as I quote above.
Can you go into a bit more detail about the -j bug(s) in minios ? I''d
be happy to take patches to add ".NOTPARALLEL:" in appropriate places.
> stubdom itself has further build issues regarding relative paths to
> configure scripts, which I was going to around to fixing after some of
> my more important tasks.
Fixes welcome of course.
Ian.
# HG changeset patch
# Parent bcacd62f0460bd461f0df477e41f45144e19f2e8
stubdom: do not build "c-stubdom"
This does not build:
/u/iwj/work/xen-unstable-tools.hg/stubdom/mini-os-x86_32-c/test.o: In function
`app_main'':
/u/iwj/work/xen-unstable-tools.hg/extras/mini-os/test.c:441: multiple
definition of `app_main''
/u/iwj/work/xen-unstable-tools.hg/stubdom/mini-os-x86_32-c/main.o:/u/iwj/work/xen-unstable-tools.hg/extras/mini-os/main.c:187:
first defined here
It''s not built by the default toplevel "make all", which is
how this
has gone unnoticed. c-stubdom appears to be some kind of test
application; in any case it is not currently used.
Fixing this means that the contents of the stubdom build: target are
aligned with the contents of the install: target. Now
"make -C stubdom build" only builds the things that would be built
(for installation) by "make -C stubdom install".
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
diff -r bcacd62f0460 stubdom/Makefile
--- a/stubdom/Makefile Tue Aug 07 19:14:31 2012 +0100
+++ b/stubdom/Makefile Thu Aug 09 17:42:23 2012 +0100
@@ -81,7 +81,7 @@ CROSS_MAKE := $(MAKE) DESTDIR .PHONY: all
all: build
ifeq ($(STUBDOM_SUPPORTED),1)
-build: genpath ioemu-stubdom c-stubdom pv-grub xenstore-stubdom
+build: genpath ioemu-stubdom pv-grub xenstore-stubdom
else
build: genpath
endif