Xen devel - Dec 2011 - Stubdom build problem in 4.1.2 but not in xen-unstable

Hello,

I''ve found the following problem while building Xen 4.1.2 stubdoms:

gcc -DCONFIG_QEMU -mno-red-zone -O1 -fno-omit-frame-pointer
-fno-optimize-sibling-calls  -m64 -mno-red-zone -fno-reorder-blocks
-fno-asynchronous-unwind-tables -m64 -g -fno-strict-aliasing
-std=gnu99 -Wall -Wstrict-prototypes -Wno-unused-value
-Wdeclaration-after-statement -Wno-unused-but-set-variable  -nopie
-fno-stack-protector -fno-exceptions -fno-builtin -Wall -Werror
-Wredundant-decls -Wno-format -Wno-redundant-decls
-fno-stack-protector -fgnu89-inline -Wstrict-prototypes
-Wnested-externs -Wpointer-arith -Winline -g -DGNT_DEBUG
-DGNTMAP_DEBUG -D__INSIDE_MINIOS__ -m64 -mno-red-zone
-fno-reorder-blocks -fno-asynchronous-unwind-tables -Os
-fomit-frame-pointer -isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/include
-D__MINIOS__ -DHAVE_LIBC -isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/include/posix
-isystem /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../tools/xenstore
 -isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/include/x86
-isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/include/x86/x86_64
-U __linux__ -U __FreeBSD__ -U __sun__ -nostdinc -isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/include/posix
-isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/cross-root-x86_64/x86_64-xen-elf/include
-isystem /usr/lib/gcc/x86_64-alpine-linux-uclibc/4.6.2/include
-isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/include
-isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/include/ipv4
-I/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/include
-I/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../xen/include
-isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/mini-os/include
-D__MINIOS__ -DHAVE_LIBC -isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/mini-os/include/posix
-isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../tools/xenstore
-DHAVE_LWIP -isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/include
-isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/include/ipv4
-D__XEN_INTERFACE_VERSION__=0x00030205  -isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/mini-os/include/x86
-isystem
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/mini-os/include/x86/x86_64
-c console/xencons_ring.c -o
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/mini-os-x86_64-ioemu/console/xencons_ring.o
console/xencons_ring.c: In function
''xencons_ring_send_no_notify'':
console/xencons_ring.c:26:41: error: inlining failed in call to
''xencons_interface'': call is unlikely and code size would grow
[-Werror=inline]
console/xencons_ring.c:38:18: error: called from here [-Werror=inline]
cc1: all warnings being treated as errors

Although I can build xen-unstable without any problems on the same
machine. I''ve been trying to find the patch that solved this on
unstable, but so far I haven''t found it. Anyone know which patch fixed
it? I would like to have it backported to xen-4.1.

Info about the system:

Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-alpine-linux-uclibc/4.6.2/lto-wrapper
Target: x86_64-alpine-linux-uclibc
Configured with:
/home/buildozer/aports/main/gcc/src/gcc-4.6.2/configure --prefix=/usr
--mandir=/usr/share/man --infodir=/usr/share/info
--build=x86_64-alpine-linux-uclibc --host=x86_64-alpine-linux-uclibc
--target=x86_64-alpine-linux-uclibc --with-pkgversion=''Alpine
4.6.2-r1'' --disable-altivec --disable-checking --disable-fixed-point
--disable-libssp --disable-libstdcxx-pch --disable-multilib
--disable-nls --disable-werror --enable-__cxa_atexit --enable-cld
--enable-esp --enable-cloog-backend
--enable-languages=c,c++,objc,java,go --enable-shared
--enable-target-optspace --enable-tls --enable-threads
--with-dynamic-linker=ld64-uClibc.so.0.9.32
--with-dynamic-linker-prefix=/lib --with-system-zlib
--without-system-libunwind
Thread model: posix
gcc version 4.6.2 (Alpine 4.6.2-r1)

Thanks, Roger!

>>> On 22.12.11 at 18:03, Roger Pau
Monné<roger.pau@entel.upc.edu> wrote:
> Hello,
> 
> I've found the following problem while building Xen 4.1.2 stubdoms:
> 
> gcc -DCONFIG_QEMU -mno-red-zone -O1 -fno-omit-frame-pointer
> -fno-optimize-sibling-calls  -m64 -mno-red-zone -fno-reorder-blocks
> -fno-asynchronous-unwind-tables -m64 -g -fno-strict-aliasing
> -std=gnu99 -Wall -Wstrict-prototypes -Wno-unused-value
> -Wdeclaration-after-statement -Wno-unused-but-set-variable  -nopie
> -fno-stack-protector -fno-exceptions -fno-builtin -Wall -Werror
> -Wredundant-decls -Wno-format -Wno-redundant-decls
> -fno-stack-protector -fgnu89-inline -Wstrict-prototypes
> -Wnested-externs -Wpointer-arith -Winline -g -DGNT_DEBUG
> -DGNTMAP_DEBUG -D__INSIDE_MINIOS__ -m64 -mno-red-zone
> -fno-reorder-blocks -fno-asynchronous-unwind-tables -Os
> -fomit-frame-pointer -isystem
>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/includ
> e
> -D__MINIOS__ -DHAVE_LIBC -isystem
>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/includ
> e/posix
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../tools/xenstore
>  -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/inc
> lude/x86
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/inc
> lude/x86/x86_64
> -U __linux__ -U __FreeBSD__ -U __sun__ -nostdinc -isystem
>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/includ
> e/posix
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/cross-root-x86_64/x8
> 6_64-xen-elf/include
> -isystem /usr/lib/gcc/x86_64-alpine-linux-uclibc/4.6.2/include
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/inclu
> de
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/inclu
> de/ipv4
> -I/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/include
> -I/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../xen/include
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/m
> ini-os/include
> -D__MINIOS__ -DHAVE_LIBC -isystem
>
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/mini-
> os/include/posix
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../tools/xe
> nstore
> -DHAVE_LWIP -isystem
>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/include
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/inclu
> de/ipv4
> -D__XEN_INTERFACE_VERSION__=0x00030205  -isystem
>
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/mini-
> os/include/x86
> -isystem 
> /home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/m
> ini-os/include/x86/x86_64
> -c console/xencons_ring.c -o
>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/mini-os-x86_64-ioemu/conso
> le/xencons_ring.o
> console/xencons_ring.c: In function 'xencons_ring_send_no_notify':
> console/xencons_ring.c:26:41: error: inlining failed in call to
> 'xencons_interface': call is unlikely and code size would grow
> [-Werror=inline]
> console/xencons_ring.c:38:18: error: called from here [-Werror=inline]
> cc1: all warnings being treated as errors
> 
> Although I can build xen-unstable without any problems on the same
> machine. I've been trying to find the patch that solved this on
> unstable, but so far I haven't found it. Anyone know which patch fixed
> it? I would like to have it backported to xen-4.1.
It may well be that there is no such changeset, but the compiler just
decides to evaluate the likelihoods differently. I'm assuming that the
build problem would go away by removing -Winline from
extras/mini-os/minios.mk:DEF_CFLAGS (which is questionable in
conjunction with the -Werror forced there too - a more reasonably
approach would seem to be -Winline -Wno-error=inline, apparently
available since gcc 4.2.x).

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

2011/12/23 Jan Beulich <JBeulich@suse.com>:
>>>> On 22.12.11 at 18:03, Roger Pau
Monné<roger.pau@entel.upc.edu> wrote:
>> Hello,
>>
>> I've found the following problem while building Xen 4.1.2 stubdoms:
>>
>> gcc -DCONFIG_QEMU -mno-red-zone -O1 -fno-omit-frame-pointer
>> -fno-optimize-sibling-calls  -m64 -mno-red-zone -fno-reorder-blocks
>> -fno-asynchronous-unwind-tables -m64 -g -fno-strict-aliasing
>> -std=gnu99 -Wall -Wstrict-prototypes -Wno-unused-value
>> -Wdeclaration-after-statement -Wno-unused-but-set-variable  -nopie
>> -fno-stack-protector -fno-exceptions -fno-builtin -Wall -Werror
>> -Wredundant-decls -Wno-format -Wno-redundant-decls
>> -fno-stack-protector -fgnu89-inline -Wstrict-prototypes
>> -Wnested-externs -Wpointer-arith -Winline -g -DGNT_DEBUG
>> -DGNTMAP_DEBUG -D__INSIDE_MINIOS__ -m64 -mno-red-zone
>> -fno-reorder-blocks -fno-asynchronous-unwind-tables -Os
>> -fomit-frame-pointer -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/includ
>> e
>> -D__MINIOS__ -DHAVE_LIBC -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/includ
>> e/posix
>> -isystem
>> /home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../tools/xenstore
>>  -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/inc
>> lude/x86
>> -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/inc
>> lude/x86/x86_64
>> -U __linux__ -U __FreeBSD__ -U __sun__ -nostdinc -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../extras/mini-os/includ
>> e/posix
>> -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/cross-root-x86_64/x8
>> 6_64-xen-elf/include
>> -isystem /usr/lib/gcc/x86_64-alpine-linux-uclibc/4.6.2/include
>> -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/inclu
>> de
>> -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/inclu
>> de/ipv4
>> -I/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/include
>> -I/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/../xen/include
>> -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/m
>> ini-os/include
>> -D__MINIOS__ -DHAVE_LIBC -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/mini-
>> os/include/posix
>> -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../tools/xe
>> nstore
>> -DHAVE_LWIP -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/include
>> -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/lwip-x86_64/src/inclu
>> de/ipv4
>> -D__XEN_INTERFACE_VERSION__=0x00030205  -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/mini-
>> os/include/x86
>> -isystem
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/extras/mini-os/../../extras/m
>> ini-os/include/x86/x86_64
>> -c console/xencons_ring.c -o
>>
/home/royger/aports/testing/xen/src/xen-4.1.2/stubdom/mini-os-x86_64-ioemu/conso
>> le/xencons_ring.o
>> console/xencons_ring.c: In function
'xencons_ring_send_no_notify':
>> console/xencons_ring.c:26:41: error: inlining failed in call to
>> 'xencons_interface': call is unlikely and code size would grow
>> [-Werror=inline]
>> console/xencons_ring.c:38:18: error: called from here [-Werror=inline]
>> cc1: all warnings being treated as errors
>>
>> Although I can build xen-unstable without any problems on the same
>> machine. I've been trying to find the patch that solved this on
>> unstable, but so far I haven't found it. Anyone know which patch
fixed
>> it? I would like to have it backported to xen-4.1.
>
> It may well be that there is no such changeset, but the compiler just
> decides to evaluate the likelihoods differently. I'm assuming that the
> build problem would go away by removing -Winline from
> extras/mini-os/minios.mk:DEF_CFLAGS (which is questionable in
> conjunction with the -Werror forced there too - a more reasonably
> approach would seem to be -Winline -Wno-error=inline, apparently
> available since gcc 4.2.x).
I've ended up removing -Werror from all the Makefiles:

find -name '*.mk' -o -name 'Make*' | xargs sed -i -e
's/-Werror//g'

Not the prettiest solution but it works...

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Fri, 2011-12-23 at 08:29 +0000, Jan Beulich wrote:
> >>> On 22.12.11 at 18:03, Roger Pau
Monné<roger.pau@entel.upc.edu> wrote:
> > console/xencons_ring.c: In function
'xencons_ring_send_no_notify':
> > console/xencons_ring.c:26:41: error: inlining failed in call to
> > 'xencons_interface': call is unlikely and code size would grow
> > [-Werror=inline]
> > console/xencons_ring.c:38:18: error: called from here [-Werror=inline]
> > cc1: all warnings being treated as errors
> > 
> > Although I can build xen-unstable without any problems on the same
> > machine. I've been trying to find the patch that solved this on
> > unstable, but so far I haven't found it. Anyone know which patch
fixed
> > it? I would like to have it backported to xen-4.1.
> 
> It may well be that there is no such changeset, but the compiler just
> decides to evaluate the likelihoods differently. I'm assuming that the
> build problem would go away by removing -Winline from
> extras/mini-os/minios.mk:DEF_CFLAGS (which is questionable in
> conjunction with the -Werror forced there too - a more reasonably
> approach would seem to be -Winline -Wno-error=inline, apparently
> available since gcc 4.2.x).
Although in this specific case (and likely in many cases) the use of
inline on xencons_interface seems unnecessary -- lets just leave it up
to the compiler.

I thought "inline" was just a hint so it seems odd that gcc would
error
out.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

>>> On 23.12.11 at 10:25, Ian Campbell <Ian.Campbell@citrix.com>
wrote:
> I thought "inline" was just a hint so it seems odd that gcc would
error
> out.
That''s what -Winline is for - to tell you that your intention to have a
function inlined wasn''t fulfilled. If one asks for this warning,
occasionally having the compiler emit it is the result. The question to
me is what the purpose was to enable that warning in the first place.

Jan

On Fri, 2011-12-23 at 09:32 +0000, Jan Beulich wrote:
> >>> On 23.12.11 at 10:25, Ian Campbell
<Ian.Campbell@citrix.com> wrote:
> > I thought "inline" was just a hint so it seems odd that gcc
would error
> > out.
> 
> That''s what -Winline is for - to tell you that your intention to
have a
> function inlined wasn''t fulfilled. If one asks for this warning,
> occasionally having the compiler emit it is the result. The question to
> me is what the purpose was to enable that warning in the first place.
It seems to be only extras/mini-os which uses it and it came from the
first changeset, copy Samuel in case he remembers why.

Whatever the reason this specific instance is surely not worth force
inlining.

Linux uses __attribute__((always_inline)) and various similar
constructs, presumably to work around varying behaviours across gcc''s.

Ian.

Ian Campbell, le Fri 23 Dec 2011 09:43:17 +0000, a écrit
:
> On Fri, 2011-12-23 at 09:32 +0000, Jan Beulich wrote:
> > >>> On 23.12.11 at 10:25, Ian Campbell
<Ian.Campbell@citrix.com> wrote:
> > > I thought "inline" was just a hint so it seems odd that
gcc would error
> > > out.
> > 
> > That''s what -Winline is for - to tell you that your intention
to have a
> > function inlined wasn''t fulfilled. If one asks for this
warning,
> > occasionally having the compiler emit it is the result. The question
to
> > me is what the purpose was to enable that warning in the first place.
> 
> It seems to be only extras/mini-os which uses it and it came from the
> first changeset, copy Samuel in case he remembers why.
Well, I didn''t even know Xen at the time :)

Sun Jan 25 03:00:54 2004
From Kip Macy, we now are stricter when compiling the mini os.

So I don''t think there is any particular reason, and it''s
probably not
worth the trouble keeping it.

Samuel

On Fri, 2011-12-23 at 11:03 +0000, Samuel Thibault
wrote:
> Ian Campbell, le Fri 23 Dec 2011 09:43:17 +0000, a écrit :
> > On Fri, 2011-12-23 at 09:32 +0000, Jan Beulich wrote:
> > > >>> On 23.12.11 at 10:25, Ian Campbell
<Ian.Campbell@citrix.com> wrote:
> > > > I thought "inline" was just a hint so it seems odd
that gcc would error
> > > > out.
> > > 
> > > That's what -Winline is for - to tell you that your intention
to have a
> > > function inlined wasn't fulfilled. If one asks for this
warning,
> > > occasionally having the compiler emit it is the result. The
question to
> > > me is what the purpose was to enable that warning in the first
place.
> > 
> > It seems to be only extras/mini-os which uses it and it came from the
> > first changeset, copy Samuel in case he remembers why.
> 
> Well, I didn't even know Xen at the time :)
Sorry, I forgot that stubdoms != minios.
> 
> Sun Jan 25 03:00:54 2004
> From Kip Macy, we now are stricter when compiling the mini os.
> 
> So I don't think there is any particular reason, and it's probably
not
> worth the trouble keeping it.
> 
> Samuel


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel